JP2009032118A - Information structuring device, information structuring method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information structuring device and an information structuring method and a program, for arranging enterprise information, product information and technical information by classifying and structuring applied information sources. <P>SOLUTION: This information structuring device 1 is provided with: an information source cutting part 100 for dividing information sources into indivisible elements in a grammar level; an information source reconstruction part 200 for forming an information source tree by combining the indivisible elements into a tree structure; and an information source discrimination part 300 for discriminating an adapted information source tree related with producer goods from an unadapted information source tree unrelated with producer goods by referring to a business domain dictionary in which business domains are registered. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an information structuring apparatus, an information structuring method, and a program for structuring an input information source according to a predetermined rule, and more particularly, for example, an information structuring apparatus that supports product planning using web information as an information source. The present invention relates to an information structuring method and a program.

  When a company plans a product for a new product, it has traditionally taken into account market trends, technologies owned by other companies, and proprietary technologies. Specifically, it is usually considered whether the market of the business domain is growing, what products are selling, or how to make use of the technology in-house that does not exist in other companies. It is.

  In the case of product planning for products (consumer goods) purchased by general consumers, the characteristics of the products are often determined by questionnaire-type marketing for general consumers. In recent years, consumer-generated media / consumer generation in which consumers use the Internet, such as word-of-mouth sites, Q & A communities, social networking services (SNS), blogs, COI (Community Of Interest) sites, etc. An attempt is being made to collect word-of-mouth information using media (Consumer Generated Media: CGM).

  On the other hand, in the case of product planning of products (production goods) purchased by a company, it is difficult to perform marketing based on customer feedback. The reason for this is that companies purchase production goods and the purchaser population is small, the types of production goods to be purchased, and the number of purchases are information related to the investment strategy of the purchasing company. It is because it is not made public.

  By the way, Patent Document 1 discloses an information extraction system that analyzes a sentence described in a natural language, extracts information in a predetermined field, and outputs information including a relationship between words included in the sentence in a certain format. It is disclosed. In the technique described in Patent Document 1, the morpheme analysis unit divides an input sentence into morpheme units, and then assigns syntax information and keyword information related to the field of information to be extracted to each morpheme. The parsing rule storage unit stores rules for parsing input sentences using syntax information, and the keyword relation calculation rule storage unit generates semantic structures indicating keyword relationships while controlling parsing rules based on keyword information. The document information extraction unit analyzes the morpheme sequence output from the morpheme analysis unit using the parsing rules and the relationship rules between keywords, and outputs the meaning consisting of the keywords and the relationship between keywords. It is.

Further, in Patent Document 2, for example, for the purpose of effectively distributing advertisements for sales promotion of products and services, a specific product or service is selected from a population of customers to be promoted. A customer screening method for screening potential customers who are considered highly likely to actually purchase the product is disclosed. The technology described in Patent Document 2 is an information classification device that automatically classifies a file containing text, and morphologically analyzes the text portion of the file output by the file storage means and outputs it together with the file identifier. A keyword storage means for storing a keyword for classifying a file, a keyword storage means, an output of the keyword storage means and an output of the morpheme analysis means, and collecting keywords from the result of the morpheme analysis and outputting them together with the file identifier Keyword collecting means, user input means for requesting a classification result different from the recent classification result, output of the input means and output of the keyword collecting means as input, and classifying the file identifiers by keyword Information classification means, information classification means output and file storage means output as inputs. And it provides a result of classification means classifies the user.
JP-A-7-85071 Japanese Patent Laid-Open No. 10-283366

  However, in the technique described in Patent Document 1 described above, information is extracted based on a keyword dictionary, and in the technique described in Patent Document 2, input information is classified based on keywords. However, if it is not registered as a keyword, information cannot be extracted. Therefore, even if a word of the same type as a certain keyword is included, if the word does not match the keyword, it is not extracted, so there is a high possibility that necessary information will be leaked. In addition, when adding a new competitive product or its feature word, there is a problem that maintenance must be performed manually, and management is complicated.

  The present invention has been made to solve such problems, and an information structuring apparatus, an information structuring method, and an information structuring method that can accurately extract necessary information from a given information source and classify it. The purpose is to provide a program.

  In the information structuring apparatus according to the present invention, an information source cutting unit that divides an information source to generate a word, an information source reconstruction unit that classifies the word according to a predetermined format rule, and a business domain are registered. An information source selection section for selecting whether the information source is a conforming information source related to a production good or a non-conforming information source not related to a production good, and a technical term and its synonyms By referring to a technical dictionary in which the set is classified based on the superordinate concept and / or subordinate concept in the business domain, an information source having a high probability of producing goods is extracted from the non-conforming information sources, and is used as the conforming information source. A group generation unit.

  An information structuring method according to the present invention is a method for executing the above information structuring process, and a program according to the present invention causes a computer to execute the above information structuring process.

  According to the present invention, it is possible to provide an information structuring apparatus, an information structuring method, and a program capable of organizing company information, product information, and technical information by classifying and structuring given information sources. it can.

  Hereinafter, specific examples to which the present invention is applied will be described in detail with reference to the drawings. The information structuring apparatus according to the present embodiment is an apparatus for organizing corporate information, product information, and technical information on the web, etc., and providing product planning support. Judging from the desire to invest, discover companies with more purchasing power. In addition, it is possible to select only production goods purchased by a trusted other company by adding information on delivery results to other companies to the product and compare products based on the purchaser's performance. . Furthermore, by discovering new customers and new parts suppliers from published information such as news, it is possible to expand sales channels and parts supply, thereby eliminating bottlenecks.

  Here, in this embodiment, for the sake of explanation, a company that manufactures parts is a parts delivery company, a company that delivers parts from a parts delivery company is a competitor, and a company that delivers products is a customer company. explain. A description will be given assuming that a user who is a competitor uses this system. Needless to say, the customer companies and the parts supplier companies are competitors, but in this embodiment, they are a parts supplier company and a customer company.

First Embodiment of the Invention FIG. 1 is a diagram showing an information structuring apparatus 1 according to a first embodiment of the present invention. As shown in FIG. 1, the information structuring apparatus 1 according to the present embodiment classifies and structures a given information source, and more specifically, as a news title expressed in a single sentence. Information sources are structured using division / reconstruction rules, and past structured information sources are used to organize company information, product information, and technical information. The information structuring apparatus 1 includes an information source cutting unit 100, an information source reconstruction unit 200, an information source selection unit 300, and a group generation unit 400. The information source cutting unit 100 divides the character string of the information source into inseparable elements at the grammar level. The information source reconstruction unit 200 generates element words by combining inseparable elements. The information source selection unit 300 selects information sources that match the business domain from the information sources. The group generation unit 400 selects similar information sources from a technical viewpoint. Each block will be described in detail below.

  FIG. 2 is a block diagram showing the information source cutting unit 100 in this embodiment. As illustrated in FIG. 2, the information source cutting unit 100 includes an information source (title) storage unit 101, a morpheme analysis unit 102, and a morpheme storage unit 103. Each of these means generally operates as follows.

  The information source storage unit 101 holds an information source character string. As shown in FIG. 3, the information source character string includes date and time and character string information. This information source character string is a short sentence having a plurality of “,” and one “.”.

  The morpheme analyzing unit 102 converts a character string expressed in one sentence such as a title character string “NEC, optical access product“ DW4000 ”to be released”, which is an information source, into a morpheme (“NEC”, ",", "Light", "access", "product", "" "," DW "," 4000 "," "" "," "," released ") and morpheme analysis. Is the smallest character string that is meaningless if it is made finer than this.The morphological analysis means knowledge of the target language grammar (gathering of grammar rules) and dictionary (word list with information such as parts of speech). Used to divide a sentence written in natural language into morpheme strings (segment processing (split writing)) and identify each part of speech.The result of morpheme analysis is, for example, as shown in FIG. , Stored in the morpheme storage unit 103 in a data format consisting of words, parts of speech, assistance, and reading It is.

  FIG. 5 is a diagram illustrating the information source reconstruction unit 200 in the present embodiment. As illustrated in FIG. 5, the information source reconstruction unit 200 includes a format analysis unit 201, a format rule storage unit 202, an element word storage unit 203, and an information source meta information storage unit 206. Each of these means generally operates as follows.

The format analysis unit 201 receives the part of speech set obtained by dividing the information source into morphemes from the morpheme storage unit 103. Consecutive nouns are combined in this part-of-speech set. In the above example, for example, “optical”, “access”, and “product” are connected to form an “optical access product”. Thereafter, by referring to the format rule storage unit 202, a rule set that matches the expression specific to the information source is extracted, and keywords such as the subject, action, product name, and feature word are extracted, and the product name is extracted from the morpheme set. A subject, action, feature word, and synonym (element word) are generated and stored in the element word storage unit 203. The element words stored in the element word storage unit 203 are stored in the information source meta information storage unit 206 for each information source.
If you follow the formatting rules below,
Subject: NEC Product Name: DW4000
Action: Release feature word: Generate element word of Hikari Access product.

The rules stored in the format rule storage unit 202 are as follows.
1) Product / derived product extraction rules (a) Extract consecutive nouns enclosed in angle brackets “” as product names (a) If there are English letters + numbers, extract them as product model numbers (c) Mark (R) Extract the continuous nouns that come before the product name as the trade name (d) Take the difference from the product name extracted in the past, and extract the difference part as the derived product name (e) If there is no corresponding, there is no product name Is not extracted

2) Subject extraction rule (a) When only nouns appear in the range from the beginning of a line to a punctuation point (,), the noun with the longest match is the company name. (A) Case particles ("GA", " The name of the longest matching noun that comes before “)” is the company name (C) Co., Ltd., and the longest matching noun that comes immediately after (Yes) is the company name. Extract company name from source URL

3) Action extraction rules (a) The last variable noun is a company action (b) The last appearing verb is a company action (c) If there is no such thing, the action of announcement is extracted

4) Action target rule (a) When the action is “delivery”, the noun that comes before the particles “he” and “ni” is the delivery company. (B) When the action is “recruitment”, The extracted company is the delivery company.
(C) If there is no applicable item, it is determined that there is no action target and is not extracted.

5) Synonym extraction rule (a) A set of consecutive nouns before and after the parenthesis “(” is extracted as a keyword and its synonym.
(B) If there is no applicable one, it is determined that there is no synonym and is not extracted.

6) Feature word extraction rule (a) A continuous noun leaked in another rule set is extracted as a feature word.

  FIG. 6 is a block diagram showing the information source selection unit 300 in this embodiment. As illustrated in FIG. 6, the information source selection unit 300 includes an information source selection unit 301, a domain dictionary storage unit 302, a compatible information source storage unit 303, and a non-conforming information source storage unit 304. Each of these means generally operates as follows.

  The information source selection unit 301 refers to the domain dictionary in the domain dictionary storage unit 302 and selects only information sources that match the business domain as matching information sources. Specifically, first, if there are words including actions such as “delivery” and “adoption” in the information source, it is determined that the information source is related to the production goods.

  Next, the information source selection unit 301 refers to the domain dictionary, and element words constituting the information source determined as the information source related to the production goods, that is, the subject, the product name, the action, the feature word, the synonym, and the domain dictionary The business domain specialized company name, product name, and feature word stored in the storage unit 302 are compared. FIG. 7 is a diagram showing the structure of the domain dictionary, and FIG. 8 is a diagram showing an example of the domain dictionary. In the domain dictionary, a plurality of word strings, such as a word / phrase column, a word / phrase two column,..., Are registered for each category such as a subject, a product name, and a feature word.

  Then, the information source selection unit 301 compares each element word of the information source determined to be an information source related to the production goods and a word in the domain dictionary, and if one or more matches are found, the information source is It is determined as a production good that matches the business domain, and is stored in the conforming information source storage unit 303. If there is no match, it is stored in the nonconforming information source storage unit 304. Then, among the “product name”, “feature word”, and “synonym” constituting the information source stored in the conformity information original storage unit 303, those not registered in the domain dictionary are stored in the domain dictionary of the domain dictionary storage unit 302. sign up. The “subject” that constitutes the information source measures the number of times each element word of the information source matches or does not match the domain dictionary in order to distinguish the specialist from the general manufacturer in the business domain. Regarding information sources of the same subject, if the number of information sources equal to or greater than a predetermined ratio th1 (0 <th1 <1) with respect to the total number of information sources is a conforming information source, the subject of the information source is designated as a specialist. If it is registered in the subject category of the domain dictionary and less than a predetermined ratio, it is registered in the subject category as a general manufacturer.

  FIG. 9 is a block diagram illustrating the group generation unit 400 in the present embodiment. As illustrated in FIG. 9, the group generation unit 400 includes an information source multiplication unit 401 and a technical dictionary storage unit 402. Each of these means generally operates as follows.

  The information source multiplication unit 401 obtains an information source that has not been determined as a business domain production good from the non-conforming information source storage unit 304, refers to the technical dictionary in the technical dictionary storage unit 402, and matches the business domain. Inspecting for non-conforming information sources is promoted to conforming information sources.

  The technical dictionary is a set of technical terms and synonyms classified based on the superordinate concepts and subordinate concepts in the business domain. For example, FIG. 10 is a technical dictionary based on the product configuration. In this technical dictionary, based on the module configuration for producing production goods, the technical terms related to parts are at the lowest node, the module part composed of parts is placed at the top, and the product composed of modules is the top. A hierarchical structure such as a higher node is defined. That is, “optical amplification”, “optical access”, “OC-12”, and “SRM-16” are defined at the bottom, and “DWDM” and “SONET / SDH” are defined at the top. Defines an “optical transmission device”. The distance between each node is also set when creating the technical dictionary. FIG. 11 is a diagram showing the structure of the technical dictionary. Terms that are used at the same time are set to have a short distance, and terms that are not used at the same time are set to have a long distance. For example, the distance between “optical amplification” and “optical access” is 0.6, and the distance between “optical amplification” and “SONET / SDH” is 2.3. Since production goods also have various domains, prepare items that match the business domain to be analyzed in advance.

  The nodes of the technical dictionary and the distance between the nodes may be freely defined in accordance with the business domain shown in FIG. 8 and FIG. 35 described later.

  By using this technical dictionary, even if a word of the same type as a certain keyword is included and the word does not match the keyword, it can be used as a compatible information source. Therefore, it is possible to reduce the possibility that necessary information is leaked, and to easily add new competitive products and their characteristic words.

  In the present embodiment, a technical dictionary will be described. However, any word dictionary in which a set of words and synonyms thereof is classified based on a higher concept and / or a lower concept in the business domain may be used. For example, when dealing with company news in the fresh food industry, the technical dictionary has a “food” at the top node, “meat” and “fish” at its lower nodes, and “pork” and “beef” at its lower nodes. It is also possible to make a tree dictionary having nodes such as “,“ chicken ”.

  First, the feature words of the product goods that match the business domain are acquired from the domain dictionary in the domain dictionary storage unit 302, and compared with the feature words of the non-conforming information source to calculate the similarity. The similarity is calculated based on the above technical dictionary, and if the distance between the non-conforming information source and the technical dictionary is equal to or less than a predetermined threshold, the non-conforming information source is determined to be information related to the production goods in the business domain and is used as the conforming information source. The information is stored in the compatible information source storage unit 303. That is, a non-conforming information source that has a high probability of conforming to the business domain is extracted and re-registered in the conforming information source storage unit 303. At this time, among the element words constituting the non-conforming information source promoted to the conforming information source, the technical dictionary is expanded by registering the word not registered in the technical dictionary in the technical dictionary. In addition, feature words and product names that are not registered in the domain dictionary storage unit 302 are newly added to the domain dictionary storage unit 302 to multiply the domain dictionary.

