JP2018085020A - Information processing device, information processing system, control method therefor, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a mechanism that enables appropriate retrieval of a target from a plurality of categories by using information on the degree of certainty of the plurality of categories.MEANS FOR SOLVING THE PROBLEM: A information processing device receives an input character string that is a retrieval condition; specifies a plurality of categories corresponding to the input character string and the degree of certainty of correspondence between each of the categories and the input character string; selects a plurality of categories in which retrieval of a retrieval target is performed, on the basis of information on the degree of certainty of the specified plurality of categories; and performs control so as to perform retrieval processing using the selected plurality of categories as a constraint.SELECTED DRAWING: Figure 10

Description

  The present invention relates to an information processing apparatus, an information processing system, a control method thereof, and a program.

  2. Description of the Related Art Conventionally, with the digitization of text documents stored on the Internet or in organizations, the number of atypical documents has become enormous. For this reason, there is an increasing need to search information such as text documents stored in a storage device of a computer at high speed and with high accuracy.

  There are various techniques for retrieving data intended by a user from a large amount of text documents. There is also provided a technique for improving search suitability by narrowing down text documents to be searched by determining a category of text input as a search condition.

  In Patent Document 1, the problem of word ambiguity is solved by classifying query categories. For example, the word “DV” is an abbreviation meaning “domestic violence” in addition to “digital video”.

  Therefore, if the search is performed by dividing the query into words by morphological analysis without narrowing down the category of the text document to be searched, a document related to both meanings can be obtained as a search result. Patent Document 1 describes a technique for improving search accuracy by assigning a category to a document group to be searched, and by determining a query category and narrowing down the search target to eliminate word ambiguity. Yes.

JP 2006-227823 A

  In the technique described in Patent Document 1, since the query category is determined and the search target is narrowed down, when the accuracy of determining the category is high, the precision of the search result for the query can be improved.

  However, the accuracy of the query category determination is not always high. If a query is classified into an incorrect category, the correct category appropriate to the query will not be used for the search, and the accuracy of the search results for the query will be reduced. .

  Note that the high precision here is the relative value height among a plurality of categories corresponding to the input search key character string. Even if a search is performed only for the highest category, a desired search result may not be obtained.

  In particular, Embodiment 3 of Patent Document 1 describes a technique for transmitting the same query to a plurality of servers and displaying the results received from the respective servers together.

  However, for example, in a search for a document group managed separately by different servers, different scoring is statistically performed on each server depending on the appearance frequency of words on each server. Therefore, even if it is the same word, different scoring may be performed depending on the document group managed by the server, and even if the results of scoring on each server are merged, they are not necessarily scored properly. I can't say that.

  An object of the present invention is to provide a mechanism capable of easily specifying a plurality of categories as a category to be searched using information on the certainty level of a plurality of categories.

An information processing apparatus according to the present invention includes an input character string receiving unit that receives an input character string serving as a search condition, a plurality of categories corresponding to the input character string, and a certainty of correspondence between each category and the input character string. An information processing apparatus comprising: a specifying unit that specifies a selection unit that selects a plurality of categories for searching for a search target based on information on certainty factors of the plurality of categories specified by the specifying unit; Search control means for controlling to perform a search process using the category selected by the selection means as a constraint.
In addition, when the category selected by the selection unit does not satisfy a criterion that is a restriction condition of search processing, the selection unit adds another category identified by the identification unit until the criterion is satisfied. The search control means controls to perform a search process using the category selected by the selection means that satisfies the criteria as a restriction condition.

  According to the present invention, it is possible to provide a mechanism that can easily specify a plurality of categories as a category to be searched using information on the certainty level of a plurality of categories.

It is a figure which shows an example of the function structure which concerns on embodiment of this invention. It is a block diagram which shows an example of the hardware constitutions of the information processing apparatus which concerns on embodiment of this invention. It is a figure which shows an example of the text document used as the search object which concerns on embodiment of this invention. It is a figure which shows an example of a structure of the various data which concern on embodiment of this invention. It is a figure which shows an example of the technique of the category specification to which the input character string which concerns on embodiment of this invention belongs. It is a figure for demonstrating an example of the similarity between categories which concerns on embodiment of this invention. It is a figure for showing an example of a restriction category selection standard concerning an embodiment of the present invention. It is a figure for demonstrating an example of the similar category list | wrist which concerns on embodiment of this invention. It is a figure which shows an example of the flowchart of the outline | summary of the search process which concerns on embodiment of this invention. It is a figure which shows an example of the flowchart of a process of the category acquisition part which concerns on embodiment of this invention. It is a figure which shows an example of the flowchart of the category acquisition process based on the total certainty degree concerning embodiment of this invention. It is a figure which shows an example of the flowchart of the category acquisition process based on the similar category which concerns on embodiment of this invention. It is a figure which shows an example of the flowchart of the category acquisition process which gives priority to the category with the high certainty factor which concerns on embodiment of this invention. It is a figure for showing the image of the result of having recalculated the score of the text document of a search result based on the certainty factor of the category concerning an embodiment of the present invention. It is a figure which shows an example of the flowchart of the process which recalculates the score of the text document of a search result based on the certainty factor of the category which concerns on embodiment of this invention. It is a figure for showing the image of the result of having recalculated the score of the text document of a search result based on the similar category network which concerns on embodiment of this invention. It is a figure which shows an example of the flowchart of the process which recalculates the score of the text document of a search result based on the similar category network which concerns on embodiment of this invention. It is a figure which shows an example of the flowchart of the acquisition process of the category which the value of reliability which concerns on embodiment of this invention satisfy | fills designation | designated conditions.

<First Embodiment>
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. First, an example of a functional configuration according to an embodiment of the present invention will be described with reference to FIG.

  The input character string receiving unit 101 receives an input character string serving as a search condition for searching for a text document.

  When the information processing apparatus according to the present invention is a computer that functions as a client device, the input character string is a functional unit that directly receives an input from a user who uses the computer.

  Based on the information stored in the category specification reference storage unit 121, the category specifying unit 102 specifies the category of the input character string and its certainty factor in order to narrow down the category to be searched based on the input character string.

  The certainty factor is a value indicating how appropriate the inputted character string is classified into the target category, and is a score indicating the degree of matching between the character string and the category. The higher the certainty factor, the higher the possibility that the target character string belongs to the category.

  There are multiple types of categories as will be described later. In the present invention, the category specifying unit 102 does not select one category but expresses the possibility of all the specified categories as a certainty factor.

  However, for example, in the category A corresponding to the character string, the category A with the highest certainty is 30% certainty, and the second highest category B is 29%, only the first category A If the search is performed with the, the user may not be able to extract a desired search result that may be in the category B having almost the same certainty.