  The information source multiplication unit 401 can promote a non-conforming information source determined to be non-conforming using the domain dictionary to a conforming information source using the technical dictionary. Further, the information source multiplication unit 401 refers to an existing domain dictionary or technical dictionary, and a subject or a feature word included in the information source newly added to the matching information source storage unit 303 as a matching information source is an existing domain. Determine whether it is registered in a dictionary or technical dictionary. If it is not registered, newly registering it in the domain dictionary or technical dictionary can proliferate words that are not in the domain dictionary or technical dictionary using the element words included in the information source that matches the domain. it can.

  Conventionally, it is necessary to collect benchmarks, technology trends, market trends, etc. of competitor products in order to plan product for production goods. For this reason, conventionally, for example, it has been necessary to conduct an external external environmental survey in cooperation with a consultant. Moreover, as an information source useful for product planning, it is laborious to collect and collect homepages scattered on the Internet, interview information held by sales representatives, magazine articles, etc. without omission. Since it is a work that is continuously performed by a person, it depends heavily on the ability of the investigator and it is difficult to plan a product based on objective information. For this reason, in product planning for production goods, new products from competitors were released, but they seemed to be side-by-side products, and there was a problem that it was difficult to understand what differentiating factors were .

  In response to such a problem, the information structuring apparatus 1 according to the present embodiment can extract element words such as feature words from the information source, so what differentiating factors are retained by the new products of competitors. The above problem can be solved. In addition to referring to the business domain dictionary for information sources, information sources that are judged to be non-conforming information sources can be promoted to conforming information sources by referring to the technical dictionary, and there is no leakage from the information sources. Necessary information sources can be extracted.

  Next, the operation of this embodiment will be described. FIG. 12 is a flowchart illustrating the operation of the information structuring apparatus according to the present embodiment. This example is an example of detecting that an unknown competitive product has been released to the market. As a premise, “DW4000” is registered in the “product name” row and “word / phrase 1” column of the domain dictionary of the domain dictionary storage unit 302 so as to match the business domain. (See Figure 8)

As an information source, for example
"NEC releases optical access product" DW4000 ""
A case where the character string "" is input to the system will be described. The morpheme analysis unit 102 of the information source cutting unit 100 performs morpheme analysis of the information source, and decomposes the information source into morphemes to obtain the morpheme shown in FIG. (Step S100). This morpheme is stored in the morpheme storage unit 103.

  When the continuous noun appears in the morpheme set stored in the morpheme storage unit 103, the format analysis unit 201 combines them. (Step S200). FIG. 14 is a diagram showing the connected morphemes. Next, a format analysis rule is applied to determine whether the combined continuous noun is a subject, a product name, an action, or a feature word (step S300).

  Next, the format analysis rule application method will be described. The format analysis unit 201 combines the continuous nouns of the morpheme set, and extracts the subject, the product name, the action, the feature word, and the synonym from the morpheme or the continuous noun according to the format rule.

  First, the product name extraction method will be described. FIG. 15 is a flowchart showing a product name extraction method. When extracting the product name, as shown in FIG. 15, it is determined from the ordered morpheme set whether or not there is a key bracket (step S301). If there is a key bracket, the morpheme array in the key bracket is determined. Extract (step S302). Then, it is determined whether or not only nouns or symbols are present in the morpheme array in the brackets (step S303), and if there are consecutive alphanumeric characters (step S304), a noun in which letters and numbers are consecutive is determined. Combine (step S305). Further, it is determined whether or not a trademark symbol exists in the morpheme array in the brackets (step S306). If it exists, the trademark symbol is deleted, and the noun before the trademark symbol is used as the product name in the product name. The noun after the trademark symbol is registered and adopted as a feature word (step S307). It is determined whether or not the product name extracted in the above has been registered as a product name in the past (step S308), and if the product name has been registered in the past and there is a noun separated by a trademark in parentheses, It is registered in the feature word of the element word storage unit 203 as a derived product modifier (step S309), and if the product name has not been registered in the past, it is registered in the product name of the element word storage unit 203 as a new product name (step S309). S310), the product name extraction flow is terminated.

  FIG. 16 is a flowchart showing the subject extraction method. As shown in FIG. 16, in the ordered morpheme set, it is determined whether or not only the noun exists from the beginning of the line to the punctuation point (step S311). If only the noun exists, it is extracted as the subject (step S312).

  On the other hand, if it is not only a noun from the beginning of a line to a punctuation point, it is determined whether or not (stock) and (present) symbols exist (step S313). If it exists, the noun before and after the symbol is extracted as the subject (step S314). If the symbols “stock” and “present” do not exist, it is determined whether “has” or “has” exists (step S315). If it exists, the noun before the particle is extracted as the subject. Further, when neither “ga” nor “ha” exists, the sender is set as the subject (step S317). The extracted subject is recorded in the element word storage unit 203.

  FIG. 17 is a flowchart showing the behavior extraction method. As shown in FIG. 17, in the ordered morpheme set, it is determined whether or not there is a saun noun at the end of the line (step S321). If there is, a continuous noun including a sa strange noun is extracted as an action (step S322). If there is no sa variable noun, it is determined whether or not a verb is present (step S323). If a verb exists, the last appearing verb is extracted as an action (step S324). If the verb does not exist, the action “announcement” is added to the action of the element word storage unit 203 shown in FIG. 5 (step S325).

  FIG. 18 is a flowchart showing the action target extraction method. As shown in FIG. 18, it is determined whether or not the action is “delivery” in the ordered morpheme set (step S331). The action target extraction process is performed after the action extraction process. If the extracted action is “delivery”, the longest matching noun before “to” and “to” is extracted as an action target (step S332). If the action is not “delivery”, it is determined whether or not it is “adoption” (step S333). If “adopted”, the company of the subject is extracted as an action target (step S334). The extracted action is recorded in the element word storage unit 203. On the other hand, if the action is not “adopt”, the process is terminated (step S335).

  FIG. 19 is a flowchart showing a synonym extraction method. As shown in FIG. 19, it is determined whether or not parentheses are present in the ordered morpheme set (step S341). If parentheses are present, a set of consecutive nouns that precede and follow the right parenthesis A synonym is extracted (step S342). The extracted word source and its synonym are recorded in the element word storage unit 203. On the other hand, when there is no parenthesis, no synonym is recorded in the element word storage unit 203 (step S343).

  FIG. 20 is a flowchart showing the feature word extraction method. As shown in FIG. 20, it is determined whether there is a noun that cannot be classified by applying another rule in the ordered morpheme set (step S351). If there is, the noun is extracted as a feature word (step S352) and recorded in the element word storage unit 203.

As described above, the product name extracting step shown in FIG. 15, the subject extracting step shown in FIG. 16, the action extracting step shown in FIG. 17, the action target extracting step shown in FIG. 18, the synonym extracting step shown in FIG. Through the feature word extraction process indicated by (2), an analysis result as shown in FIG. 21 is obtained (step S300). In other words, from the information source “NEC releases optical access product“ DW4000 ””
Subject: NEC Product Name: DW4000
Action: Launch Feature word: The result of the analysis is an optical access product.

Next, a method in which the information source selection unit 301 selects a compatible information source from the information sources with reference to the domain dictionary will be specifically described. As an information source (news information source)
"Opto Sales Deliver" FX4200 "with Raman Method for Optical Amplification to NEE"
Is input to the information structuring apparatus 1. The information structuring apparatus 1 executes steps S100 to S300 in FIG. 12 and obtains the format analysis result shown in FIG.

  FIG. 23 is a flowchart showing a compatible information source selection method. The information source selection unit 301 collates the element words constituting the information source with the business domain dictionary. That is, as shown in FIG. 23, the information source selection unit 301 detects whether or not “delivery” and “adoption” are included in the action node from the information source (step S401). “Delivery” and “recruitment” are included, and the words constituting the information source determined to be the information source related to the production goods are matched with the words of each category in the domain dictionary (step S402). Then, it is determined whether or not one or more, a part or all of them are identical (step S403), and the information source is registered as a compatible information source (step S404).

  Next, product names / characteristics constituting the information source that are not registered in the domain information are added to the domain dictionary (step S405). Next, it is determined whether or not the number of adaptations is equal to or greater than a predetermined ratio (step S406). If the predetermined number of conditions is met, the information source subject is added to the domain dictionary (step S407). Thereby, it is possible to distinguish whether the “subject” constituting the information source is a specialist in the business domain or a general manufacturer. In this example, the element words constituting the information source shown in FIG. 22 and the business domain dictionary shown in FIG. 8 do not partially match. Therefore, in this business domain selection step S400, the input information source matches the business domain. It is determined not to. If they do not match, the information source is stored in the unsuitable information source storage unit 304.

  Next, the information source multiplication unit 401 refers to the technical dictionary, and selects the business domain candidates from the information sources in the non-conforming information source storage unit 304, thereby proliferating the information sources and proliferating the words in the technical dictionary and the domain dictionary. (Step S500). FIG. 24 is a flowchart showing an information source proliferation method.

  As shown in FIG. 24, the information source multiplication unit 401 first collates all the nodes of the technical dictionary shown in FIG. 10 with each element word of the non-conforming information source stored in the non-conforming information source storage unit 304. (Step S411). In this example, since the feature word “light amplification” matches one node of the technical dictionary, it is listed as a business domain candidate. (Step S412) Then, the distance between this “optical amplification” and the domain dictionary feature word “optical access” is calculated using the technical dictionary (step S413). In this case, the distance between “optical amplification” and “optical access” is 0.6. Here, if the threshold is set to be less than 1.5 as the business domain (step S414), the information source including the characteristic word of optical amplification is also determined to be related to the business domain by the information source multiplication unit 401. Is done. Then, the non-conforming information source determined to be related to the business domain is stored in the conforming information source storage unit 303 (steps S415 and S600).

  Then, the information source multiplication unit 401 adds “Raman method” as a feature word based on the non-conforming information source determined to be related to the business domain (step S416). Further, the information source multiplication unit 401 fixes, for example, a “Raman” node as a child node of the optical amplification node of the technical dictionary based on the information source determined to be related to the business domain among the non-conforming information sources. Add at a distance of 0.1. In this case, it is possible to set a fixed value of the system as a distance for the word that is automatically propagated in this way. For example, an original sentence (information) of the word to be added and the word to which the word is added is added. It is also possible to calculate based on the positional relationship in the (source). Further, it is determined whether or not the number of times of matching is equal to or greater than a predetermined ratio (step S417). If the number of times of matching is equal to or greater than the predetermined ratio, “opt sales” is added as the subject of the domain dictionary. (Step S418). Business domain dictionaries and technical dictionaries are automatically expanded by inputting information sources that match the business domain.

  Next, effects of the first exemplary embodiment of the present invention will be described. In this embodiment, the information source is decomposed into the subject, action, product name, and feature word, and compared with other information sources using the business domain dictionary and technical dictionary, so the competing product of the company that has been unknown until now Can be discovered and presented for purchase. Then, the morpheme sets of information sources are combined with continuous nouns, and the subject, product name, action, feature word, and synonym can be extracted by the format rule.

  In addition, by using a domain dictionary, only information sources belonging to a production domain and a target business domain can be extracted as relevant information. Furthermore, an information source that is determined to be non-conforming by using a domain dictionary can be promoted as a conforming information source by using a technical dictionary, and a word that does not exist in the domain dictionary or technical dictionary conforms to the domain. Can be propagated using the feature words. In other words, feature words of new competitive products can be added to the business domain dictionary and technical dictionary, so that the terms of the business domain / technical dictionary can be automatically expanded even when a new product is released. This can reduce the man-hours required for manual maintenance. Furthermore, since the information source inputted later is also compared including the feature word which grew, it becomes possible to discover a product more widely.

Modification of the first embodiment of the present invention Next, a second embodiment of the present invention will be described. The first embodiment is that the information source reconstruction unit has a tree generation unit that makes the information source a tree structure, and the information source selection unit and the group generation unit handle the information source in the state of the information source tree. Different from the example. The other points are the same.

  FIG. 25 is a diagram showing an information structuring apparatus 1b according to a modification of the first embodiment of the present invention. As shown in FIG. 1, the information structuring apparatus 1b according to the present modification includes an information source cutting unit 100, an information source reconstruction unit 200b, an information source selection unit 300b, and a group generation unit 400b.

  FIG. 26 is a diagram illustrating the information source reconstruction unit 200b according to the present embodiment. As shown in FIG. 26, the information source reconstruction unit 200b includes a format analysis unit 201, a format rule storage unit 202, an element word storage unit 203, a tree generation unit 204, and an information source tree storage unit 205. Other than the tree generation unit 204 and the information source tree storage unit 205 are the same as those in the first embodiment.

The tree generation unit 204 generates an information source tree for rooting the subject from unconnected element words. FIG. 27 is a diagram showing an information source tree. As shown in FIG. 27, the subject name is the root node, and the product name having the trademark is its child node. Furthermore, derivative products, actions, and feature words are child nodes of product names having trademarks. Furthermore, synonyms and feature words are child nodes of the derived product, and the action target is a child node of the action. This information source tree is generated according to the following rules.
(1) The subject name is the root node. (2) The product name is a word having universality in relation to trademarks. The product name is a child node of the company name. (3) The derived product name is a child node of the product name having the trademark. (4) The company action is the child node of the product or derivative product. Is a child node of an action (6) A feature word is a child node of a product name or a derived product name (7) A synonym is a node at the same level as a keyword to be paraphrased

  The information source tree generated according to the above rule corresponds to one information source. The generated information source tree is stored in the information source tree storage unit 205.

  By this information source reconstruction unit 200b, morpheme sets of information sources are combined with continuous nouns, subject, product name, action, feature word, and synonym are extracted by format rules, and a tree structure is constructed from these element words can do. By making the extracted element words into a tree structure, an information source tree in which the element words are structured can be obtained.

  FIG. 28 is a block diagram showing the information source selection unit 300b in the present embodiment. As illustrated in FIG. 28, the information source selection unit 300b includes a tree selection unit 301b, a domain dictionary storage unit 302, a compatible information source tree storage unit 303, and a non-conforming information source tree storage unit 304b. The difference from the first embodiment is that the information source tree is handled instead of the information source.

  FIG. 29 is a block diagram illustrating the group generation unit 400b in the present embodiment. As illustrated in FIG. 29, the group generation unit 400b includes a tree multiplication unit 401b and a technical dictionary storage unit 402. Similar to the information source selection unit 300b, it is different from the first embodiment in that an information source tree is handled instead of an information source.

  Next, the operation of this embodiment will be described. FIG. 30 is a flowchart illustrating the operation of the information structuring apparatus according to the present embodiment. Similarly to the above, “DW4000” is registered in the “product name” row and “word / phrase 1” column of the domain dictionary of the domain dictionary storage unit 302 so as to match the business domain. (See Figure 8)

And as an information source,
"NEC releases optical access product" DW4000 ""
Is input to the system, the information source cutting unit 100 performs morphological analysis in the same manner as described above to obtain the morpheme shown in FIG. (Step S100). Then, the format analysis unit 201 applies the format analysis rule (step S300). As a result, an analysis result as shown in FIG. 21 is obtained (step S300). In other words, from the information source “NEC releases optical access product“ DW4000 ””
Subject: NEC Product Name: DW4000
Action: Launch Feature word: The result of the analysis is an optical access product.

  From the analysis result, the tree generation unit 204b generates an information source tree as shown in FIG. 31 according to the predetermined rule shown in FIG. 27 as described above (step S360). The generated information source tree is stored in the information source tree storage unit 205b. Then, the tree selection unit 301b of the information source selection unit 300b determines whether or not the information source tree is related to the production goods depending on whether or not the information source tree includes actions such as “delivery” and “adoption”. judge. Then, the subject, the product name, and the feature word constituting the information source tree determined to be related to the production goods are compared with the domain dictionary in the domain dictionary storage unit 302. When a word in the domain dictionary and an element word constituting the information source tree partially match or all match, the information source tree is determined as a matching information source tree. In this example, in the domain dictionary shown in FIG. 8, it can be seen that the product name “DW4000” matches and the feature word “optical access product” partially matches “optical access” in the domain dictionary. The source is determined to be a compatible information source tree that matches the business domain, and the information source tree is stored as a compatible information source tree. (Step S400). The information source tree determined to be the compatible information source tree is stored in the compatible information source tree storage unit 303.

Next, a method in which the tree selection unit 301 selects a compatible information source tree from the information source tree with reference to the domain dictionary will be specifically described. As an information source (news information source)
"Opto Sales Deliver" FX4200 "with Raman Method for Optical Amplification to NEE"
Is input to the information structuring apparatus 1b. The information structuring apparatus 1b executes steps S100 to S300 in FIG. 12 and obtains the format analysis result shown in FIG. The tree generation unit 204b generates an information source tree as shown in FIG. 32 from the format analysis result, and the tree selection unit 301b collates the element words constituting the information source tree with the business domain dictionary. In this example, the element words constituting the information source tree shown in FIG. 22 and the business domain dictionary shown in FIG. 8 do not partially match. Therefore, in this business domain selection process, the input information source matches the business domain. It is determined not to. If they do not match, the information source tree is stored in the unsuitable information source tree storage unit 304b.

  Thus, when the information source is a non-conforming information source tree (step S410: Yes), the tree breeding unit 401b refers to the technical dictionary, and from the information tree of the non-conforming information source tree storage unit 304b, business domain candidates. The information source tree is propagated by selecting and the words of the technical dictionary and the domain dictionary are propagated (step S500). Then, the non-conforming information source tree determined to be related to the business domain is stored in the conforming information source tree storage unit 303b (step S600).

  In this modified example, as in the first embodiment, the information source is decomposed into the subject, action, product name, and feature word, and compared with other information sources using the business domain dictionary and technical dictionary. It is possible to discover competing products of companies that were not known before and present them as purchase targets. Then, the morpheme sets of the information sources are combined with continuous nouns, and the subject, product name, action, feature word, and synonym are extracted according to the format rules, and a tree structure can be constructed.

Second Embodiment of the Invention Next, a second embodiment of the present invention will be described. This embodiment is different from the first embodiment in that the information source selection unit has a subject dictionary for determining what category the subject is a company. In the following embodiment, an example in which an information source is handled as an information source tree as in the above-described modification will be described, but it goes without saying that the information source does not have to have a tree structure. FIG. 33 is a diagram illustrating the information source selection unit according to the present embodiment. Other configurations are the same as those in the first embodiment, and a detailed description thereof will be omitted. As shown in FIG. 33, the tree selection unit 301 of the information source selection unit 300c refers to the subject dictionary 305 when classifying the information source tree into the matching information source tree, and classifies the subject information for each subject in each information source tree.

  FIG. 34 is a diagram showing a subject dictionary. The subject dictionary is composed of a combination of a company category and the company name, such as a customer company: NEE, a competitor company: Tameshita Industry, and a parts supplier company: Shinkawa Electric. The tree selection unit 301 determines whether the information source tree is a customer company, a competitor company, or a parts delivery company, for example, extracts only the competitor company information and displays it on a display unit (not shown). To do.

  Next, the operation of the second embodiment of the present invention will be described in detail with reference to the flowchart of FIG. In this embodiment, an example is shown in which it is detected that an unknown competitive product has been delivered to an unknown customer. In the present embodiment, the values of the business domain dictionary (FIG. 35) and the technical dictionary (FIG. 36) proliferated as a result of the first embodiment are used.

As source news,
"Tameshita Industry Delivers Lightest" J703iU "for ADDI"
Is input to the information structuring apparatus. Similar to the information structuring apparatus 1, the information structuring apparatus executes steps S100 to S300 and obtains the format analysis result shown in FIG. With reference to the business domain dictionary shown in FIG. 35, it is determined that this information source does not belong to the business domain, that is, is a non-conforming information source tree (step S400). Since the information source is a non-conforming information source tree (step S410), the technical dictionary shown in FIG. 36 is referred to and it is determined that this information source is not a business domain candidate (step S500).

Then, as news of information sources,
"Tameshita Sangyo to deliver" DDI5000 "DWDM product to ADDI"
Is input to the information structuring apparatus 21. Similarly, the information structuring apparatus 21 executes steps S100 to S300 and obtains the format analysis result shown in FIG. When an information source including words such as “delivery” and “adoption” is input, the tree selection unit 301 refers to the business domain dictionary, and the information source tree is a conforming information source tree that matches the business domain. It is determined whether or not there is (step S400). In this example, with reference to the business domain dictionary shown in FIG. 35, it is determined that this information source does not belong to the business domain.

  Since the information source is a non-conforming information source tree (step S410), referring to the technical dictionary shown in FIG. 36, the information source matches with the keyword DWDM, and the feature word for the product belonging to the business domain In comparison, since the distance is equal to or smaller than the threshold, it is determined that the tree is a compatible information source tree (step S500). Here, the subject “Tameshita Sangyo” has a conforming information source tree that matches the business domain and a conforming information source tree that does not match, and the ratio is 50%, and is registered as the subject of the business domain dictionary. Is not registered in the business domain dictionary. However, the product name (“ADDI15000”) and the feature word (“DWDM product”) included in the information source determined to be the compatible information source tree are registered in the business domain dictionary (FIG. 39). Here, in this embodiment, based on the subject dictionary shown in FIG. 34, the tree selection unit 301 can classify the compatible information source tree for each subject. Thus, for example, only competitor products can be presented to display means (not shown) connected to the information structuring device 21.

  Also, the subject can determine whether the news matches the business domain. If the name of a specialist is included in the news, even if it is a feature word that is not in the technical dictionary, it is determined that the product matches the business domain. It can be determined that it is suitable. Therefore, the customer who purchased it can be authorized as a new customer. In addition, if the general manufacturer K is registered as the subject person, if a K laundry machine is erroneously determined to be an optical network product or there is a coin laundry company that has purchased a large number of K laundry machines, May be misidentified as a new customer who has purchased a HIKARI business domain product, but by determining the subject without registering the general manufacturer as the subject, all the products of a manufacturer match the business domain. Preventing you from doing.

  In this embodiment, as in the first embodiment, even if the product name, competitor, and customer company are unknown, the technical dictionary is used to determine whether the matching information source tree matches the business domain. Use to determine. For this reason, it is possible to classify delivery news that is threatening to the company and those that are not. In addition, in order to determine whether the subject is a specialist or a comprehensive manufacturer, taking into account past news, if the subject has developed various conforming information source trees, a new product in a different business domain will be mistakenly And a customer who has purchased a new product in this different business domain is also less likely to be erroneously determined as a new customer. Furthermore, in the present embodiment, since the subject can be classified, it is possible to extract company information of a necessary category at a desired timing and display it.

Third Embodiment of the Invention Next, a third embodiment of the present invention will be described. FIG. 40 is a diagram showing an information structuring apparatus 31 according to the third example of the present invention. The third embodiment is an example of dynamically determining a customer company. As shown in FIG. 40, the information structuring apparatus 31 further includes a quantification apparatus 32 in addition to the information structuring apparatus 1 in the first embodiment. The quantification device 32 includes an information source value estimation unit 500 and a company power estimation unit 600. Other configurations are the same as those of the first embodiment, and a detailed description thereof will be omitted. The information source value estimation unit 500 is a definition table (word quantification) in which leaf score calculation methods constituting the lowest layer of the enterprise strength measurement tree that represents the company characteristics in a tree structure are defined by element words constituting the information source tree. ) And the information source tree, the score of the leaf is calculated as an initial score. The enterprise power estimation unit 600 calculates the score of the node constituting the uppermost layer of the enterprise strength measurement tree as the enterprise strength based on the initial score. The enterprise strength measurement tree is composed of a leaf and its upper layer nodes, and the score calculation method for each node is defined. The upper layer score is calculated based on the lower layer score. Details will be described later. Next, the information source value estimation unit 500 and the enterprise power estimation unit 600 will be described in detail.

  FIG. 41 is a block diagram showing the information source value estimation unit 500 in the present embodiment. As shown in FIG. 41, the information source value estimation unit 500 includes a subject determination unit 501, a subject dictionary storage unit 502, a subject classified information source storage unit 503, an information source quantification unit 504, a word It has a quantification map storage unit 505, an information reduction initial score storage unit 506, and an enterprise strength measurement tree storage unit 507.

  The subject person dictionary storage unit 502 stores the subject person dictionary. As shown in FIG. 34 described above, the subject person dictionary has a structure including a set of a category, {customer, competition, parts delivery} and a company name (name). The name of the competitor, the name of the customer company, and the name of the parts delivery company are stored in the company name of the subject dictionary.

  The subject determination unit 501 refers to the matching information source tree storage unit 303 and compares the subject node with the subject dictionary in the subject dictionary storage unit 502. Then, the conforming information source tree is classified by the subject, and stored in the information source storage unit 503 as a competitor company information source tree for a competitor company, a customer company information source tree for a customer company, or a delivery company information source tree for a delivery company. To do.

  The information source quantification unit 504 is stored in the corporate strength measurement tree storage unit 507, which will be described later, from the word quantification map having the structure shown in FIG. 42 stored in the subject information source tree and the word quantification map storage unit 505. The score of the leaf constituting the lowest layer of the corporate strength measurement tree is calculated as the initial score.

  First, the word quantification map shown in FIG. 42 will be described. The word quantification map is a combination of “subject”, “product name”, “feature word”, “action”, and “action target” in the information source tree. Rules and corresponding initial score calculation methods are described.

  Specifically, the “word” column that acts as a trigger for starting the calculation, the “score calculation method” column that describes how the score is calculated from the information source tree, and the initial score allocation destination are determined. And a “distribution company” column, an “allocation leaf” column, and an “allocation rate” column.

  There are roughly four types of score calculation methods. The first is a method of giving a fixed value. If a word is included in the information source, a value designated by the fixed value is given as an initial score. The second is to calculate using the value of the leaf of the distribution destination, and is a method applied when the value of the target leaf is doubled or tripled. The third method is to calculate the words of the information source tree in combination, and applies the face value such as “10 billion” and “capital increase” as the score. The fourth method is to calculate according to a method defined by an external calculation unit, and is a method applied when calculating a score using a plurality of information sources.

  Next, the corporate strength measurement tree will be described. FIGS. 43 to 45 are diagrams showing the structures of a customer enterprise strength measurement tree, a competitive enterprise strength measurement tree, and a parts delivery enterprise strength measurement tree, respectively. The enterprise strength measurement tree storage unit 507 includes a competitive enterprise strength measurement tree for the competitor shown in FIG. 43, a customer enterprise strength measurement tree for the customer enterprise shown in FIG. 44, and a component delivery company for the component supplier shown in FIG. Three types of enterprise strength measurement trees are prepared. As shown in FIG. 43 to FIG. 45, the enterprise strength measurement tree has two types of data structures, a node structure and a leaf structure, and uses the enterprise name as a root node, and enterprise characteristics for expressing the competitiveness of each enterprise. It is possible to have nodes hierarchically. The lowest layer is a leaf for assigning an initial score generated from an information source. A leaf cannot have nodes.

  FIG. 46 is a diagram showing a word / node correspondence table included in the word quantification map. The information source quantifying unit 504 determines whether an element word that matches the word column in the word / node correspondence table is included in the subject-specific information source tree. If it is included, the score specified in the score field is set as the initial score for the leaves (node 1 to node 3) that constitute the lowest layer of the enterprise ability measurement tree of the type corresponding to the word field. . For example, if the subject (type) of the information source tree by subject is a competitor and the information source tree includes “exhibition”, the score (initial score) of the “delivery record” leaf is set to 2. Also, for example, when the subject (type) of the information source tree by subject is “competitor or delivery company” and “development” is included in the information source tree, “product variation”, “development speed”, Double the score (initial score) of each “functional feature” or “part variation”, “development speed”, and “part highest performance ranking” and multiply by 0.4, 0.2, and 0.4 respectively. To do. As described above, the information source value estimation unit 500 uses the word quantification map to cause the information source quantification unit 504 to combine the action existing in the information source tree and the feature word associated therewith to obtain the initial information source. A score can be calculated.

  FIG. 47 is a block diagram showing the corporate strength estimation unit 600 in the present embodiment. As shown in FIG. 47, the enterprise power estimation unit 600 includes an enterprise power calculation unit 601 and an information source final score storage unit 602.

  The enterprise power calculation unit 601 receives the initial score given to the leaves constituting the lowest layer of the enterprise strength measurement tree, and the word / node allocation allocation rate of the word quantification map stored in the word quantification map storage unit 505 Then, the score of each node is calculated from the enterprise strength measurement tree information stored in the enterprise strength measurement tree storage unit 507, and the enterprise strength (the highest node) is calculated. Then, a value obtained by subtracting the past corporate strength (score) from the current corporate strength (score) is generated as the final score of the information source. FIG. 48 is a diagram showing a node structure and a leaf structure. 43 to 45, the value of each node is calculated from a lower leaf value or node value, and a calculation method can be defined for each node.

  The enterprise power calculation unit 601 first determines to which enterprise type the information source tree belongs using the subject dictionary stored in the subject dictionary storage unit 502 and is stored in the word quantification map storage unit 505. The initial score calculated by the information source quantification unit is distributed to the corresponding node of the corresponding company type according to the word / node allocation distribution rate described in the word quantification map. The word / node allocation allocation ratio describes which leaf of the enterprise strength measurement tree the initial score is allocated to. Here, the company characteristic measurement tree has different company characteristics for each company type of the subject, and has different characteristic values for each company of the subject. For this reason, even if the initial score calculated from the information source is the same, if the subject to be applied is different, the final score is different.

  If the subject is a customer company, as shown in FIG. 44, the information source is quantified by an enterprise strength measurement tree prepared for the customer company. The customer enterprise strength measurement tree is composed of child nodes of “investment power” and “service occupation power”. The “investment power” node includes child nodes of “investment aggressiveness” and “new investment power”. “Investment aggressiveness” has leaves of “current assets” and “fixed assets”, and “new investment power” has leaves of “net assets” and “liabilities”. The “service occupancy” node has child nodes of “regional” and “demand”. Furthermore, the “regional” node has “service density” and “service area” leaves. The “demand” node has a “cycle” and a “product matching” leaf.