  The category specification criterion is specification criterion information of a category to which the input character string belongs, which is a determination criterion for which category the input character string belongs (in which category the search based on the input character string should be performed). The information on the category specifying standard is stored in, for example, an external memory of the information processing apparatus 100. For example, it is a rule that determines that a word having the same meaning as the word having the highest appearance frequency in the input character string is a category to which the input character string belongs (corresponds). For example, it corresponds to the profile set in Japanese Patent Laid-Open No. 2003-141129. Other known techniques may be used as the identification technique of the category.

  The details of specifying the category for the input character string and calculating the certainty factor for each category will be described with reference to FIG. 5. For example, the input character string (for example, for questions) and the category corresponding to the input character string are used for learning Then, machine learning is performed, and the category corresponding to the input character string newly accepted based on the learning result is specified and the certainty factor is calculated.

  It goes without saying that other methods and techniques for determining the category of the input character string, such as a method of determining by a rule or the like, a method of determining by performing semantic analysis, etc. are included. In this case, information used in rules and semantic analysis, or conditions coded as a program are also stored in the category specification reference storage unit 121.

  The category acquisition unit 103 acquires a similar category group based on the conditions described in the restriction category selection criteria stored in the restriction category selection criteria storage unit 122. That is, a category similar to the search target candidate category is specified.

  In addition, among some examples described later, one that acquires a category group based on the similarity between categories is also described. In that case, similar categories (including similar source and similar destination categories) stored in the similar category storage unit 123 are also referred to.

  Finally, the document search unit 104 searches the text document stored in the text document storage unit 124 based on the input character string and category group acquired in the above-described procedure.

  The search may be performed not on the text document itself but on an index generated in advance from the text document. Since the technique relating to the search and generation of the index of the text document is a well-known technique, the details are omitted.

  FIG. 2 is a block diagram illustrating an example of a hardware configuration of the information processing apparatus according to the embodiment of the present invention.

  As shown in FIG. 2, the information processing apparatus 100 and the application server 140 include a central processing unit (CPU) 201, a random access memory (RAM) 202, a read only memory (ROM) 203, and an input controller 205 via a system bus 204. The video controller 206, the memory controller 207, the communication I / F controller 208, etc. are connected. The CPU 201 comprehensively controls each device and controller connected to the system bus 204.

  Further, the ROM 203 or the external memory 211 will be described later, which is necessary for realizing the functions executed by each server or each PC, such as BIOS (Basic Input / Output System) and OS (Operating System) which are control programs of the CPU 201. Various programs are stored. Further, information necessary for carrying out the present invention is stored. The external memory may be a database.

  The RAM 202 functions as a main memory, work area, and the like for the CPU 201. The CPU 201 implements various operations by loading a program or the like necessary for executing the processing from the ROM 203 or the external memory 211 to the RAM 202 and executing the loaded program.

  The input controller 205 controls input from a keyboard (KB) 209 or a pointing device such as a mouse (not shown).

  The video controller 206 controls display on a display device such as the display 210. The display device may be a display device such as a liquid crystal display. These are used by the administrator as needed.

  The memory controller 207 is an external storage device (hard disk (HD)), flexible disk (FD), or PCMCIA (Personal Computer) that stores a boot program, various applications, font data, user files, editing files, various data, and the like. Controls access to an external memory 211 such as a Compact Flash (registered trademark) memory connected to a Memory Card International Association (Card Memory) card slot via an adapter.

  The communication I / F controller 208 connects and communicates with an external device via a network, and executes communication control processing on the network. For example, communication using TCP / IP (Transmission Control Protocol / Internet Protocol) is possible.

  Note that the CPU 201 can display on the display 210 by executing an outline font rasterization process on a display information area in the RAM 202, for example. Further, the CPU 201 enables a user instruction using a mouse cursor (not shown) on the display 210.

  Various programs to be described later for realizing the present invention are recorded in the external memory 211 and executed by the CPU 201 by being loaded into the RAM 202 as necessary.

  FIG. 3 is a diagram showing an example of a text document to be searched according to the embodiment of the present invention. Reference numerals 301a to 301e represent examples of individual text documents.

  Each text document is divided into several parts called fields. For example, the document shown in FIG. 3 includes three fields: a “question” (part Q) and a “answer” (part A) category.

  The field is not necessarily just the above three parts. For example, there are a part to be searched as a morpheme (or an N-gram token) analyzed by morpheme, a character string, a numerical value, a date, and the like. Either of these can be specified as a constraint condition.

  In this example, “question” (Q portion) and “answer” (A portion) are used as search targets such as morphemes (tokens). In addition, “category” is used as a constraint condition.

  That is, it is determined whether to extract each document shown in FIG. 3 as a search result using the category. This “category” is a category that is meant in the claims of the present invention. The above is the description of FIG.

  Next, an example of a list of categories according to the embodiment of the present invention will be described with reference to FIG. In the present embodiment, “question / answer” information of the software product is used as an example. Therefore, in the description of the present invention, “specifications”, “functions”, “error indications” to “users” generally appearing in “questions / answers” commonly used in software products as indicated by 400 “Information” etc. will be described as an example of a category. The above is the description of FIG.

  Next, with reference to FIG. 5, an example of a method for determining the category of an input character string according to the embodiment of the present invention will be described. In this figure, a method for generating a category specifying standard using machine learning will be described as an example.

  A technique for classifying text categories using machine learning is known, for example, in WATSON (registered trademark), an artificial intelligence of IBM (registered trademark) (https://www.ibm.com/watson/developercloud/ doc / nl-classifier /). For example, the categories are classified as shown in the image diagram of FIG.

  In the example of FIG. 5, first, the values of “question” (Q: 431) and “category” (432) given to each document are extracted from the text document (430). Reference numeral 430 denotes learning data 430 in FIG. 4, which is stored in the external memory of the information processing apparatus 100.

  The question does not necessarily have to be extracted from the text document. For example, as a response described in the text document, a sentence that the questioner may input may be created separately and stored in an external memory as another document.

  In the explanation example of the present invention, a text document in which a question and an answer are integrated is used.

  For example, a question sentence for learning may be manually created and stored in the external memory 211 as a document. In any case, it is assumed that “question” has already been prepared in the description of FIG.

  The information processing apparatus 100 inputs the “question” sentence and the “category” as a pair to the category specification reference learning program 501 and outputs a category specification reference for use in the processing of the present invention as a result.

  For this category identification (determination) / classification, for example, a technique described in JP-A-2003-141129 is used. Since this is a well-known technique, a detailed description is omitted. The category specification criterion is stored in the category specification criterion storage unit 121.

  When the input character string receiving unit 101 of the information processing apparatus 100 of the present invention receives an input character string serving as a search condition (502), the input character string is transferred to the category specifying unit 102, and the above-described category specifying unit 102 Based on the information stored in the category specifying reference storage unit 121, the category to which the input character string belongs (for example, corresponding to the character string “question”) is specified.

  The category certainty list 440 expresses as a list which category the input character string belongs to, and each element of the list includes a character string representing the identified category and a numerical value representing the certainty level. (The unit is% in the example).

  In the example shown in the figure, the input character string 502 (character string in the “question” field) has a certainty factor of 40% for a question regarding “installation”, a certainty factor of 30% for a question regarding “trouble”, and the like. It represents as. The above is the description of FIG.

  Next, an example of the restriction category selection criterion according to the embodiment of the present invention will be described with reference to FIG.

  The restriction category selection criterion 700 is a rule (rule) that defines a criterion for how the category acquisition unit 103 uses the category certainty list 440 of the category specifying unit 102.

  That is, according to the restriction category selection criterion 700, a list of categories specified for the input character string and its certainty factor (category certainty factor list 440) is used. The constraint category selection criterion 700 is stored in the constraint category selection criterion storage unit 122.

  The restriction category selection criterion 700 is composed of 701 to 704. A name 701 is a name of a restriction category selection criterion and is identification information. The category selection condition 702 indicates which category is narrowed down (selected) as a target category for document search.

  In the description of the present invention, a category that is selected (narrowed down) using the constraint category selection criterion 700 as a search constraint condition is referred to as a constraint category.

  The minimum standard 703 is the minimum standard to be satisfied by the category narrowed down by the category selection condition 702. The additional category selection condition 704 is a process that is additionally executed when the minimum standard 703 is not satisfied, and indicates which category is added to the narrowed-down category.

  That is, when the condition of 703 is narrowed down and the condition of 703 is not satisfied, the process of 704 is repeated until the condition of 703 is satisfied, and the restriction category is narrowed down.

  FIG. 7 shows an example of six constraint category selection criteria (700a to 700g). In the first embodiment of the present invention, it is assumed that the constraint category selection criterion 700a is designated as the selection criterion to be executed first.

  Note that it may be possible to arbitrarily specify which constraint category selection criterion is used to select a category corresponding to the input character string. For example, all the constraint category selection criteria shown in 700a to 700f are stored in the constraint category selection criteria storage unit 122, and any constraint category is selected by a setting file (not shown) or user selection (not shown user interface). It is assumed that selection criteria 700 is first used or specified.

  Further, the name 701 of another restriction category selection criterion can be described in the additional category selection condition 704 and used. In that case, the restriction category selection criteria 700 also needs to be stored. Details will be described later.

  Next, with reference to FIG. 9, the search process according to the embodiment of the present invention will be described. Each step of the flowchart of FIG. 9 is executed by the CPU 201 on the information processing apparatus 100.

  In step S <b> 901, the input character string receiving unit 101 receives an input character string as a condition for searching for a text document by the document search unit 104 described later. For example, on a display screen (not shown), an input of a character string “A message appears during installation and installation does not operate normally” by a user operation is accepted.

  In step S902, it is determined and specified to which category the input character string received in step S901 is classified.

  Specifically, the category identification unit 102 uses a category identification criterion stored in the category identification criterion storage unit 121 to classify a plurality of pairs of categories for classifying an input character string and the certainty of classification for the category, Obtained as a specific result and stored in the category certainty list 440 on the RAM 202.

  Also, for example, at the timing of step S902, the similarity of each category specified as one of the category certainty lists 440 is specified and stored in the RAM 202 as a similar category list. 800a to 800c (generically referred to as the similar category list 800) in FIG. 8 are examples of the similar category list.

  In the description of the embodiment of the present invention, 800c will be described as an example, but the similar category list may be generated and stored in a configuration such as 800a or 800b.

  The similar category list 800a indicates which category (similarity source category 801) is similar to which category (similarity destination category 802). The numerical value in parentheses in the column of the similar destination category 802 is the similarity between the category described immediately before the parenthesis and the similar source category 801.

  The similar category list 800b shows an example of another storage form of the similar category list 800. In the similar category list 800b, there is no distinction between the similar source category and the similar destination category.

  The similar category list 800c indicates which category (similarity source category 801) is similar to which category (similarity destination category or constraint release 804). If a restriction release value is stored in the similar destination category or restriction release 804 column, the restriction based on the category stored in the restriction category list 410 (FIG. 4) is released and a search is performed. Details will be described later.

  The restriction category list is a list of categories used as restriction conditions, and is stored on the RAM of the information processing apparatus 100. In a later step S904, the information processing apparatus 100 performs a search using the input character string for the category documents stored in the restriction category list 410.

  When the similar category is specified / acquired using 800a or 800c, the target category (for example, the category having the first certainty in the category certainty list 440 of the category specifying unit 102) is selected as the similar source category 801. By searching in the column, the category group described in 802 or 804 corresponding to 801 is specified as a category similar to the target category. For example, if the category with the highest certainty factor is “trouble”, categories such as “install” and “uninstall” are included in the similar category group from the first line.

  When a similar category is specified / acquired using 800b, another category group stored in the same data row 803 as the target category is specified as a similar category. For example, a search is made as to which of the similar categories 803 includes “trouble”, and categories included in a row including “travel” are regarded as similar categories.

  A method for identifying similar categories will be described with reference to FIG. FIG. 6 is a diagram for explaining an example of similarity between categories according to the embodiment of the present invention.

  In FIG. 6, the character string in the circle indicates the category name, and the lines (solid line and dotted line) indicate the similarity between the categories. For convenience, it will be called “similar category network” between categories.

  For example, focusing on “Trouble” as a category, the categories “Operating environment”, “Function”, “Uninstall”, “Install”, “Error display” and “Character string input as question” combined with solid lines The trend is similar.

  That is, the similarity in the “similar network” does not depend on whether the category name itself, the content of the answer, or the technical substance represented by the answer is similar.

  In the sense, “a character string input as a question” is learned by the category specifying reference learning plum 501 in FIG. 5 and determined by the category specifying unit 102, so that it tends to be determined erroneously. It is similar.

  The similar network may designate “similarity” between similar categories. Moreover, although the similarity is described only between the categories adjacent on the similar network (directly connected by a line), for example, the similarity between categories that are not adjacent may be defined.

  Alternatively, for example, when connected via another category of “function” such as “trouble” and “specification”, the similarity may be calculated based on two similarities. .

  For example, since “trouble” and “function” are “0.8” and “function” and “specification” are “0.9”, “0.8 × 0.9 = 0.72” is “trouble”. It is a similarity of “specification”. Of course, the calculation method is not limited to this example.

  Furthermore, not all categories need be connected on a similar network. In the example of FIG. 6, the two categories of “user information” and “license information” are not connected to other categories. That is, there is almost no probability that the question text for searching text documents of two categories is similar to the question text for other categories.

  Of course, it is actually not zero and may be a very low probability. It goes without saying that whether or not to divide the network is merely for convenience and is merely an example.

  The calculation of the similarity between the categories can be replaced by the conventional technique. The above is the description of FIG.