The initial value of each node and leaf is 1, and the enterprise power calculation unit 601 calculates the score of each node using the word / node correspondence table shown in FIG. Each node or leaf of each company type receives the word contained in the information source and updates its score according to the rules. For example, a rule “use the score described in the information source as it is” is assigned to the “score calculation method” column of the fixed asset leaf and the current asset leaf. If the information source tree contains information of “100 billion yen investment”, the fixed asset leaf will increase the face value by “100 billion” and the current asset leaf will reduce the face value by “100 billion”. As described above, the initial score of the lowest leaf is calculated by the information source quantification unit 504. The enterprise power calculation unit 601 calculates the scores of the middle and top nodes by the following calculation formula.
Intermediate node: Level 2
“Investment aggressiveness” node score = “Fixed asset” leaf score / “Current asset” leaf score “New investment power” node score = “Net asset” leaf score / “Fixed debt” leaf score “Regional” Node score = (α01 × “service density” leaf score) + (α02 × “service area” leaf score)
“Demand” node score = (β01 × “period” leaf score) + (β02 × “product matching” leaf score)
Intermediate node: Level 1
“Investment power” node score = α11 × “Investment aggressiveness” node score × α12 × “New investment power” node score “Service occupancy” node score = β11 × “Regional” node score × β12 × “ “Demand” node score Top node “Customer company score” = α21 × “Investment power” node score × β21 × “Service occupancy”

Here, the weights from α01 to β21 can be freely set, and it is possible to select important characteristics among the corporate strengths belonging to the customer enterprise type. The score of each node can be obtained by the following calculation formula.
Leaf score (t + 1) = initial score of information source (t + 1) node score (t + 1) = node score (t) + h (node_child1, node_child2, ..., node_childN)
Here, h () is a function that can be freely set for each node. Specifically, the score of the above-mentioned “investment aggressiveness” node = the score of the “fixed asset” leaf / the score of the “current asset” leaf It is a mathematical formula like this. The value of the function h () may be negative, and as a result, the value of each node can be decreased.

  Then, the enterprise strength calculation unit 601 stores the above-mentioned node scores in the customer enterprise strength measurement tree. The enterprise strength calculation unit 601 generates a different enterprise strength measurement tree for each subject. Since the information source increases the score of the node of the enterprise strength measurement tree corresponding to each enterprise, the enterprise with a lot of news releases accumulates the enterprise strength compared with the enterprise with few. The final score of the information source is equal to a score obtained by subtracting the value of the previous highest node from the score obtained at the highest node, and the enterprise power calculation unit 601 stores the final score in the information source final score storage unit 602. .

Similarly, if the subject is a competitor, the information source is quantified based on a corporate strength measurement tree prepared for the competitor as shown in FIG. The corporate strength measurement tree is composed of child nodes of “customer convenience”, “credit strength”, “sales power”, and “development power”. The “customer convenience” node has a child leaf of “support base”. The “credit” node has child leaves of “rating” and “delivery record”. The “sales power” node has child nodes of “sales results”. The “development power” node has child leaves of “product variation”, “development speed”, and “functional characteristics”. Each node and leaf has an initial score of 1, and the initial score is increased according to the contents of the information source. As described above, for the lowest leaf, the information source quantification unit 504 directly calculates a score from the contents of the information source. For example, if information “start sales of three products” is posted in the information source, the information source quantifying unit 504 increases the leaf score of the product variation by 3 with reference to FIG. The enterprise power calculation unit 601 calculates the middle and top nodes by the following calculation formula.
Intermediate node “customer convenience” node score = (α11 × “1 / access time” leaf score) + (β11 × “number of service locations” leaf score) + (γ11 × “number of support personnel” leaf score)
“Credit” node score = (α21 × “delivery result” leaf score) + (β22 × “rating” leaf score)
Score of “development ability” node = (α31 × “product variation” leaf score) + (β32 × “development speed” leaf score) + (γ33 × “functional feature” leaf score)
Topmost node “competitor score” = (α41 × “customer convenience” node score) + (β42 × “credit” node score) + (γ42 × “sales power” node score)

  Similarly, if the subject is a supplier of parts, the information sources are quantified based on the enterprise strength measurement tree prepared for the parts delivery company as shown in FIG. The enterprise strength measurement tree includes child nodes of “customer convenience”, “credit strength”, and “development ability”. The “customer convenience” node has a “number of deliverables” and a “delivery date” leaf. The “creditworthiness” node has a “delivery record” and “rating” leaf. The “development power” node has “part variation”, “development speed”, “part highest performance ranking”, and “differentiation” leaf. Each node and leaf has an initial score of 1, and the score is increased according to the contents of the information source. As described above, the information source quantification unit 504, for example, increases the initial score of the component highest performance ranking leaf by 2 when information “develop the world's fastest component” is input to the information source. The enterprise power calculation unit 601 calculates the middle and top nodes by the following calculation formula.

Intermediate node “customer convenience” node score = (α11 × “delivery available” leaf value) + (β11 × “delivery date” leaf score)
“Credit” node score = (α21 × “delivery result” leaf score) + (β22 × “rating” leaf score)
Score of “development ability” node = (α31 × “part variation” leaf score) + (β32 × “development speed” leaf score) + (γ33 × “part highest performance ranking” leaf score)

Top node “parts supplier score” = (α41 × “customer convenience” node score) + (β42 × “credit” node score) + (γ43 × “development” node score)
Note that the values of α, β, and γ can be freely set according to the business domain. In addition, the calculation formulas of the leaves constituting the enterprise power measurement tree may be freely redefined. Further, the node calculation formula may be implemented by combining multiplication, addition, subtraction, and other functions between leaves.

  The enterprise power calculation unit 601 uses the subject dictionary to classify the tree into customer companies / competitors / delivery companies, and uses an evaluation template (enterprise power measurement tree) for each company type to determine the initial score of the information source. The final score of the information source can be quantified by allocating it to the characteristics of corporate strength and taking into account past corporate strength. In addition, the evaluation template for each company type has a structure that allows each node or leaf to have a rule. By applying the rule to the value of the information source, the value of each node or leaf is calculated. can do. Furthermore, the evaluation template for each company type has a structure that can give weights to the links connecting nodes by “ratio”, and the company characteristics that the analyst considers important and those that are not considered important. Can be set for each company type.

  The operation of the third embodiment of the present invention will be described in detail with reference to the flowchart of FIG. As a premise, the values of the business domain dictionary (FIG. 39) and the technical dictionary (FIG. 36) which are propagated as a result of the second embodiment are used. Also assume that the client company has 5 billion net assets, 5 billion liabilities, 3 billion fixed assets, and 7 billion current assets.

As source news,
"ADDI increases capital by 10 billion yen through third-party allotment for metro business"
Is input to the information structuring apparatus 31. The information structuring apparatus 31 executes steps S100 to S400 in the same manner as the information structuring apparatus 1, and determines that the information source matches the business domain by the keyword “metro”. Thus, the information source tree shown in FIG. 50 is created and stored in the relevant information source tree storage unit 303 (step S400).

  Next, the subject determination unit 501 refers to the subject dictionary shown in FIG. 34, determines that the subject “ADDI” is a customer company, and stores it in the information source storage unit 503 as a customer company information source tree. Then, the information source quantification unit 504 reads this customer company information source tree, and based on the customer company power measurement tree shown in FIG. 45 and the word quantification map shown in FIG. 46, each element word of the information source tree. An initial score is calculated (step S700).

  Here, before the information source is input, the enterprise strength calculation unit 601 calculates the upper layer score based on the initial score of 50 for net assets, 50 for liabilities, 30 for fixed assets, and 70 for current assets. For example, since “investment aggressiveness” score = fixed asset / current asset score, 30 / 70≈0.43, and “new investing power” score = net asset / liability, 50 = 1. Since the score of “investment power” = α11 × (score of “investment aggressiveness” node) × α12 × (score of “new investment power” node), if α11 = α12 = 1, 1 × 0 .43 × 1 × 1 = 0.43.

If the initial value of service density, service area, cycle, and product matching is 1, α01 to β21 = 1,
“Regional” node score = (α01 × “service density” leaf score) + (α02 × “service area” leaf score) = 2
“Demand” node score = (β01 × “period” leaf score) + (β02 × “product matching” leaf score) = 2
“Service Occupancy” node score = β11 × “Regional” node score × β12 × “Demand” node score = 4
Thus, “customer company score” = α21 × “investment power” node score × β21 × “service occupancy” = 4.93.

  When the information source arrives, the information source quantification unit 504 updates the initial score. That is, since the word “capital increase” is included in the information source tree shown in FIG. 50 and the word “capital increase” is also included in the word / node correspondence table in FIG. 46, the initial score of the node “net asset” is set. Update. In this case, since the ratio of the “debt” leaf, which is the leaf of the new investment power node, is the same as the “net asset” leaf, the score is not allocated. That is, since the word “10 billion yen” is included in the information source tree, “% s yen” (“% s yen” indicates that the number before “yen” is used as the node score). The corresponding “10 billion” is the initial score for “Net assets”.

  The enterprise power calculation unit 601 calculates the score of each node based on the updated initial score (step S800). “New investment power” is 150/50 = 3 after the information source arrives.

  Each time ADDI's capital increase news flows, the values of the “investment aggressiveness” and “new investment power” nodes change. The investment power is obtained by multiplying the investment aggressiveness and the new investment power. If the values of the coefficients α11 and α12 are 1, respectively, 1.9 × 3.0 = 5.7 is the value of the investment power node. Since the value of the investment power node was originally 0.42 × 1.0 = 0.42, it can be seen that the increment of 5.7−0.42 = 5.28 is the change score. Here, the weights α11 and α12 may be free values depending on the business domain.

  As for the score of the service occupancy node, if both α and β are 1, the regional score × demand score = 4, and the change before and after arrival of the information source is 0. The corporate strength score = investment power × service occupation power, and 5.7 × 4 = 22.8 is the corporate strength. The change in the corporate power before and after arrival of the information source is obtained as 5.7 × 4−0.42 × 4 = 21.12, and this value becomes the score of the information source.

  In the third embodiment, the enterprise strength calculation unit 601 refers to the enterprise strength measurement tree defined for each competitor, customer, and delivery company, and based on the enterprise strength including the value of past information sources. Calculate the enterprise power generated by the information sources. The difference between the corporate power before and after the information source is given is regarded as the final information source score. This classifies the tree into customer companies / competitors / delivery companies using the subject dictionary, and assigns the initial score of the information source to the characteristics of the company power using the evaluation template for each company type. The final score of the information source can be quantified taking into account the corporate strength. In the present embodiment, the information source value is estimated using the data from the group generation unit 400 and the enterprise power is estimated. However, the information of the information source reconstruction unit 200 is used to obtain information. It is also possible to perform source value estimation and enterprise power estimation.

  Next, another operation example of the present embodiment will be described. This example is an example showing that the producer of product goods has a technical advantage. As a premise, the values of the business domain dictionary (FIG. 39) and the technical dictionary (FIG. 36) which are propagated as a result of the second embodiment are used.

As source news,
"Opto Sales develops" FX5000 "lossless optical transmission device using μ laser"
Is input to the information structuring apparatus 41. The information structuring apparatus 41 executes steps S100 to S400 and obtains an information source tree shown in FIG. Then, since the subject “Opt Sales” is registered as the subject in the business domain dictionary shown in FIG. 35, it is determined that this information source is the optical business domain, and the information source is classified into the relevant information source tree (step). S400).

  Next, in S700, the subject determination unit 501 refers to the subject dictionary shown in FIG. 34 and determines that the subject “Opt Sales” is a competitor. Then, it is stored in the information source storage unit 503 as a competitor information source tree. Next, the information source quantification unit 504 reads the competitor company information source tree in the information source storage unit 503, and performs quantification based on the competitor company strength measurement tree shown in FIG. 44 and the word / node correspondence table shown in FIG. Do. That is, first, the information quantification unit 504 calculates the initial score of the leaves that constitute the lowest layer of each enterprise strength measurement tree.

  Here, all the leaves are given 1 as an initial score. There is a keyword “development” in the competitive information source tree, and it is described from the word / node correspondence table in FIG. 46 that each leaf score is doubled. For each leaf connected to the development capability node, a ratio for calculating the score of the development capability node is defined. By setting the ratio value, the link between nodes can be weighted, and the company characteristics that the analyst considers important and the company characteristics that are not considered important can be set for each company type. It has become.

In this example, an extension function is set for the word “development”, and it is described that the score is distributed at a ratio of 0.4 for the product variation leaf, 0.2 for the development speed leaf, and 0.4 for the functional feature. Has been.
Extension function: 2 x g (f (product variation), f (development speed), f (feature))
here,
g (param1, param2, param3) {

if (param1 node == product variation) {
return param1;
}
f (product variation) = number of past released products + 1
f (development speed) = f (product variation) / cumulative release period f (feature) = number of feature words in the information source tree.

Since the opt product has “FX4200” in the past, the number of past released products is one. Therefore,
f (product variation) = the number of past released products + 1 = 2.

Since function g adopts the value of f (product variation) as the first argument as it is,
2 x g (f (product variation), f (development speed), f (feature))
= 2 × f (product variation)
= 2 × 2
= 4
It becomes. And since the distribution ratio to the product variation leaf is 0.4 in the word quantification map,
Product variation leaf value = 4 x 0.4 = 1.6
It becomes.

  That is, for the keyword “development”, a value obtained by doubling the result of the expansion function is distributed at the distribution destination and distribution rate described in the quantification map, and an initial score to be given to each leaf is determined. The extension function has three parameters as arguments and functions so that the value returned by the leaf type of each argument is different. In the case of the product variation leaf, the return value of the extension function is obtained by adding 1 to the number of past released products. In the case of the development speed leaf, the value obtained by dividing the number of products by the total release interval of each product is the return value of the extension function. In the case of a feature leaf, the number of feature words of the information source is used as the return value of the extension function.

Further, it can be seen that the product was released two years ago from the time information shown in FIG. 3 of the information source of the previous product “FX4200” of Opt Sales, and the development speed can be calculated by the following calculation formula.
Value of development speed leaf = 2 × All products sold by Opt / Release interval × 0.2 = 2 × 2/3 × 0.2 = 0.26

Further, “FX5000” includes two terms “μ laser” and “lossless optical transmission device” as characteristic words, and can be obtained by the following calculation formula.
Score of functional feature leaf = 2 × number of feature words × 0.4 = 2 × 2 × 0.4 = 1.6

Next, the corporate strength calculation unit 601 calculates the score of each node using these initial scores (step S800). Assuming that there are two types of "Opto Products" due to this information source, the score of the development power node is multiplied by the score of the leaf and the score of the information source = product variation leaf score + development speed leaf score + functional characteristics Leaf score = 1 × 1.6 + 1 × 0.26 + 1 × 1.6
= 3.46
Is required. If the score of the development capability node is an initial value 1, a difference of 3.6-1.0 = 2.6 from the current score is an increase before and after arrival of the information source at the development capability node. If the initial value of customer convenience, creditworthiness, and sales power is 1,
Top node (corporate power) score = customer convenience score + credit score + sales power score + development score = 1 + 1 + 1 + 3.6
= 6.6
It becomes. Also, the difference in corporate power before and after arrival of the information source is 6.6-4.6 = 2.6, and this 2.6 is the final score of the information source.

  In the present embodiment, the enterprise power of each company can be quantified from the keywords included in the information source, the word / node correspondence table, the structure of the competitive enterprise strength measurement tree, and the leaf value calculation formula. Furthermore, since the impact (difference of the highest node of the corporate strength measurement tree) possessed by the keyword is calculated against the past corporate activity, the score of the information source can be evaluated for each enterprise.