  Returning to the description of FIG. In step S903, the category acquisition unit 103 reads one constraint category selection criterion from the constraint category selection criterion storage unit 122, and performs a constraint category selection process that is a category for searching for a document. Details of the processing of S903 will be described later with reference to FIG.

  It should be noted that at this point in time, which constraint category selection criterion in FIG. 7 is to be read is determined in advance by setting or user selection.

  In the first embodiment, the restriction category is selected (narrowed down) using 700a in FIG. Details will be described later with reference to FIG.

  Here, with reference to FIG. 10, the process of the category acquisition part which concerns on embodiment of this invention is demonstrated. Each step of the flowchart of FIG. 10 is executed by the CPU 201 on the information processing apparatus 100.

  In step S <b> 1001, the category having the first certainty factor is acquired from the category certainty factor list 440 obtained by the category specifying unit 102. Further, it is determined whether to release the constraint condition using the category.

  Specifically, it is determined with reference to the similar category list 912 (specifically, 800c in FIG. 8) whether to release the constraint condition. When “restriction release” is described in the similar destination category or restriction release 804 column, the restriction condition is released. When the name of a similar destination category similar to the similar source category 801 is described, the restriction condition is not released.

  The release of the constraint condition is useful, for example, when the restriction by category can improve the accuracy of the search.

  For example, assume that the category with the highest certainty is “trouble”. In addition, for example, “trouble” is related (similar) to all items (installation, function, version upgrade), and if a constraint condition is given to the search target based on the category “trouble”, an erroneous constraint occurs on the contrary. Suppose you know that it will degrade the accuracy of your search.

  In such a case, as “restriction release”, it is more likely that the search accuracy is higher when no restriction is applied to the subsequent search.

  In the present embodiment, it is assumed that the CPU 201 of the information processing apparatus 100 automatically sets “restriction release” to 804 when the number of similar destination categories exceeds a predetermined number. The predetermined number of values can be arbitrarily changed by a user operation on a setting screen (not shown), and is stored in advance in the external memory of the information processing apparatus 100.

  For example, when the category having the highest certainty factor is “trouble”, it is determined whether or not the category “trouble” is designated “unrestricted” in the similar category list. If it is determined in step S1001 that “there is no restriction release instruction” (“no restriction release” is described in 804), the process proceeds to step S1002. If it is determined that “there is an instruction to release the constraint”, the process proceeds to step S1008.

  In step S1008, all the category group information stored in the restricted category list 410 is deleted (reset). Thereafter, the process of the flowchart of FIG. 10 is terminated, and the process proceeds to step S904 of FIG. That is, by resetting the category group already registered in the restriction category list 410, the document is searched in step S904 without restriction by category.

  On the other hand, in step S1002, the storage variable used for summing up the certainty of the categories thereafter is reset to zero.

  In step 1003, the specified restriction category selection criterion 700 is read from the restriction category selection criterion storage unit 122 (in the first embodiment, 700a is read), and the category selection condition 702 is acquired.

  In step S1004, according to the category selection condition 702, the process which acquires the category group used as a constraint condition is performed. The acquired category is added to the restriction category list 410 and used in the document search process described later.

  The restriction category list is a list of categories used as restriction conditions, and is stored on the RAM of the information processing apparatus 100. That is, the information processing apparatus 100 performs a search using the input character string for the category documents stored in the restriction category list in the subsequent step S904.

  Details of the processing in step S1004 using 700a will be described later with reference to FIG.

  Here, with reference to FIG. 11, a restriction category selection process in the first embodiment of the present invention will be described. Each step of the flowchart of FIG. 11 is executed by the CPU 201 on the information processing apparatus 100.

  According to 702 of 700a, steps S1101 to S1103 are repeatedly executed until the total value of the certainty factors of the categories set as the constraint conditions becomes 90% or more.

  Specifically, the categories specified by the category specifying unit 102 are acquired in descending order of confidence, and the process is applied until the certainty total condition (90% in the example of 702b in FIG. 7) is reached. Is repeated.

  In step S1101, attention is focused on one data among unprocessed categories in the category certainty list 440 (for example, a corresponding data string is read on the RAM). Then, the certainty factor of the focused category is acquired from the category certainty factor list 440 (specific result).

  In step S1102, the certainty factor acquired in step S1101 is added to and stored in a total value storage unit (total value storage table 420) that stores the total confidence factor.

  In step S1103, the category under consideration is stored in the restriction category list 410 in order to be used as restriction information for the category in the subsequent text document search.

  The CPU 201 of the information processing apparatus 100 ends the processing of steps S1001 to S1003 when the total certainty factor of the total value storage table 420 is 90% or more, and lists the list of categories stored in the restricted category list. Is determined and output and stored on the RAM 202, and the processing of FIG.

  Returning to the description of FIG. In step S1005, it is determined whether the currently selected category selection condition is the (first) condition specified by the category selection condition 702 or the additional condition specified by the additional category selection condition 704.

  If the condition is specified by the category selection condition 702, the process advances to step S1006. If it is an additional process, there is no information to be processed any more, so the workflow process of FIG. 10 is terminated.

  In step S1006, it is determined whether or not a minimum condition (minimum criterion 703) for repeating the category selection procedure is satisfied.

  For example, it is determined whether the minimum criterion 703 described in the currently selected constraint category selection criterion 700 is satisfied.

  When the minimum standard 703 is satisfied (in the case of YES), the process of this flowchart is terminated. That is, a constraint condition based on a category at the time of document search is determined, and a constraint category list as the determined constraint condition is output on the RAM 202 and stored. If the minimum standard 703 is not satisfied (in the case of NO), the process proceeds to step S1007.

  In step S1007, when the minimum standard 703 is not satisfied, the next instruction (processing) is read out from the additional category selection condition 704 and specified, and the specified condition is “reset constraint” or another constraint. It is determined whether it is the name of the category selection criterion 700 or other conditions.

  When there is an instruction to “reset constraints” in the additional category selection condition 704, the category group stored so far is reset in the constraint list 913, and the processing of the flowchart in FIG.

  When the name of another restriction category selection criterion 700 is described in the additional category selection condition 704, it is added to the category already stored in the restriction category list and described in the additional category selection condition 704. In order to store the category selected by the name constraint category selection criterion 700 in the constraint category list, the process returns to step S1003, the constraint category selection criterion 700 specified in 704 is acquired, and the processing from step S1004 is applied. To do.

  If the condition itself is described in the additional category selection condition 704, the condition is added to the category already stored in the restriction category list, and the category selected using the condition is stored in the restriction category list. In step S1004, category selection using is performed, and the process proceeds to step S1005.

  In addition, in 700a of FIG. 7, since the minimum standard 703 is not described, it is determined that there is no condition, and the process of FIG. 10 is terminated assuming that the condition is satisfied. The above is the description of FIG.

  Returning to the description of FIG. In step S904, the document search unit 104 acquires a list of text documents that meet the conditions from the text document storage unit 124 using the above-described input character string as a search condition.

  At this time, if data exists in the constraint category list 410 (if it is not an empty list), the category is used as a constraint condition. That is, a search using the input character string is performed for documents in the categories stored in the restricted category list, and a list of documents hit by the search is output and stored on the RAM 202 as a search result.

  Further, the document list of the search result is displayed on a display screen (not shown). Above, description of the flow of processing by the flowchart of FIG. 9 is completed.

  The first embodiment of the present invention has been described above. According to the first embodiment of the present invention, it is possible to provide a mechanism that can easily specify a plurality of categories as a category to be searched using information on the certainty level of a plurality of categories.

  Therefore, narrowing down the search target with more emphasis on certainty compared to narrowing down a predetermined number of categories as the search target category or simply adding a category similar to a certain category as the search target Is possible.

  For example, since the narrowing down in the document search unit 104 is in a category in which the question sentences are classified with a high probability, an effect of executing a highly accurate search can be obtained.

  In the first embodiment, in the constraint category selection criterion of 700a, the category selection condition 702 is “from the highest confidence level until the confidence level reaches 90%”, and the minimum standard 703 and the additional category selection condition 704 are Although not set, for example, the category selection condition 702 is “the category with the highest certainty”, the minimum standard 703 is “the total certainty of the category selected by the selection condition is 90% or more”, and added By executing the processing of FIG. 10 with the category selection condition 704 of “added until a total of 90% from a high certainty factor”, the same invention and its effect can be enjoyed.

  Further, as shown in 700g, by setting the above-mentioned minimum standard 703 and additional category selection condition 704 and executing the processing of FIG. 10, it is possible to perform further category specifying processing as a search target. It is.

  For example, if there is a category that is not included in similar categories among the previously selected categories, the missing similar categories are added based on the similar category storage unit 123 of FIG.

  As a result, it is possible to obtain an effect of improving accuracy by preventing a search omission for a text document of a category having a high similarity of question sentences.

  In the above-described embodiment, the instruction regarding “restriction release” is described as an example described in the similar category list 912 (illustrated in 800c of FIG. 8). Needless to say, the storage unit may be provided to check the information.

<Second Embodiment>
In the second embodiment, in the process of FIG. 10, the restriction category is selected (narrowed down) using 700b of FIG. Details will be described later with reference to FIG.

  Here, with reference to FIG. 12, the restriction category selection processing in the second embodiment of the present invention will be described. Each step of the flowchart of FIG. 12 is executed by the CPU 201 on the information processing apparatus 100. Note that description of processing that is common to the processing described in the above embodiment is omitted.

  In step S1004, the information processing apparatus 100 executes the process of FIG. 12 instead of the process of FIG. In step S1201, the information processing apparatus 100 selects a category having the highest certainty among the categories included in the category certainty list 440 (FIG. 4) that is the category identified by the category identifying unit 102 (category certainty). Pay attention to the category having the first degree of certainty in the degree list 911). Then, the category having the highest certainty factor is stored in the restriction category list 410.

  The restriction category list 410 is a category for searching for a document according to an input character string in step S904 of FIG. 9 described later.

  In step S1202, a category having the highest certainty factor and a similar category group are acquired. Specifically, the group to which the first category belongs is acquired from 800 in FIG. 8 and read onto the RAM. The information in FIG. 8 is already stored in the external memory of the information processing apparatus 100.

  Hereinafter, by repeating the processing of step S1203 and subsequent steps for all categories in the similar category list acquired in step S1202, the category having the highest certainty factor and the total certainty factor of the category similar to that category is obtained. Calculate the value.

  In step S1203, one unprocessed category in the similar category list acquired in step S1202 is acquired.

  In step S1204, the total value storage unit (total value storage table 420) that stores the total certainty factor of the category with the highest certainty factor and the certainty factor of the category once acquired in step S1203 is acquired in step S1204. The certainty value of the selected category is added and stored.

  In step S1205, the category acquired in step S1203 is stored in the restriction category list 410 in order to be used as a restriction condition for the category in the subsequent text document search. This is the end of the description of the flowchart in FIG.

  Note that similar categories may be specified by using a data structure in which similar categories are in a network form as shown in FIG. 6. In this case, all categories with links may be selected or adjacent to each other. You can choose only one, or add to the category group until it falls below a certain number, while multiplying the value as you follow many times according to the number of similarities described on the line (solid line, dotted line), etc. Also good. This is a design matter, and any method may be used for determining the category group.

  The CPU 201 of the information processing apparatus 100 determines whether or not the minimum standard 703b of 700b is satisfied in step S1005 of FIG. 10, and if not, the process proceeds to step S1007.

  That is, if the total certainty factor of the acquired constraint category is not 90% or more, the process proceeds to step S1007, the process returns to step S1004 with reference to 704b, and the category with the first certainty factor is similar to that category. 11, the process described in FIG. 11 is repeatedly executed until the total of certainty becomes 90% or more, together with the certainty of the category already stored in the restriction category list, for categories other than the category that is being used. . The above is the description of FIG.

  The second embodiment has been described above. According to the second embodiment, it is possible to provide a mechanism capable of easily specifying a plurality of categories as search target categories, taking into account similar categories and certainty factors.

  According to the second embodiment, the search by the document search unit 104 is performed by narrowing down to all categories that are assumed to have similar question sentences, thereby obtaining an effect of realizing a highly compatible search. Can do.

  In addition, if the minimum criteria are not met, if the certainty is low even if all the similar categories of the question are collected, an additional decision is made based on the idea that the accuracy of similarity determination is low in the first place. A category can be further added to the constraint condition according to the category selection condition.

  In the second embodiment, categories that have a high certainty factor are added to the restricted category list (those that are not yet included in the category group). Therefore, even if the accuracy of similarity determination is poor, it is possible to perform a search without fail and to obtain an effect of realizing a search with high reproducibility.

<Third Embodiment>
In the third embodiment, in the process of FIG. 10, the restriction category is selected (narrowed down) using 700c of FIG. Details will be described later with reference to FIG. Note that description of processing that is common to the processing described in the above embodiment is omitted.

  FIG. 13 is a diagram illustrating an example of a flowchart of the category acquisition process in which priority is given to the category with higher confidence according to the embodiment of the present invention. Each step of the flowchart of FIG. 13 is executed by the CPU 201 on the information processing apparatus 100.

  In step S1004, the information processing apparatus 100 executes the process in FIG. 13 instead of the process in FIG. 11 or FIG.

  Steps S <b> 1301 to S <b> 1303 are repetitive processes for acquiring a specified number of categories in a descending order from a category having the highest certainty factor.

  In step S1302, in accordance with the category selection condition 702c of the restriction category selection criterion 700c having “higher category” in the name 701b, the unprocessed, highest-ranked category from the categories up to the designated rank in the category certainty list 440 is displayed. Focus on one of the categories.