Fourth Embodiment of the Invention In the fourth to eighth embodiments described below, a ripple force calculation unit 700, a demand cycle estimation unit 800, a potential purchasing power estimation unit 900, and a sales force estimation unit 1000, respectively. A competitiveness enhancement component estimation unit 1100 is provided. Each block has a function of updating the score of a leaf of a specific enterprise strength measurement tree. In other words, the ripple force calculation unit 700 calculates the ripple force that one information source gives to the enterprise power of a competitor, and updates the score of the functional feature leaf of the competitor strength measurement tree. The demand cycle estimation unit 800 estimates the demand cycle of the product and updates the score of the cycle leaf of the customer enterprise strength measurement tree. The potential purchasing power estimation unit 900 estimates the potential purchasing power of a customer company, estimates the function and price of a product with high sales potential, and updates the score of the product matching leaf in the customer company power measurement tree. The sales force estimation unit 1000 estimates the sales force of the company and updates the score of the sales performance leaf in the competitive company power measurement tree. The competitiveness-enhancing component estimation unit 1100 estimates components that have an impact on competitiveness, and updates the score of the differentiated leaf in the component supplier enterprise strength measurement tree. In the following fourth to eighth embodiments, description will be made on the assumption that each process is executed using the matching information source tree output from the group generation unit 400, but the information source reconstruction unit 200 extracts it. It is also possible to execute each process using element words. First, a fourth embodiment of the present invention will be described.

  FIG. 52 is a diagram illustrating the information structuring apparatus 41 according to the present embodiment. In this embodiment, the range of influence of the current information source can be extended to companies that are not directly described by tracing past information sources. As shown in FIG. 52, the information structuring apparatus 41 has a configuration in which a ripple force calculation unit 700 is further provided as the quantification apparatus 32 in the information structuring apparatus 31 in the third embodiment. Other configurations are the same as those of the third embodiment, and detailed description thereof is omitted. The fourth embodiment is an example showing that a component manufacturer has a technical advantage.

  FIG. 53 is a block diagram illustrating the ripple force calculation unit 700 in the present embodiment. As shown in FIG. 53, the ripple force calculation unit 700 includes a supplier parts delivery information source selection unit 701, a destination company extraction unit 702, a feature word enumeration unit 703, a short-term term extraction unit 704, and a ripple force calculation unit. 705. The ripple force calculation unit 700 estimates the influence range and influence of an information source using a technical dictionary. When a feature word included in the information source is registered in the technical dictionary, another feature word at a distance close to the feature word is listed. When the target company included in the information source has developed products related to the listed feature words in the past, a ripple score is calculated from the technical distance. Each block will be described in detail.

  The parts delivery information source selection unit 701 extracts only the parts delivery information source tree of the supplier from the information source tree. The delivery destination company extraction unit 702 extracts a delivery destination company of a part, that is, a competitor company that purchased the part, and extracts news of products released by the competitor company. The feature word enumeration unit 703 lists all the feature words included in the product release information of competitors. The short-term term extraction unit 704 checks whether a technical term including a feature word of a part exists in the technical dictionary, and enumerates all terms having a certain threshold value or less. The ripple force calculation unit 705 compares the word set enumerated by the feature word enumeration unit 703 and the word set enumerated by the short-range term extraction unit 704, and if there is a match, there is a possibility that the delivered part has been adopted for the product Is determined to be high. The ripple force calculation unit 705 calculates the ripple force obtained by multiplying the reciprocal of the distance of the technical term extracted by the short-term term extraction unit 704 (the ripple coefficient) by the information source value obtained from the parts delivery company strength measurement tree. Add to the initial score of the functional feature leaf that makes up the force measurement tree.

  As a result, by following the past information source tree, it is possible to reach a ripple score to companies that do not directly describe the influence range of the current information source. For example, in the case of parts supplier news, an initial score is calculated by a closed system called a parts supplier company tree (FIG. 45), whereas a mechanism for giving a ripple score to a competitor company tree (FIG. 44) is introduced. In other words, by assigning the ripple score to the functional feature leaf of the competitor tree, the parts supplier news can affect the leaf value of the competitor tree. By classifying information sources into those that match the business domain, and calculating only the score of the matched information source from the word score and corporate strength score, companies and products with high exposure can be used as new business partners. Can be sorted.

  The operation of the fourth embodiment of the present invention will be described in detail with reference to the flowchart of FIG. As a premise, the values of the business domain dictionary (FIG. 39) and the technical dictionary (FIG. 36) which are propagated as a result of the second embodiment are used.

As source news,
"Shinkawa Electric develops the world's highest level amplification module" A200R "by Raman method"
Is input to the information structuring apparatus 51. The information structuring apparatus 51 executes steps S100 to S600, and determines that this information source is the optical business domain (conforming information source tree) because there is a keyword “Raman” in the technical dictionary. The subject determination unit 501 refers to the subject dictionary shown in FIG. 56 in order to determine whether the subject is a competitor, a delivery company, or a customer company. As a result, Shinkawa Electric is determined to be a delivery company, and is stored in the information source storage unit 503 as a delivery company information source tree.

The information source quantification unit 504 reads the delivery company information source tree in the information source storage unit 503 (step S421), and calculates an initial score (step S700). Quantification is performed based on the parts delivery company strength measurement tree shown in FIG. 45 and the word / node correspondence table shown in FIG. There is a keyword “development” in the information source tree, and it is described from the word / node correspondence table of FIG. 45 that the score is doubled. In addition, the score is distributed in a ratio of 0.4 for the product variation leaf, 0.2 for the development speed leaf, and 0.4 for the component highest performance ranking leaf. All leaves are given an initial value of 1. By referring to the business domain dictionary, it can be seen that there is no A200R product in the past, so it can be seen that it is a new product, and the component variation leaf can be obtained by the following calculation formula.
Component variation leaf value = 2 × Past component variation leaf value × 0.4 = 2 × 1 × 0.4 = 0.8
Development speed leaf value = 2 x Elapsed time from the previous product release (month) / 12 x 0.2 = 2 x 1 x 0.2 = 0.2
Value of component highest performance ranking leaf = 2 × {e ^ {− (order−1)} + 1} × 0.4 = 2 × 2 × 0.4 = 1.6

Next, the corporate strength calculation unit 601 calculates a final score (step S800). Here, the enterprise power calculation unit 601 determines the subject of the information source tree, and calculates the final score only when the subject is a customer company or a delivery company. If the subject is a competitor, the final score is calculated after calculating the ripple force described later. In this example, since it is an information source of a delivery company, its final score is calculated. That is, first, the development ability node score in the delivered company ability measurement tree can be obtained by the following calculation formula.
Development capability = 0.8 + 0.2 + 1.6 = 2.6
Assuming that the customer convenience and creditworthiness are 1 at the initial value, and the coefficients α, β, and γ of the intermediate nodes are also 1, the company power value of the parts delivery company is
Value of parts supplier = customer convenience + creditworthiness + development ability = 1 + 1 + 2.6 = 4.6
It becomes. Before the input of the information source, the value of the parts delivery company was 3 in total, so an increment of 4.6-3.0 = 1.6 is obtained as the score (value) of the information source.

As news sources,
"Shinkawa Electric Co., Ltd. Delivered Optical Amplifier Module" A200R "to Opt Sales"
Is input to the information structuring apparatus 51, the information structuring apparatus 51 executes steps S100 to S400 to obtain the information source tree shown in FIG. Since there is a keyword “optical amplification” in the technical dictionary, it is determined that this information source is the optical business domain (compatible information source tree) (steps S500 and S600). The subject determination unit 501 refers to the subject dictionary shown in FIG. 55 in order to determine whether the subject is a competitor or a delivery company. As a result, Shinkawa Electric is determined to be a delivery company, and is stored in the information source storage unit 503 as a delivery company information source tree.

  Similarly, the information source quantification unit 504 calculates an initial score, and the enterprise power calculation unit 601 calculates a final score based on the initial score when the subject is not a competitor (steps S700 and S800). That is, the subject determination unit 501 refers to the subject dictionary shown in FIG. 34, determines that the subject “Shinkawa Electric” is a delivery company, and stores it in the information source storage unit 503 as a delivery company information source tree. Then, the information source quantification unit 504 reads this delivery company information source tree, and based on the delivery company power measurement tree shown in FIG. 45 and the word quantification map shown in FIG. An initial score is calculated (step S700). The enterprise power calculation unit 601 calculates the score of each node based on the updated initial score (step S800).

  Next, the ripple force calculation unit 700 calculates a ripple score that is multiplied by the score of the functional feature node of the business force tree of the competitor (step S900). The information source tree data similar to the information source value estimation unit 500 is input to the ripple force calculation unit 700. FIG. 57 is a flowchart showing details of the ripple force calculation method. First, the parts delivery information source selection unit 701 selects an information source tree of a delivery source company (delivery company) from a set of applied information source trees (step S421). Next, the delivery destination company extraction unit 702 determines whether or not there is an information source of product news related to the delivery destination company (competitor) of the subject of the extracted information source tree (delivery source company). A search is made at 303 (step S422). Here, it is searched whether or not there is product news related to “Opt sales” to which “Shinkawa Electric” is delivered. Then, it is discovered that there is an information source of “Opto Sales develops lossless optical transmission device“ FX5000 ”using μ laser.” Then, the feature word enumeration unit 703 extracts the information source extracted in S422 ( Feature words are extracted from (product news) (step S424) Here, feature words of “μ laser” and “lossless transmission device” are extracted.

  On the other hand, the short-range term extraction unit 704 refers to the technical dictionary in the technical dictionary storage unit 402, and whether there is a word that partially matches or completely matches the feature word included in the information source tree of the supplier company selected in S421. Whether or not is searched (step S423). Here, it is found whether the “light amplification module” in the information source tree exists in the technical dictionary shown in FIG. 36, and it is found that the “light amplification” node exists as a child node of the “DWDM” node. Thereby, it is found that the improvement of optical amplification becomes the improvement of DWDM. That is, in the technical dictionary, the upper element is composed of lower elements, so if there is an excellent part in the lower element, it leads to the differentiation of the products of the upper element. Then, a word whose distance from the partially matched or completely matched word is not more than a predetermined threshold is extracted (step S425). Here, assuming that a distance of 1.5 or less is extracted from “optical amplification”, “DWDM” (distance 0.3), “optical access” (distance 0.6), and “optical transmission device” (distance 1. Three of 3) are extracted.

The ripple force calculation unit 705 compares the feature words enumerated by the feature word enumeration unit 703 with the technical words enumerated by the short-distance term extraction unit (step S426), and whether or not there is a partial or complete match. (Step S427), and if there is a match, the propagation score is calculated. Here, in the “lossless transmission apparatus” enumerated in step S424 and the “optical transmission apparatus” enumerated in step S425. Matches with "Transmission device". In the case of a match, the reciprocal of the distance between the matched description word and the technical term that matches the word in the information source tree is the ripple coefficient. The ripple coefficient is multiplied by the value of the information source of the delivery company, and this value is added to the score of the functional feature leaf in the enterprise strength measurement tree (FIG. 44) of the delivery destination company (competitor company). That is, in step S428, the ripple force calculation unit obtains the reciprocal of the distance 1.3 between the “optical amplification” node and the “optical transmission device” node≈0.77.
Ripple force = Functional feature of competitor measurement tree Increase score of leaf = Ripple coefficient × Score (value) of information source of parts supplier company = 0.77 × 1.6 = 1.232

  In this way, the initial value 1 of the functional feature leaf is updated by adding the ripple force 1.232, and based on the updated initial score, the enterprise power calculation unit 601 calculates the score of the information source. Although the present embodiment has been described on the assumption that the ripple force is calculated from the conforming information source tree, the ripple force calculating unit 700 refers to the technical dictionary or the business domain dictionary, and the conforming information source is derived from the information source tree or the information source. A tree or compatible information source may be extracted.

  In this example, by using the subject company dictionary, it is possible to measure the technical superiority of the parts supplier because it distinguishes between the producer goods company and the parts manufacturer company and estimates the performance ranking from the feature words of the parts. It has become. In addition, since the technical superiority news of parts and the news of the delivery destination (competitor) are evaluated in combination, the technical superiority of the companies adopting the parts can be updated. Furthermore, it is possible to increase the functional feature leaf of competitors by the ripple effect and influence the corporate strength of competitors.

Fifth Embodiment of the Invention Next, a fifth embodiment of the present invention will be described. In this embodiment, the best selling products are listed from both the customer's production goods purchase history and the competitor's production goods sales history, and the replacement demand period is estimated. While the fourth embodiment described above focuses on companies and measures corporate power such as which companies are actively active, in this embodiment, the focus is on products. Knowing what products are being sold, for example, it is possible to analyze what kind of viewpoint the products are made with when developing products together with information on what kind of companies are active in the fourth embodiment. .

  FIG. 58 is a diagram showing a demand cycle in the present embodiment. If time is taken on the horizontal axis and demand is taken on the vertical axis, the demand cycle of a product can be expressed. Here, from the information source, the quantity D1 of a certain product P delivered by a competitor and the quantity D2 of a customer company purchased a certain product are extracted and added together, so that the demand for the product P at that time can be known. In addition, when adding, the case where the delivery destination of a competitor is the customer company which purchased is excluded. In addition, it is possible to determine which company is actively working from the delivery record and purchase record, and in this embodiment, it is possible to display these information together on the display unit.

  FIG. 59 is a diagram showing an information structuring apparatus 51 according to the fifth example of the present invention. As shown in FIG. 59, the information structuring apparatus 51 has a configuration in which a demand cycle estimation unit 800 is further provided as the quantification unit 52 in the information structuring apparatus 41 in the fourth embodiment. Other configurations are the same as those of the fourth embodiment, and a detailed description thereof will be omitted. In the present embodiment, the demand for goods is estimated and the estimation result is reflected in the periodic leaf of the customer enterprise strength measurement tree shown in FIG.

  FIG. 60 is a block diagram illustrating the demand cycle estimation unit 800 in the present embodiment. As shown in FIG. 60, the demand cycle estimation unit 800 includes a product adoption information selection unit 801, a product delivery information selection unit 802, a duplication detection unit 803, a monthly cumulative counter 804, and a product demand display unit 805.

  The demand cycle estimation unit 800 operates such that news of customer companies and competitors generates data for predicting demand in the entire market. A value obtained by applying the sine function to the gradient of the graph in the analyzed month is adopted as the value of the periodic leaf of the customer enterprise strength measurement tree.

  In other words, the product adoption information selection unit 801 and the product delivery information selection unit 802 of the demand cycle estimation unit 800 respectively provide customer production goods purchase / recruitment news and competitors' production goods sales / delivery result news as information source storage units. Extract from 503. The duplicate detection unit 803 extracts the adopted product and the delivery source included in the information source tree extracted from the information source storage unit 503 by the product adoption information selection unit 801, and the product delivery information selection unit 802 extracts from the information source storage unit 503. The delivery product and delivery destination included in the information source tree are extracted, and it is determined whether or not the product names match and the delivery destination and delivery source match. The determination as to whether or not they are the same product can be made based on whether or not they include feature words below the child node (DWDM or SONET / SDH) of the top node of the technical dictionary. If they match, the product delivery amount (adopted amount) is subtracted from the total delivered amount (adopted amount). The monthly cumulative counter 804 plots the time transition on the horizontal axis and the sales results of the products classified into the same product group on the vertical axis based on the detection result of the duplication detection unit 803, and the product demand as shown in FIG. Generate a periodic graph. As a result, the trend of replacement demand for production goods that occurs periodically is predicted and displayed on the product demand display unit 805 to determine whether the demand of a certain product group has decreased or increased. Can be presented every time.

  Furthermore, the enterprise strength calculation unit 601 replaces the numerical value of the demand gradient with the initial score of the period of the customer enterprise strength measurement tree. The demand cycle estimation unit 800 can list the best-selling products and estimate the replacement demand period from both the customer's production goods purchase history and the competitor's production goods sales history. Conventionally, it has been difficult to understand how many products of competitors are selling, and it has been difficult to understand what kind of demand there is because the customer's reaction is not directly visible. To solve.