  In step S1303, the storage unit (total value storage table 420) that adds up and stores the certainty factors of the focused category is added to the stored value and stored.

  In step S1304, the category of interest is stored in the constraint category list 410 in order to be used as constraint information for the category in subsequent text document searches. This is the end of the description of the flowchart in FIG.

  When the accuracy of the category specifying unit 102 is very high and the number of categories is large, it is possible to reduce the load of narrowing down in the document search unit 104 by fixing the number of narrowed categories to a small number.

  As for the processing after step S1005, as in the invention of the second embodiment, if the certainty level sum of the previously selected category does not satisfy a certain standard, the total certainty level is further increased to 90%. Add a category.

  If the category identification accuracy of the category identification unit 102 initially assumed is high, the definition can be used with peace of mind by being able to handle even when the category identification unit 102 is low against expectations.

  Heretofore, the third embodiment of the present invention has been described. According to the third embodiment of the present invention, it is possible to provide a mechanism that can easily specify a plurality of categories as a search target category by taking into account the designation of the number of categories and the certainty of the designated category. .

<Fourth Embodiment>
In the fourth embodiment, in the process of FIG. 10, the restriction category is selected (narrowed down) using 700d of FIG. Details will be described later with reference to FIG. Note that description of processing that is common to the processing described in the above embodiment is omitted.

  FIG. 18 is a diagram illustrating an example of a flowchart of a category acquisition process in which the certainty value according to the embodiment of the present invention satisfies a specified condition.

  In step S1004, the information processing apparatus 100 executes the process of FIG. 18 instead of the processes of FIG. 11, FIG. 12, and FIG. Each step of the flowchart in FIG. 18 is executed by the CPU 201 on the information processing apparatus 100.

  Note that “the certainty value satisfies the specified condition” means that “the certainty factor is not 0” or “the certainty factor is 5%, as described in the constraint category selection criterion 700d in FIG. Those that satisfy any condition, such as “something above”. According to 700d, unless the certainty factor is 0 in the category certainty factor list 911, the category is acquired as a constraint condition.

  Steps S1801 to S1804 are the categories included in the category certainty factor list 440, which is the identification result of the category identifying unit 102, whose certainty satisfies the specified condition (in the example of 702d in FIG. In which the degree is not 0%) is added to the restriction category list.

  In step S1801, the certainty factor of the focused category is acquired from the category certainty factor list 911.

  In step S1802, it is determined whether or not the certainty of the focused category satisfies a specified condition.

  In this example, it is determined whether the certainty level exceeds 0% or 0%. If the specified condition is not satisfied (in this example, 0%: NO), the processing for the category of interest is completed and the next category is focused. When the specified condition is satisfied (in this example, when the certainty level exceeds 0%: YES), the process proceeds to step S1803.

  In step S1803, the certainty factor acquired in step S1802 is added to and stored in the storage unit (total value storage table 420) that stores the total certainty factor.

  In step S1804, the category under consideration is stored in the constraint category list 410 to be used as constraint information for the category in subsequent text document searches. This is the end of the description of the flowchart in FIG.

  Although not described in detail, the category selection condition 702 can describe various conditions. For example, although the condition to be acquired has been described in the above description, it may be described that the condition for not acquiring “a category with a certainty factor of less than 2% is not acquired”.

  Of course, this is the same condition as “acquire all categories with a certainty factor of 2% or more”. Further, the network of the similar category in FIG. 6 is traced twice, or the similarity on the line connecting the categories is sequentially multiplied from the first category as the similarity of the distant category, and the similarity is 0.6. It is also possible to follow the network until the following is reached and acquire all categories within that range. All of these are included as examples of the present invention.

  In addition, the definitions that can be described in the minimum standard 703 and the additional category selection condition 704 are similarly included in the scope of the embodiments of the present invention as long as they are implemented as a program or implemented in a format that can be interpreted by the program. The fourth embodiment has been described above.

<Fifth Embodiment>
In the fifth embodiment, in the process of FIG. 10, the restriction category is selected (narrowed down) using 700e of FIG. Note that description of processing that is common to the processing described in the above embodiment is omitted.

  The constraint category selection criterion 700e in FIG. 7 is the same as the category selection condition 702 and the minimum criterion 703 is the same as the constraint category selection criterion 700c. However, while the additional category selection condition 704c specifies the condition, in the case of the additional category selection condition 704d, the names of other constraint category selection criteria 700 are described.

  In other words, when the minimum criterion 703e is not satisfied, another constraint category selection criterion 700b (designation of certainty factor) that has already been defined is called to select a constraint category.

  Note that, for example, if the specified certainty level that has been called does not satisfy the minimum criterion 703b and the name of another constraint category selection criterion 700 is described in the additional category selection condition, the additional selection process of the constraint category is repeated indefinitely. For example, the number of processings can be limited by storing in advance the number of repetitions of the repeated processing based on the additional category selection condition 704.

  Alternatively, when it is repeated a predetermined number of times or more, it is determined that it is inappropriate to narrow down the category and execute the document search unit 104, and it is possible to determine that the narrowing down by category is not performed.

<Sixth Embodiment>
In the sixth embodiment, in the process of FIG. 10, the restriction category is selected (narrowed down) using 700f of FIG. Note that description of processing that is common to the processing described in the above embodiment is omitted.

  The constraint category selection criterion 700f is the same as the category selection condition 702 and the minimum criterion 703 is the same as the constraint category selection criterion 700b. However, the additional category selection conditions 704b and 704f are different.

  “Reset constraint” described in 704f resets the constraint condition for the category at the time of document search when the minimum standard 703f is not satisfied. That is, it is specified that, instead of executing another process for selecting a category, a search is performed on text documents of all categories at the time of document search.

  Therefore, when the minimum standard 703f is not satisfied, for example, the constraint category included in the constraint category list is further increased, and the text document storage is performed without the constraint, rather than using the category group as a constraint condition at the time of retrieval. The rule is to release the constraint condition when it is determined that all text documents stored in the unit 124 should be searched. The sixth embodiment of the present invention has been described above.

  As described above, according to the present invention, it is possible to provide a mechanism capable of easily specifying a plurality of categories as a category to be searched using information on the certainty level of a plurality of categories.

<Seventh Embodiment>
In addition, the priority of search result extraction and output is determined by the score (score of each document indicating how valid the document is as a search result) specified from the query text that is the input character string and the document itself. Can do.

  However, the high certainty level and rank of the category to which the question sentence belongs does not necessarily match the high score and rank specified by the direct relation between the question sentence itself and the document.

  In order to output the search result more accurately, it is necessary to perform more flexible scoring (ranking) using the certainty factor information and present the search result to the user.

  Flexible selection of categories determined from queries improves reproducibility in search accuracy (the percentage of correct answers included in search results), and what is the question text even if it contains incorrect answers? It is desired to provide a search and scoring technique for making a plausible answer higher by adjusting the scoring in consideration of the analysis result.

  In the seventh embodiment, by weighting the search score based on the certainty factor for the category of the question sentence, not only the search score but also the certainty of the specific result indicating which category the question sentence belongs to is taken into account. Re-scoring.

  Hereinafter, the seventh embodiment of the present invention will be described with reference to FIGS. In addition, description is abbreviate | omitted about the process same as the process described in embodiment mentioned above.

  FIG. 14 is a diagram illustrating an image of a result of recalculating a score of a text document as a search result based on the certainty of a category according to the embodiment of the present invention. The input character string 502 (question sentence) and the category certainty list 440 are the same as those in FIG.

  The search result 1401 and the search result 1402 are data of a document search result, and are information in which each document of the search result is associated with a category to which the document belongs and a score for the question sentence of the document.

  A search result 1401 is a result of a search performed by the document search unit 104, and is a state in which weighting adjustment based on the certainty of category described later in the description of FIG. 15 is not performed. The first place in the search result is the category “trouble” and the score “80”. The second place is a category 1 “install” with a score of “75”.

  In the seventh embodiment, according to the category presented in the category certainty factor list 440 and its weighting, for example, the score of 1401 is adjusted by the formula shown in “calculation example” 1403.

  A search result 1402 is a search result obtained by adjusting weighting according to the category.

  Here, the reliability of the category to which the text document belongs is acquired from the category reliability list 440 and multiplied by the search score of 1401.

  In the example, “trouble” (score 80), which is first in 1401, is 24 according to the certainty (30%). Two cases of “Installation” (scores 75 and 70) that are also ranked second in 1401 become 30 and 28 depending on the certainty factor (40%), and the scores after adjusting with the certainty factor of the category are switched. become. A state in which the order has been changed is shown in 1402.

  In other words, by weighting the search score based on the certainty for the question category, not only the search score, but also the certainty of the specific result that the question sentence belongs to affects the score. It is possible to obtain an effect of performing more accurate scoring (ranking) on the result and presenting it to the user.

  With reference to FIG. 15, the process of recalculating the score of a text document as a search result based on the certainty of the category according to the embodiment of the present invention will be described. That is, details of the contents described in FIG. 14 will be described. Each step of the flowchart of FIG. 15 is executed by the CPU 201 on the information processing apparatus 100 instead of step S904.

  In step S1501, the text document storage unit 124 is searched to acquire a text document group that satisfies the search condition.

  At the time of this search, if there is a description of one or more categories in the restriction category list 410 described above, the search is limited to text documents including those categories.

  Further, for example, scoring is performed on each document using a conventional technique disclosed in Japanese Patent Application Laid-Open No. 2008-225957.

  Thereafter, the processes in steps S1502 to S1506 are repeatedly executed on the text document group of the search result (for example, 1401 in FIG. 14) acquired in step S1501.

  In step S1502, attention is focused on one of the text documents to be processed from the acquired text document group.

  In step S1503, the category information specified in the category field and the search result score assigned in step S1501 are acquired from the text document under consideration. For example, “category” and “score” 1401 in FIG. 14 are acquired.

  For example, paying attention to the document in the first line 1401, the trouble that is the category and the value of 80 that is the score of the document are acquired.

  In step S1504, the certainty factor of the category information acquired from the text document under attention is acquired from the category certainty factor list 911. For example, the acquired category = 30% of the value of the certainty of trouble is obtained from the category certainty list 440.

  In step S1505, the score of the text document under attention is recalculated based on the score of the text document under attention (obtained in S1504) and the certainty of the category (obtained in S1505).

  In the seventh embodiment, the score is recalculated by multiplying the score of the acquired document by the certainty of the category to which the document belongs.

  That is, re-scoring is performed in consideration of not only the search score, but also the certainty of the specific result that the question sentence belongs to.

  In S1506, the score calculated in S1506 is reset as the score of the text document under consideration.

  The CPU 201 of the information processing apparatus 100 applies the processing in steps S1502 to S1506 to all search result documents, and then applies the processing to the re-scored search result document list (for example, FIG. 14). The search results 1402) are sorted in descending order of score, and are determined (determined) as search result information, and stored / updated in the RAM 202. That is, search result output processing is performed. This is the end of the description of the workflow processing in FIG.

  According to the seventh embodiment, more accurate search results can be specified using the certainty factor information.

  As described above, according to the present invention, it is possible to provide a mechanism that can easily specify a plurality of categories as a category to be searched using information on the certainty level of a plurality of categories.

<Eighth Embodiment>
Also, there are cases where it is desired to specify a more accurate search result by taking into account not only the certainty of the category but also the relationship between the categories.

  In the eighth embodiment, a more accurate search result is specified in consideration of the certainty of the category and the relationship between the categories.

  Specifically, in the eighth embodiment, the score of a document is recalculated using a category with a high certainty factor and the similarity between the category and the category of the document.

  Hereinafter, an eighth embodiment of the present invention will be described with reference to FIGS. In addition, description is abbreviate | omitted about the process same as the process described in embodiment mentioned above.

  FIG. 16 is a diagram illustrating an image of a result of recalculating a score of a text document as a search result based on the similar category network according to the embodiment of the present invention.

  In the description of FIG. 14 described above, the score of the text document is recalculated based on the certainty factor corresponding to the category to which each text document belongs. On the other hand, in FIG. 16, the document is re-scored using the information of the similar category network as shown in 1601.

  Note that reference numeral 1601 in FIG. 16 is an excerpt of the similar category network described in FIG. The values “0.8” and “0.9” described in 1601 indicate the similarity between the categories.

  First, as in FIG. 14, it is assumed that the category “installation” is ranked first in the category certainty factor list 440.

  In the eighth embodiment, re-scoring is performed using the similarity between the category of the document acquired from the search result 1401 and the category having the highest certainty factor in the category certainty factor list 440.

  For example, the similarity from the category “installation” is “0.8” in the case of “trouble”.

  Regarding “error display”, there is no direct link (similar relationship) to the category with the highest confidence level, but it is connected via “trouble” (similar relationship is specified). Multiply by the similarity of the two connections. Therefore, as indicated by 1601 in FIG. 16, the similarity between “install” and “error display” is “0.8 × 0.9 = 0.72”.

  In the eighth embodiment, when the score of a document in the category of “error display” in the search result 1401 is multiplied by this, a search result (score after category weighting) 1602 is obtained.

  That is, the score is readjusted according to the category of the text document.

  It is assumed that the “installation” document having the first certainty factor is multiplied by similarity = 100% (that is, a value of “1”). This completes the description of FIG.

  With reference to FIG. 17, a process of recalculating the score of a text document as a search result based on the similar category network according to the embodiment of the present invention will be described. That is, details of the contents described in FIG. 16 will be described. Each step of the flowchart of FIG. 17 is executed by the CPU 201 on the information processing apparatus 100 instead of step S904.