The operation of the fifth embodiment of the present invention will be described in detail with reference to the flowchart of FIG.
"NEE adopts" FX4200 "sold by Opt for US-Japan backbone line"
The news arrives. Since NEE is a customer company, the product adoption information sorting unit 801 detects the product adoption information source tree from the information source storage unit 503 (step S431), and lists the adopted products and the delivery source (step S432). . In this example, the adopted product is “FX4200” and the delivery source is “Opt sales”. Next, the product delivery information selection unit 802 creates a product delivery information source tree of competitors from the past information sources stored in the information source storage unit 503.
"Opto Sales Deliver" FX4200 "with Raman Method for Optical Amplification to NEE"
Are detected (step S433), and the delivery destinations of the delivered goods are listed (step S434). In this example, the delivered product is “FX4200” and the delivery destination is “NEE”.

  Next, the duplicate detection unit 803 determines whether the product name, delivery destination, and delivery source match. If all match, it is determined that the two information sources are duplicated. In this example, since the product names match, the two pieces of information are added to the monthly demand as one sale (product sales) (step S436). If none of the product name, delivery destination, and delivery source match, the monthly demand is added as a separate supply / demand (step S438). Similarly, sales of products are recorded on a monthly basis based on customer company recruitment information and competitors' delivery information. For example, suppose that April has 5 sales, May has 6 sales, and June has 5 sales. In this case, a Bezier curve is drawn using the sales results (demand) for three months from April to June (step S437). In this example, the increase rate in May is sin (BezierFunction (5,6,5)) = sin (0) = 0. That is, it can be estimated that the demand tends to be flat. This Bezier curve is displayed by the commodity demand display unit 805.

  In the present embodiment, the demand period is estimated using the conforming information source tree. However, the demand period estimating unit 800 refers to the domain dictionary or the technical dictionary and adapts from the information source tree or the information source. An information source tree or a compatible information source may be extracted.

  In this embodiment, the demand for goods can be estimated by combining information about customers and information about competitors and using equivalence of information sources. That is, it is possible to list the best-selling products and estimate the replacement demand period from both the customer's production goods purchase history and the competitor's production goods sales history.

Sixth Embodiment of the Invention Next, a sixth embodiment of the present invention will be described. FIG. 62 is a diagram showing an information structuring apparatus 61 according to the sixth example of the present invention. In the present embodiment, the potential purchasing power of a customer is estimated from the financial statements and the customer's production goods purchase history, or the sales performance of competitors, and a good customer is estimated. As shown in FIG. 62, the information structuring apparatus 61 according to the present embodiment has a configuration in which a potential purchasing power estimation unit 900 is further provided as the quantification apparatus 62 in the information structuring apparatus 51 in the fifth embodiment. Other configurations are the same as those of the fifth embodiment, and detailed description thereof is omitted.

  FIG. 63 is a block diagram showing a potential purchasing power estimation unit 900 in the present embodiment. As shown in FIG. 63, the potential purchasing power estimation unit 900 is configured together with the product adoption information selection unit 801, the product delivery information selection unit 802, and the duplication detection unit 803 of the demand cycle estimation unit 800 in the fifth embodiment. The potential purchasing power estimation unit 900 further includes another company product feature word selection unit 904, own company similar product selection unit 905, other company product adoption customer selection unit 906, financial information acquisition unit 907, investment performance calculation unit 908, investment similarity calculation unit 909. , A customer group selection unit 910 and a non-adopted customer selection unit 911.

  The potential purchasing power estimation unit 900 calculates the asset margin from the customer's financial statements and superimposes the purchase histories of the production goods, so that it is the customer who purchases the production goods with what characteristics at what fund level. Presenting purchasing characteristics. Based on the investing attitude of the client company and the products actually purchased, the customer company that sells its own products is indexed, and the corporate strength renewal department adds it to the product matching characteristics of the customer company. As a result, it is possible to rank good customers for the company, and to estimate the potential purchasing power of the customer from the financial statements, the customer's production goods purchase history, or the sales performance of competitors. This is a means for solving the problem that it has been difficult to understand which of the potential customers who have the capacity to purchase production goods and the customers who have traditionally had business relationships.

  FIG. 64 is a diagram showing the characteristics of products with delivery / purchase results of each company. In FIG. 64, it is generally assumed that the higher the function, the higher the price. A company that acquires structured information using the information structuring device 61 is referred to as company A. Company A is assumed to belong to a competitor. The customer a manufactures a product with a lower function and a lower price, and the customer b manufactures a product with a lower function and a lower price. It is assumed that competitors B and C manufacture products with different price ranges and functions from company A, and each has a delivery record. Here, the product cannot be sold unless the product of the function and price required by the client company is manufactured. In addition, a product similar to a product that has already been manufactured and delivered by a rival company is already manufactured and is unlikely to be sold. In other words, it is highly possible that a product that has a function and price required by a client company and that is not manufactured by another competitor can be sold. For example, when company A plans to manufacture a high-performance, high-priced product, analysis is performed using this information structuring device 61, so that the function and price are more likely to be sold. You can review the product.

  Returning to FIG. 63, the product adoption information selection unit 801 selects the product adoption information source tree of the customer company from the information source tree. The product delivery information selection unit 802 selects a product delivery information source tree of a competitor from the information source tree. The duplicate detection unit 803 calculates the equivalence between the product adoption information source tree and the product delivery information source tree of the customer company, and when it is determined that they are the same, discards one of the information source trees.

  Next, another company's product feature word selection unit 904 extracts product feature words that have been adopted by customers from the information source tree. The in-house similar product selection unit 905 determines whether the in-house product has a feature word of a product that has been adopted by a competitor, and if so, picks up the in-house product. The third-party product adoption customer selection unit 906 extracts a customer list from the subject dictionary and lists, by product, customer companies that have adopted the third-party product from the information source tree obtained from the duplication detection unit 903. On the other hand, the financial information acquisition unit 907 acquires financial information published by the client company. For example, it can be obtained by extracting financial information from a company's financial report published every quarter or half year, or by extracting financial information contained in an information source. Or you may make it input from the outside. An investment performance calculation unit 908 calculates the amount of investment in fixed assets. The investment similarity calculation unit 909 sorts the client companies according to the standard (investment attitude) obtained from fixed assets / current assets. The customer group selection unit 910 receives a customer list that has purchased another company's products and an investment attitude list of all customers, and selects customers who have similar investment attitudes and have not purchased other company's products. In this case, enumerate all the companies before and after the company that adopted the products of other companies.

  The customer selection section 911 that has not yet adopted another company's product also compares the feature word of the other company's product that has been delivered with the feature word of the company's product, and determines the functional similarity (number of parts / perfect match words / number of other company's product The number of feature words is calculated. In addition, the investment attitude similarity is calculated by 1−e ^ (− (/ | fixed asset / current asset−fixed asset / current asset |)), and the multiplied values are shown in FIG. The initial score of the product matching leaf.

  FIG. 65 is a flowchart illustrating the operation of the information structuring apparatus according to the present embodiment. FIG. 66 is a table created by the information structuring apparatus of this embodiment. In the present embodiment, it is estimated what products are likely to be purchased from all registered customer companies, and companies that are highly likely to purchase the products are selected. First, the financial information acquisition unit 907 acquires a customer company list from the subject dictionary shown in FIG. 34, and obtains financial statements of the customer companies, for example, “ATT”, “real communication”, and “OVOP” (step S445). Specifically, the financial information acquisition unit 907 acquires fixed assets and current assets of each customer company. The fixed assets are 1 billion, 4 billion and 3 billion, respectively, and the current assets are 1 billion, 6 billion and 5 billion. Next, the investment performance calculation unit 908 obtains the investment performance ratio of each company to the fixed assets as fixed assets / current assets. Here, 1.0, 0.67, and 0.6 are obtained (step S446). Then, when the investment similarity calculation unit 909 sorts each customer company by the value of the investment attitude (step S447), as shown in FIG. 66, “OVOP (0.6)” “Real communication (0.67)”. The order is “ATT (1.0)”.

On the other hand, as delivery / adoption news of other companies, the product delivery information selection unit 902 and the product adoption information selection unit 901 respectively.
"Orion delivers" OR2500 "optical access product to OVOP",
"ATT adopts Nikko Hikari Metro product" BB-P012 ""
Is newly selected from the information source storage unit 503 (S441). After duplication inspection of these information source trees by the duplication detection unit 903, the “other products” feature word selection unit 904 extracts “optical access product” as the former feature word and “metro product” as the latter feature word. (S442). Next, the self-similar product selection unit 905 receives the release news of past self-product from the information source storage unit 503.
"NEC releases optical access product" DW4000 ""
And “DW4000” is extracted. The in-house similar product selection unit 905 compares the feature word of DW4000 with the feature words of OR2500 and BB-P012, and determines that DW4000 is similar to a product that has been delivered by other companies. Thereby, it is estimated that DW4000 is a product that is easy to deliver (step S443).

  Next, the other company's product adoption customer selection unit 906 stores “OVOP” and “ATT” as a delivery destination company having a delivery record. Further, since there is no news adopted by the real communication and there is no news delivered by the competitor, it is determined that the “real communication” has no record (S444). 66. The customer set sorting unit 910 receives the result sorted by the investment similarity calculation unit 909 and the product information adopted by the other company product adoption customer sorting unit 906, and creates a table as shown in FIG. From this table, it can be seen that “real communication” has an investment performance ratio of 0.67, and the investment attitude is similar to “OVOP”. The customer set sorting unit 910 extracts a customer company having a purchase record in this table and one company before and after that, and inputs it to the non-commercial product non-adopted customer sorting unit 911. In the case of this example, information of all customer companies corresponds to this. The other company product non-adopted customer selection unit 911 is supplied with product information having characteristics similar to those of other company products that have been used by the in-house similar product selection unit 905, and the column of whether or not the company can be manufactured in the table is described.

  As a result, it is understood that “OVOP” has already purchased an optical access product having the same characteristic word as that of its own product. On the other hand, real communication has no product purchase record and is similar in investment attitude to OVOP. In other words, real communication has a high possibility of adopting a product adopted by OVOP having a similar investment attitude. Therefore, “real communication” becomes a promising company as a sales candidate of “DW4000”, which is its own product. The other company product non-adopted customer selection unit 911 extracts another customer company that has a similar investment attitude to that of the customer company that has adopted such other company's product and has not yet adopted the other company's product, and a display unit that is not shown in the drawing. (Step S911).

Further, the other company product non-adopted customer selection unit 911 calculates the value of the product matching leaf of the customer company strength measurement tree shown in FIG. 43 of the customer company “real communication” as follows.
Functional similarity = (number of parts / perfect match words / number of previous feature words of other company's products and own products) = 1/1
Investment attitude similarity = 1−e ^ (− (1 / | fixed asset / current asset−fixed asset / current asset |)) = 1−e (− (1 / | 40 / 60−30 / 50 |)) = 1−e ^ (− 0.07) ≈1
Product matching leaf score = functional similarity x investment attitude similarity = 1
As a result, the initial score of product matching in the customer enterprise power tree is updated to 1.

  In this embodiment, the potential purchasing power is estimated using the matching information source tree. However, the potential purchasing power estimation unit 900 refers to the domain dictionary or the technical dictionary and matches from the information source tree or the information source. An information source tree or a compatible information source may be extracted. In addition, since the potential purchasing power estimation unit 900 compares the characteristic words of the company's product and the other company's product, the present embodiment can be applied without separating the information source into a conforming information source or a non-conforming information source.

  According to the present embodiment, by using fixed assets and current assets in the financial statements publicly disclosed by the client companies, the similarity of the customer's investment attitude is calculated, and the customer and competitors deliver.・ Listing companies that have yet to be delivered that are likely to accept the company's products from the recruitment results news, making it possible to estimate customers who are likely to sell their products. In addition, the corporate strength of the client company can be determined more accurately.

Seventh Embodiment of the Invention Next, a seventh embodiment of the present invention will be described. FIG. 67 is a diagram showing an information structuring apparatus 71 according to the seventh example of the present invention. In this embodiment, the development history and sales history of production goods are combined to estimate the profitability of a competitor. As shown in FIG. 67, the information structuring apparatus 71 has a configuration in which a sales force estimation unit 1000 is further provided as a quantifying apparatus 72 in the information structuring apparatus 61 in the sixth embodiment. Other configurations are the same as those of the sixth embodiment, and a detailed description thereof will be omitted. The seventh embodiment is an example in which the sales force of a competitor is estimated.

  68 (a) and 68 (b) are graphs showing the relationship between sales force and the period until sales, and the relationship between sales force and function, respectively. As shown in FIG. 68, in general, it can be seen that the sales force is higher when the sales time-development completion (release) time is shorter. Also, in general, when a function is low but sold, it can be seen that the sales power is higher than when the function is high and sold. The function may be a function per unit price or a price. If there are few functions per unit price and it sells, or if the price per unit price is high, it can be seen that the sales power is high. In the present embodiment, the sales force of a company is estimated based on such product function and period information.

  FIG. 69 is a block diagram showing the sales force estimation unit 1000 in the present embodiment. As shown in FIG. 69, the sales force estimation unit 1000 is configured together with the product adoption information selection unit 801, the product delivery information selection unit 802, and the duplication detection unit 803 of the demand cycle estimation unit 800 in the fifth embodiment. The sales force estimation unit 1000 further includes a product-specific feature word collection unit 1004, a product development information selection unit 1005, an average feature word number calculation unit 1006, and a competitive product sales rate calculation unit 1007.

  The sales force estimation unit 1000 operates to generate data for predicting the sales force of a competitor, and the sales force is estimated based on the product development release news of the competitor, the delivery history news of the product, or the customer adoption history news. Calculate

  First, the product-specific feature word collection unit 1004 collects information sources for each competitive product, counts the feature words included in each information source, and obtains the cumulative feature word number αi (i is an arbitrary product) for each product. The average feature word number calculation unit 1006 calculates the average feature word number β per product from the cumulative feature word number and the product number of all products. By obtaining the average number, it is possible to determine whether the features collected by the product-specific feature word collection unit 1004 are more or less for products in the industry.

  Next, the product adoption information sorting unit 801 and the product delivery information sorting unit 802 respectively extract only the production product purchase / recruitment news of the customer and the production product sales / delivery result news of the competitors from the information source storage unit 503. . The duplicate detection unit 803 extracts the adopted product and the delivery source included in the information source tree extracted from the information source storage unit 503 by the product adoption information selection unit 801, and the product delivery information selection unit 802 extracts from the information source storage unit 503. The delivery products and delivery destinations included in the information source tree are extracted, and the equivalence between the delivery news and the recruitment news is determined by determining whether the product names match and whether the delivery destination and the delivery source match. to decide. Discard any duplicates. The product development information selection unit 1005 receives the release time and the sales amount for each product from the duplication detection unit 803, and calculates the ratio γi = (number of sales news) / (delivery time−release time). This corresponds to FIG.

  The competitive product sales rate calculation unit 1007 calculates the average feature word number / cumulative feature word number × adopted news number / (delivery time−from the average feature word calculation unit 1006 and αi, β from the product development information selection unit 1005. The sales rate (sales power) calculated by the release time is calculated. β / α represents a product having a smaller function, that is, less attractive as the value is larger. When β / α × γi is large, it indicates that the period until the sale is short in spite of being unattractive, that is, a value indicating that the sales force is high.

  The enterprise strength calculation unit 601 adds this index to the sales performance leaf of the competitive enterprise strength measurement tree. As shown in FIG. 68 (a), how efficiently the developed product is linked to sales by plotting the time from development to sales on the horizontal axis and the sales record on the vertical axis. Can be presented by a display unit (not shown). As a result, it is possible to link the development history of the production goods and the sales performance history and estimate the profitability of the competitor. In this embodiment, the conventional problem that it is difficult to understand the difference between the product development capability of competitors and the product development capability of the company is solved.

Next, the operation of this embodiment will be described in detail with reference to the flowchart of FIG. First, the product-specific feature word collection unit 1001 measures the number of feature words for the release news of each competitive product.
"Opto Sales develops" FX5000 "lossless optical transmission device using μ laser"
Is stored in August, the product-specific feature word collection unit 1001 collects two (αi = 2) of “μ laser” and “lossless optical transmission device” as feature words. Thus, the number of feature words is collected for all competing products (step S451). Then, the average feature word calculation unit 1002 divides the number of feature words by the number of competing products that have been investigated (step S452). In this case, it is assumed that a value of average β = 1.4 is obtained. next,
"Opt sales, delivering" FX5000 "to ATT",
"Optical sales, lossless optical transmission" FX5000 "delivered to real communications"
If the news is added and stored in October and November, respectively, product information is extracted from this information source by the product adoption information selection unit 801, the product delivery information selection unit 802, and the duplication detection unit 803. Then, the number of delivery / adoption news is calculated for each competing product (step S454).

Next, the product development information selection unit 1005 sets the ratio γi = number of sales news / (delivery time−development time), where i is the product number.
Is calculated. In this example, the number of sales news delivered in October and November is 1 each.
Ratio γi = 1 / (10−8) + 1 / (11−8) ≈0.5 + 0.3 = 0.8
(Step S455). The competitive product sales rate calculation unit 1007
Sales force = Σβ / αi × γi
Ask for. In this example, the sales rate of “FX5000” is obtained as 1.4 / 2 × 0.8 = 0.56. As a characteristic of business sales power, there is a tendency that if a product with no features is developed and delivered immediately, it becomes higher. The competitive product sales rate calculation unit 1007 updates the value of the sales performance leaf of the competitive enterprise strength measurement tree with the value of the sales power.

  In this embodiment, the sales force is estimated by using the compatible information source tree. However, the sales force estimation unit 1000 refers to the domain dictionary or the technical dictionary and matches the information from the information source tree or the information source. An information source tree or a compatible information source may be extracted. Further, since the sales force estimation unit 1000 uses the average number of feature words of the product, the present embodiment can be applied without separating the information source into the conforming information source or the non-conforming information source.

  In this embodiment, the sales force of a competitor can be estimated by using the number of features of the developed product, its average value, the number of sales news, and the period from development to delivery of the developed product as indices. Further, by setting the obtained sales force as the value of the sales performance leaf in the competitive enterprise strength measurement tree, the competitive enterprise strength can be updated to a value taking into account the sales capability, and can be accurately estimated.

Eighth Embodiment of the Invention Next, an eighth embodiment of the present invention will be described. FIG. 71 is a diagram showing an information structuring apparatus 81 according to the eighth example of the present invention. In the present embodiment, the product produced by the customer and the parts delivered to the product development maker are linked to estimate which part is the competitiveness of the product. As shown in FIG. 71, the information structuring apparatus 81 has a configuration in which a competitiveness enhancement component estimation unit 1100 is further provided as the quantification apparatus 82 in the information structuring apparatus 71 in the seventh embodiment. Other configurations are the same as those of the seventh embodiment, and detailed description thereof is omitted. The eighth embodiment is an example in which excellent parts are estimated.

  FIG. 72 is a block diagram showing a competitiveness enhancement component estimation unit 1100 in the present embodiment. As shown in FIG. 72, the competitiveness-enhancing component estimation unit 1100 includes a component adoption rate higher-ranking company selection unit 1101, a competitor company product delivery performance measurement unit 1102, and a component contribution rate calculation unit 1103. The competitiveness enhancement component estimation unit 1100 uses the output results of the product adoption information selection unit 801, the product delivery information selection unit 802, and the duplication detection unit 803 described above.

  The competitiveness-enhancing part estimation unit 1100 uses a part delivery result news or a competitor's part adoption news and a competitor's product sales record or a customer's product adoption record to incorporate a part into a competitor's product, Calculate the rate of final sales to customers. This index is input by the competitiveness-enhancing component estimation unit 1100 to the enterprise strength estimation unit 600 to be added to the initial score of the differentiated leaf of the delivered enterprise strength measurement tree. In addition, by plotting the parts delivered on the horizontal axis and the delivery results to customers on the vertical axis, it is shown which parts are the source of competitiveness. As a result, it is possible to link the product produced by the customer and the parts delivered to the product development maker and estimate which parts are competitive with the product. This solves the conventional problem that it is difficult to understand the parts with high technological advantages adopted by other companies' products, and the products used by customers and the parts delivered to the manufacturers Present mapping and extract highly competitive parts.

  The part adoption rate top company selection unit 1101 selects the top two companies having a high part occupancy ratio of the supplier from among the competitors adopting the parts of the supplier. For example, a case will be described in which the part occupancy rate is calculated from the sales performance news of the parts supplier A1 and the adoption news of the supplier competitor B1. For example, if the supplier competitor B1 releases the adoption of the parts of the supplier A1 three times and the supplier competitor B1 releases the adoption of the parts of another supplier A2 once, the parts supplier A1 It can be seen that the competitor with a high component occupancy is the supplier competitor B1.

Competitor product delivery performance measurement department 1102 determines the adoption rate of products developed by each company to customer companies,
Adoption rate α = calculated by the number of products used / the number of products developed. Even if the product is delivered to the competitor, if the product using it is not delivered to the customer company, the part is not adopted.

The part contribution rate calculation unit 1103 calculates the contribution rate Pi = 1 / the number of delivered parts × the adoption rate from the relationship between the parts delivered to the competitor, where i is the product type. The contribution rate is calculated in the same way for the remaining top one company and added to the previous contribution rate. Calculate all parts owned by the supplier. The calculation result is input to the enterprise power calculation unit 601 and adopted as the initial score of the differentiated leaf of the component delivery enterprise power measurement tree.

The operation of the eighth embodiment of the present invention will be described in detail with reference to the flowchart of FIG. 73 and the news relation diagram of FIG. News to use
"Shinkawa Electric develops high-concentration μ laser",
"Shinkawa Electric develops small optical amplification module",
"Shinkawa Electric starts supplying optical cross-connect (XC) parts"
3 news related to the development of Shinkawa Electric,
"Shinkawa Electric supplies μ laser to Orion",
"Orion fully adopts Shinkawa Electric's optical amplification module"
Two parts supply news for Orion, a competitor
"Shinkawa Electric Co., Ltd. Delivered μ Laser to Nippon Optical Transmission",
"Shinkawa Electric Co., Ltd. delivers optical amplification module to Nippon Optical Transmission",
"Shinkawa Electric Co., Ltd. delivers XC parts to Nippon Optical Transmission"
3 parts supply news for Nippon Kogyo, a competitor
"Shinkawa Electric supplies μ-laser to Sakura-dori"
One part delivery news for Sakura Tsuru, a competitor
"Orion delivers metro products to carrier ATT",
"Real communication uses Orion's DWDM part"
Two news stories about the adoption of Orion products
"Nippon Hikari Transmission and Metro products delivered to OVOP"
One news about the adoption of Nikko transmission products of a customer company,
"Sakura-Dori, Access product AX-1000 delivered to ADDI"
This is one news story about the adoption of Sakura products by a customer company. These news related diagrams are shown in FIG.

The component adoption rate higher-level company selection unit 1101 lists competitors as delivery destinations for each delivered part (step S461). In the above example, “Orion”, “Nippon Optical Transmission”, and “Sakura-dori” are listed as the competitors of the delivery destination. The companies that deliver the most parts from "Shinkawa Electric" are from the top, "Nippon Hikari Transmission" (3), "Orion" (2), "Sakura-dori" (1). The unit 1101 selects the top two companies (step S462). The competitor product delivery record measuring unit 1102 extracts delivery record news for the top two customers (step S463). In this example,
"Orion delivers metro products to carrier ATT",
"Real communication uses Orion's DWDM part"
2 cases, and
"Nippon Hikari Transmission and Metro products delivered to OVOP"
It becomes.

  On the other hand, the competitor product delivery record measuring unit 1102 calculates the number of adoption record news to the customer for each part (step S464). That is, the paths for parts to reach customers through competitors are listed. Regarding “μ laser”, two cases of Orion → ATT / real communication and one case of Nippon Optical Transmission → OVOP are applicable. Similarly, there are 3 cases of light amplification and 1 case of XC. Assuming that there are three types of Orion's own products and two types of Nihon Kohtsu's own products, there are two types of products that Orion has received in its customers. ) = Number of products employed / number of products developed = 2/3, α (Nippon Optical Transmission) = 1/2 is obtained (step S465).

Next, the component contribution rate calculation unit 1103 calculates the contribution rate Pi = α × (1 / all components) for each component that has a track record in the competitors (step S466). In this example, since there are two products delivered to “Orion”, “μ laser” and “optical amplification”, the component contribution rate calculation unit 1103 obtains the component contribution rate Pi as follows. .
Pi (μ laser) = (2/3) × 1/2 = 1/3,
Pi (optical amplification) = (2/3) × 1/2 = 1/3
Similarly, there are three types of parts delivered to Nihon Kohtsu,
Pi + 1 (μ laser) = (1/2) × 1/3 = 1/6,
Pi + 1 (optical amplification) = (1/2) × 1/3 = 1/6,
Pi + 1 (XC) = (1/2) × 1/3 = 1/6
It becomes. When the contribution rate calculated for each company is totaled for each product,
P (μ laser) = 1/3 + 1/6 = 1/2,
P (light amplification) = 1/2,
P (XC) = 1/6
(Step S467). Therefore, when products are sorted by this value, the order is “μ laser”, “optical amplification”, and “XC”, and the recommended product of Shinkawa Electric is “μ laser” or “optical amplification”.

  In the present embodiment, description has been made on the assumption that the competitiveness enhancing component is estimated using the matching information source tree. However, the competitiveness enhancing component estimation unit 1100 refers to the domain dictionary or the technical dictionary, and the information source tree or You may make it extract a suitable information source tree or a suitable information source from an information source.

  In this embodiment, the relationship between related news and delivery results is indexed in the range of parts suppliers, competitors, and customer companies, so that it is possible to estimate superior parts based on results. In addition, the calculated contribution rate is replaced with the initial score of the differentiated leaf in the component delivery company strength measurement tree, thereby making it possible to estimate the enterprise strength reflecting the quality of the sales performance of the components.

The first to eighth embodiments of the present invention described above have the following effects. That is,
The first effect is that it is possible to automatically determine whether an unknown product is a competing product that matches the business domain. The reason for this is that the information source is decomposed into the subject, action, product name, and feature word, and compared with other information sources using business domain dictionaries and technical dictionaries. This is because it can be presented as a purchase target. In addition, past product release news history, comprehensive companies that release products that do not match the business domain, and specialists that only release products that match the business domain, are mistakenly classified as new products in different business domains. Is reduced.

  The second effect is that it is possible to find competitors that match the business domain and to distinguish between specialists and general manufacturers. The reason is that the product is matched with the business domain using a universal technical dictionary in the business domain and compared with the feature words of the new product release news, and the name of the developed company can also be obtained. Because. Furthermore, it is because it is possible to measure whether a product handled by a competitor belongs to a business domain and determine whether a product matching the business domain is released at a rate equal to or higher than a threshold.

  The third effect is that a client company that matches the business domain can be found. This is because companies that purchase production goods in the business domain can be listed from past product delivery news and presented together. As a result, it is possible to find a customer who has not made a direct transaction in the past. In addition, by preparing the financial statements of client companies, it is possible to determine which customers are likely to purchase production goods by calculating the level of excellent customers based on the investment attitude of the customers and surplus funds. it can.

  The fourth effect is that keywords related to the business domain can be automatically collected. The reason is that word verbs appearing simultaneously from information sources determined to belong to the business domain can be stored as closely related words. Since feature words of new competitive products are added to the business domain dictionary and the technical dictionary, business domain terms can be automatically expanded, and the number of man-hours required for manual maintenance can be reduced. Furthermore, since the information source inputted later is also compared including the feature word which grew, it becomes possible to discover a product more widely. In addition, by reflecting the keywords that appear simultaneously in the news information source in the distance of the technical dictionary, it is possible to automatically construct a technical word corresponding to the depth of association.

  The fifth effect is that the difference in corporate power among customers, competitors, and parts suppliers can be quantified. The reason for this is that templates for measuring corporate strength are prepared for the three types of customers, competitors, and parts suppliers, keywords included in information sources, word / node correspondence tables, and competitive enterprise strength measurement tree structures, This is because the impact of keywords is calculated from the leaf value calculation formula against the past business activities. Further, since the influence on the company that employs the parts supplier is directly quantified using the technical dictionary, the ripple effect can be quantified more accurately.

  The information structuring process as described above can be realized by a computer such as a dedicated computer or a personal computer (PC). However, the computer does not need to be physically single, and a plurality of computers may be used when performing distributed processing. As shown in FIG. 75, the computer 2100 has a CPU 2101 (Central Processing Unit), a ROM 2102 (Read Only Memory), and a RAM 2103 (Random Access Memory), which are connected to each other via a bus 2104. Although explanation of OS software for operating the computer is omitted, it is assumed that a computer for constructing the information processing apparatus is also provided.

  An input / output interface 2105 is also connected to the bus 2104. The input / output interface 2105 includes, for example, an input unit 2106 made up of a keyboard, mouse, sensor, etc., a display made up of a CRT, LCD, etc., an output unit 2107 made up of headphones, speakers, etc., a storage unit 2108 made up of a hard disk, etc. A communication unit 2109 including a modem and a terminal adapter is connected.

  The CPU 2101 executes various processes according to various programs stored in the ROM 2102 or various programs loaded from the storage unit 2108 to the RAM 2103. In the present embodiment, for example, information source cutting processing, information source reconstruction processing, information source selection processing, group generation processing, information source value estimation processing, ripple power calculation processing, enterprise power estimation processing, demand cycle estimation processing, latency Processing such as purchasing power estimation processing, sales power estimation processing, or competitiveness enhancement component estimation processing is executed. The RAM 2103 also appropriately stores data necessary for the CPU 2101 to execute various processes.

  For example, the communication unit 2109 performs communication processing via the Internet (not shown), transmits data provided from the CPU 2101, and outputs data received from a communication partner to the CPU 2101, the RAM 2103, and the storage unit 2108. A storage unit 2108 communicates with the CPU 2101 to save and erase information. The communication unit 2109 also performs communication processing of analog signals or digital signals with other devices.

  The input / output interface 2105 is also connected to a drive 2110 as necessary. For example, a magnetic disk 2111, an optical disk 2112, a flexible disk 2113, or a semiconductor memory 2114 is appropriately mounted, and a computer program read from them is required. Is installed in the storage unit 2108 accordingly.

  It should be noted that the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the present invention.

  For example, in the above-described embodiments, the hardware configuration has been described. However, the present invention is not limited to this, and any processing can be realized by causing a CPU (Central Processing Unit) to execute a computer program. It is. In this case, the computer program can be provided by being recorded on a recording medium, or can be provided by being transmitted via the Internet or another transmission medium.

It is a figure which shows the information structuring apparatus concerning 1st Example of this invention. It is a block diagram which shows the information source cutting part in the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows the structure of an information source character string. It is a figure which shows the morphological analysis result in the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows the information source reconstruction part in the information structuring apparatus concerning 1st Example of this invention. It is a block diagram which shows the information source selection part in the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows the structure of the domain dictionary which the information structuring apparatus concerning 1st Example of this invention uses. It is a figure which shows an example of the domain dictionary which the information structuring apparatus concerning 1st Example of this invention uses. It is a block diagram which shows the group production | generation part in the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows the technical dictionary based on a product structure. It is a figure which shows the structure of the technical dictionary which the information structuring apparatus concerning 1st Example of this invention uses. It is a flowchart which shows operation | movement of the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows an example of the morphological analysis result in the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows after the information reconstruction part in the information structuring apparatus concerning 1st Example of this invention connected the morpheme. It is a flowchart which shows the product name extraction method in the information structuring apparatus concerning 1st Example of this invention. It is a flowchart which shows the subject extraction method in the information structuring apparatus concerning 1st Example of this invention. It is a flowchart which shows the action extraction method in the information structuring apparatus concerning 1st Example of this invention. It is a flowchart which shows the action target extraction method in the information structuring apparatus concerning 1st Example of this invention. It is a flowchart which shows the synonym extraction method in the information structuring apparatus concerning 1st Example of this invention. It is a flowchart which shows the feature word extraction method in the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows the result which the information source reconstruction part in the information structuring apparatus concerning 1st Example of this invention extracted the subject, the product name, the action, the characteristic word, and the synonym according to the format rule. It is a figure which shows an example of the format analysis result in the information structuring apparatus concerning 1st Example of this invention. It is a flowchart which shows the suitable information source tree selection method in the information structuring apparatus concerning 1st Example of this invention. It is a flowchart which shows the propagation method of the information source tree in the information structuring apparatus concerning 1st Example of this invention. It is a figure which shows the information structuring apparatus concerning the modification of the 1st Example of this invention. It is a figure which shows the information source reconstruction part in the information structuring apparatus concerning the modification of the 1st Example of this invention. It is a figure which shows the information source tree which the information structuring apparatus concerning the modification of the 1st Example of this invention produces | generates. It is a block diagram which shows the information source selection part in the information structuring apparatus concerning the modification of the 1st Example of this invention. It is a block diagram which shows the group production | generation part in the modification of the 1st Example of this invention. It is a flowchart which shows operation | movement of the information structuring apparatus concerning the modification of the 1st Example of this invention. It is a figure which shows an example of the information source tree which the information structuring apparatus concerning 1st Example of this invention produced | generated. It is a figure which shows an example of the information source tree which the information structuring apparatus concerning 1st Example of this invention produced | generated. It is a block diagram which shows the information source selection part in the information structuring apparatus concerning 2nd Example of this invention. It is a figure which shows the structure of the subject dictionary which the information structuring apparatus concerning 2nd Example of this invention refers. It is a figure which shows an example of the business domain dictionary which the information structuring apparatus of this invention expanded. It is a figure which shows an example of the technical dictionary which the information structuring apparatus of this invention expanded. It is a figure which shows an example of the format analysis result in the information structuring apparatus concerning 2nd Example of this invention. It is a figure which shows an example of the format analysis result in the information structuring apparatus concerning 2nd Example of this invention. It is a figure which shows an example of the business domain dictionary which the information structuring apparatus of 2nd Example of this invention propagated. It is a figure which shows the information structuring apparatus concerning the 3rd Example of this invention. It is a block diagram which shows the information source value estimation part in the information structuring apparatus concerning the 3rd Example of this invention. It is a figure which shows the word quantification map which the information structuring apparatus concerning the 3rd Example of this invention refers. It is a figure which shows the structure of the customer enterprise power measurement tree which the information structuring apparatus concerning 3rd Example of this invention refers. It is a figure which shows the structure of the competitive enterprise power measurement tree which the information structuring apparatus concerning 3rd Example of this invention refers. It is a figure which shows the structure of the parts delivery company power measurement tree which the information structuring apparatus concerning 3rd Example of this invention refers. It is a figure which shows the word and node correspondence table contained in the word quantification map which the information structuring apparatus concerning 3rd Example of this invention refers. It is a block diagram which shows the enterprise power estimation part in the information structuring apparatus concerning the 3rd Example of this invention. It is a figure which shows a node structure and a leaf structure. It is a flowchart which shows operation | movement of the information structuring apparatus concerning the 3rd Example of this invention. It is a figure which shows an example of the information source tree which the information structuring apparatus concerning the 3rd Example of this invention produced | generated. It is a figure which shows an example of the information source tree which the information structuring apparatus concerning the 3rd Example of this invention produced | generated. It is a figure which shows the information structuring apparatus concerning the 4th Example of this invention. It is a block diagram which shows the ripple force calculation part in the 4th Example of this invention. It is a flowchart which shows operation | movement of the information structuring apparatus concerning the 4th Example of this invention. It is a figure which shows the subject dictionary which the information structuring apparatus concerning the 4th Example of this invention refers. It is a figure which shows an example of the information source tree which the information structuring apparatus concerning the 4th Example of this invention produced | generated. It is a flowchart which shows the detail of the ripple force calculation method in the information structuring apparatus concerning the 4th Example of this invention. It is a figure which shows the demand cycle of a certain goods. It is a figure which shows the information structuring apparatus concerning the 5th Example of this invention. It is a block diagram which shows the demand period estimation part in the 5th Example of this invention. It is a flowchart which shows operation | movement of the information structuring apparatus concerning the 5th Example of this invention. It is a figure which shows the information structuring apparatus concerning the 6th Example of this invention. It is a block diagram which shows the potential purchasing power estimation part in the information structuring apparatus concerning the 6th Example of this invention. It is a figure explaining the balance of the function and price of a product. It is a flowchart which shows operation | movement of the information structuring apparatus concerning the 6th Example of this invention. It is a table | surface produced with the information structuring apparatus of the 6th Example of this invention. It is a figure which shows the information structuring apparatus concerning the 7th Example of this invention. (A) shows the relationship between the sales force and the period, and (b) shows the relationship between the sales force and the function. It is a block diagram which shows the sales force estimation part in the information structuring apparatus concerning the 7th Example of this invention. It is a flowchart which shows operation | movement of the information structuring apparatus concerning the 7th Example of this invention. It is a figure which shows the information structuring apparatus concerning the 8th Example of this invention. It is a block diagram which shows the competitiveness increase component estimation part in the information structuring apparatus concerning the 8th Example of this invention. It is a flowchart which shows operation | movement of the information structuring apparatus concerning the 8th Example of this invention. It is a figure which shows a news relation figure. It is a figure which shows the computer concerning the Example of this invention.

Explanation of symbols

1, 1b, 21, 31, 41, 51, 61, 71, 81, 91 Information structuring device 2, 12, 22, 32, 43, 53, 63, 72, 82, 92 Quantifying device 100 Information source cutting unit 101 Information Source Storage Unit 102 Morphological Analysis Unit 103 Morphological Storage Unit 200 Information Source Reconstruction Unit 201 Format Analysis Unit 202 Format Rule Storage Unit 203 Element Word Storage Unit 204 Tree Generation Unit 205 Information Source Tree Storage Unit 206 Information Source Meta Information Storage Unit 300, 300b, 300c Information source selection unit 301 Information source selection unit 301b Tree selection unit 302 Domain dictionary storage unit 303 Applicable information source storage unit 303b Applicable information source tree storage unit 304 Non-conforming information source storage unit 304b Non-conforming information source tree storage Unit 305 Subject dictionary 400, 400b Group generation unit 401 Information source multiplication unit 401b Tree multiplication unit 40 2 Technical dictionary storage unit 500 Information source value estimation unit 501 Subject determination unit 502 Subject dictionary storage unit 503 Information source storage unit 504 Information source quantification unit 505 Word quantification map storage unit 506 Information reduction initial score storage unit 507 Measurement tree storage unit 600 Enterprise power estimation unit 601 Enterprise power calculation unit 602 Information source final score storage unit 700 Ripple force calculation unit 701 Parts delivery information source selection unit 702 Delivery destination company extraction unit 703 Feature word enumeration unit 704 Short-range term extraction unit 705 Ripple calculation unit 800 Demand cycle estimation unit 801 Product adoption information selection unit 802 Product delivery information selection unit 803 Duplicate detection unit 804 Monthly cumulative counter 805 Product demand display unit 900 Potential purchasing power estimation unit 901 Product adoption information selection unit 902 Product delivery information Sorting unit 903 Duplicate detection unit 904 Competitor product feature word sorting unit 905 In-house similar product selection 906 Financial information acquisition unit 908 Investment performance calculation unit 909 Investment similarity calculation unit 910 Customer group selection unit 911 Other company product non-adopted customer selection unit 1000 Sales force estimation unit 1001 Product feature word collection unit 1002 Average feature word calculation unit 1003 Product development information selection unit 1004 Product-specific feature word collection unit 1005 Product development information selection unit 1006 Average feature word number calculation unit 1007 Competitive product sales rate calculation unit 1100 Competitiveness enhancement component estimation unit 1101 Component adoption rate top companies Sorting unit 1102 Competitor product delivery results sorting unit 1103 Part contribution rate calculation unit 2100 Computer 2101 CPU
2102 ROM
2103 RAM
2104 Bus 2105 Input / output interface 2106 Input unit 2107 Output unit 2108 Storage unit 2109 Communication unit 2110 Drive 2111 Magnetic disk 2112 Optical disk 2113 Flexible disk 2114 Semiconductor memory

Claims (25)

An information source cutting unit that divides the information source to generate words;
An information source restructuring unit for classifying the word according to a predetermined format rule;
An information source selection unit that refers to a business domain dictionary in which business domains are registered, and selects the information source as a conforming information source related to a production good or a non-conformity information source not related to a production good;
Referring to a word dictionary in which a set of words and their synonyms is classified based on a superordinate concept and / or subordinate concept in the business domain, and extracting information sources having a high probability of producing goods from the non-conforming information sources, An information structuring apparatus having a group generation unit as a compatible information source. The information structuring apparatus according to claim 1, wherein the information source cutting unit includes a morpheme analysis unit that divides the information source into morphemes. The information structuring apparatus according to claim 1, wherein the information source restructuring unit forms the information source tree having the subject as a root by combining the words. The information source reconstruction unit
A format analysis unit for classifying the inseparable elements into product names, subjects, feature words, and actions;
The information structuring apparatus according to claim 3, further comprising: a tree generation unit configured to generate the information source tree from the information source divided into the inseparable elements based on the format analysis result. The information structuring apparatus according to claim 4, wherein the format analysis unit generates an element word by combining consecutive nouns in the morpheme. The said group production | generation part refers to the subject person dictionary by which the subject person was classified for every classification, and classifies the said conformity information source for every subject person. The Claim 1 thru | or 5 characterized by the above-mentioned. Information structuring device. The said information source selection part registers the word which does not exist in the said business domain dictionary among the words which comprise the said conforming information source in the said business domain dictionary. The information structuring apparatus described. The information according to any one of claims 1 to 6, wherein the group generation unit registers, in the word dictionary, a word that does not exist in the word dictionary among words constituting the matching information source. Structured device. The word dictionary is a set of distances between words based on the similarity between words,
With reference to the word dictionary, according to the distance between the word included in the information source for which the supplier company is the subject and the word included in the information source regarding the destination company, The information structuring apparatus according to claim 1, further comprising a ripple force calculation unit that calculates an influence force. The system further comprises a demand cycle estimation unit for estimating a demand cycle of the product based on the product and delivery source in the product adoption information source of the customer company and the product and delivery destination in the product delivery information source of the competitor company. Item 9. The information structuring apparatus according to any one of Items 1 to 8. 10. A potential purchasing power estimation unit for estimating the potential purchasing power of the customer company based on the financial information of the customer company and the product purchase history of the customer company or the product sales performance of the competitor company. The information structuring apparatus according to any one of the above. The information dictionary according to any one of claims 1 to 11, wherein the word dictionary is a technical dictionary in which a set of technical terms and synonyms thereof is classified based on a superordinate concept and / or a subordinate concept in the business domain. Structured device. An information source cutting step of dividing the information source to generate words;
An information source reconstruction step of classifying the word according to a predetermined format rule;
An information source selection step of selecting a business domain dictionary in which business domains are registered and selecting the information source as a conforming information source related to a production good or a non-conformity information source not related to a production good;
Referring to a technical dictionary in which a set of technical terms and synonyms thereof is classified based on a superordinate concept and / or subordinate concept in the business domain, and extracting information sources having a high probability of producing goods from the non-conforming information sources; And an information structuring method including a group generation step as the matching information source. The information structuring method according to claim 13, further comprising: a morpheme analysis unit that divides the information source into morphemes in the information source cutting step. The information structuring method according to claim 13 or 14, wherein, in the information source reconstruction step, the information source tree is formed by combining the words and having the subject as a root. The information source reconstruction step includes:
A format analysis step for classifying the inseparable elements into product names, subjects, feature words, and actions;
16. The information structuring method according to claim 15, further comprising: a tree generation step of generating the information source tree from the information source divided into the inseparable elements based on the format analysis result. The information structuring method according to claim 16, wherein in the format analysis step, element words are generated by combining consecutive nouns of the morphemes. The said information source selection process WHEREIN: The word which does not exist in the said business domain dictionary among the words which comprise the said conforming information source is registered into the said business domain dictionary. The any one of Claim 13 thru | or 17 characterized by the above-mentioned. Information structuring method described. 19. The group generation process according to claim 13, wherein in the group generation step, the subject refers to a subject dictionary classified by type, and classifies the conforming information source for each subject. Information structuring method. The information according to any one of claims 13 to 19, wherein in the group generation step, a word that does not exist in the technical dictionary among words constituting the matching information source is registered in the technical dictionary. Structured method. The technical dictionary has a distance between words set based on the similarity between words,
With reference to the technical dictionary, according to the distance between the word included in the information source for which the supplier company is the subject and the word included in the information source regarding the destination company, The information structuring method according to any one of claims 13 to 20, further comprising a ripple force calculation step of calculating influence. Further comprising a demand cycle estimation step of estimating a demand cycle of the product based on the product and delivery source in the product adoption information source of the customer company and the product and delivery destination in the product delivery information source of the competitor company. Item 22. The information structuring method according to any one of Items 13 to 21. 23. A potential purchasing power estimation step for estimating the potential purchasing power of the customer company based on the financial information of the customer company and the product purchase history of the customer company or the product sales performance of the competitor company, further comprising: The information structuring method according to any one of the above. 24. The information dictionary according to claim 13, wherein the word dictionary is a technical dictionary in which a set of technical terms and synonyms thereof is classified based on a higher concept and / or a lower concept in the business domain. Structured method. A program for causing a computer to execute processing for structuring an input information source,
An information source cutting step of dividing the information source to generate words;
An information source reconstruction step of classifying the word according to a predetermined format rule;
An information source selection step of selecting a business domain dictionary in which business domains are registered and selecting the information source as a conforming information source related to a production good or a non-conformity information source not related to a production good;
Referring to a technical dictionary in which a set of technical terms and synonyms thereof is classified based on a superordinate concept and / or subordinate concept in the business domain, and extracting information sources having a high probability of producing goods from the non-conforming information sources; A program including a group generation step as the matching information source.

Images