  The difference from the processing of the seventh embodiment described in FIG. 15 is that the score used for recalculating the score of each text document is not the category certainty itself but the similarity between categories.

  The same processes as those in FIG. 15 are given the same step numbers as in FIG. Hereinafter, processing different from FIG. 15 will be described.

  In step S1704, the CPU 201 of the information processing apparatus 100 acquires the category having the highest certainty factor from the category certainty factor list 440. Further, the category of the text document under attention acquired in step S1504 is acquired, and the similarity in the similar category network of FIG. 6 is acquired.

  For example, as shown in 1602 of FIG. 16, when the category of the focused document is “trouble”, a value of similarity 0.8 (similarity = 80%) with “install” is obtained from the data of the similar category network. get. When the category of the focused document is “error display”, a similarity of 0.72 (similarity = 70%) with “installation” is acquired.

  In step S1705, based on the similarity acquired from the similar category network in S1704 and the score of the focused text document acquired in S1504, the score of the focused text document is recalculated.

  In the eighth embodiment, the score of the acquired document is recalculated by multiplying the score of the acquired document by the similarity value between the category to which the document belongs and the category with the highest certainty factor acquired in step S1704. .

  In S1506, the score calculated in S1705 is reset as the score of the text document under consideration.

  The CPU 201 of the information processing apparatus 100 applies the processing in steps S1502 to S1506 in FIG. 17 to all search result documents, and then applies the processing to the re-scored search result document list ( For example, the search result 1602 in FIG. 16 is determined as information of the search result, and stored / updated on the RAM 202. That is, search result output processing is performed. This is the end of the description of the workflow processing in FIG.

  The eighth embodiment of the present invention has been described above. According to the eighth embodiment of the present invention, a more accurate search result can be specified in consideration of the certainty of the category and the relationship between the categories.

  As described above, according to the present invention, it is possible to provide a mechanism capable of specifying a more appropriate search result from the search result using the certainty factor of the category to which the search result belongs.

  It should be noted that the configuration and contents of the various data described above are not limited to this, and it goes without saying that the various data and configurations are configured according to the application and purpose.

  Further, the input character string receiving unit 101 may be a functional unit that receives from another application program that operates on the computer.

  Moreover, when the information processing apparatus 101 according to the present invention is a server on a network, the input character string reception unit 101 may be a functional unit that receives communication information from another information processing apparatus via the network.

  The configuration of this text document is merely an example. It should be noted that the “category” as a constraint condition for narrowing down the search target in the present invention does not necessarily have to be the name of the aforementioned category. This name is only an example. For example, the name “classification” may be used.

  Any material may be used as long as it indicates the contents of individual text documents (content classification based on a certain standard).

  Further, although the search target is a text document, it is not necessarily a text document. For example, it may be a PDF document or image data in which text is described.

  In addition, the category identification standard generation method using machine learning is merely an example of a generation method. For example, a human profile is created by, for example, examining a text document previously assigned with a category to be searched by a human and extracting the feature to extract a word that is a feature of each category. Also good. Needless to say, the category identification standard may be generated using a method such as natural language processing or statistical analysis other than machine learning.

  Although several embodiments have been described above, the present invention can take an embodiment as, for example, a system, apparatus, method, computer program, or recording medium, and more specifically, a plurality of devices. The present invention may be applied to a system configured from the above, or may be applied to an apparatus including a single device.

  For example, an information processing system in which the client PC includes the input character string receiving unit 101 and the category specifying unit 102, the category acquiring unit 103, and the document searching unit 104 operate on different server devices is configured. The servers may transmit and receive each other to execute processing by each unit, and the result may be transmitted to a client PC (client device) that can communicate with each server for display output.

  Further, for example, when the server that performs the processing in FIG. 10 receives the information for which the processing in step S903 has been completed from the server of another device, the processing in FIG. 10 is performed, and when the processing in FIG. For a server device having a search function (for example, a function for executing the processes of FIGS. 15 and 17), a search based on the input character string is performed via the network using the constraint category selected in the process of FIG. 10 as a constraint condition. The search control instructing to perform may be performed, and the server device that receives the instruction may perform the document search. In this case, it is assumed that the instruction includes a restriction category list and an input character string.

  In addition, instead of steps S901 to S903, input characters using a publicly available API for a service for specifying a category to which an input character string belongs and a score providing service provided via a network such as WAN, for example. A sequence may be transmitted to identify a plurality of categories to which the input character string may belong and obtain scores for each category. That is, the processing in steps S901 to S903 and the processing in step S904 may be executed by separate devices.

  The computer program according to the present invention is a computer program capable of executing the processing method of the flowcharts shown in FIGS. 9 to 13, 15, 17, and 18, and the storage medium of the present invention is illustrated in FIGS. 13, a computer program capable of executing the processing methods of FIGS. 15, 15, and 18 is stored.

  Note that the computer program in the present invention may be a computer program for each processing method of each apparatus in FIGS. 9 to 13, 15, 17, and 18.

  As described above, a recording medium in which a computer program for realizing the functions of the above-described embodiments is recorded is supplied to the system or apparatus, and the computer (or CPU or MPU) of the system or apparatus is stored in the recording medium. It goes without saying that the object of the present invention can also be achieved by reading and executing a program.

  In this case, the computer program itself read from the recording medium realizes the novel function of the present invention, and the recording medium storing the computer program constitutes the present invention.

  As a recording medium for supplying a computer program, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a DVD-ROM, a magnetic tape, a nonvolatile memory card, a ROM, an EEPROM, Silicon disks, solid state drives, etc. can be used.

  Further, by executing the computer program read by the computer, not only the functions of the above-described embodiments are realized, but also an OS (operating system) running on the computer based on the instructions of the computer program. It goes without saying that a case where the function of the above-described embodiment is realized by performing part or all of the actual processing and the processing is included.

  Furthermore, after the computer program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function is based on the instructions of the computer program code. It goes without saying that the CPU or the like provided in the expansion board or the function expansion unit performs part or all of the actual processing and the functions of the above-described embodiments are realized by the processing.

  It goes without saying that the present invention can also be applied to a case where the present invention is achieved by supplying a computer program to a system or apparatus. In this case, by reading the recording medium storing the computer program for achieving the present invention into the system or apparatus, the system or apparatus can enjoy the effects of the present invention.

  Furthermore, by downloading and reading out a computer program for achieving the present invention from a server, database, etc. on a network using a communication program, the system or apparatus can enjoy the effects of the present invention.

  In addition, all the structures which combined each embodiment mentioned above and its modification are also included in this invention.

DESCRIPTION OF SYMBOLS 100 Information processing apparatus 101 Input character string reception part 102 Category specification part 103 Category acquisition part 104 Document search part 121 Category specification reference | standard storage part 122 Restriction category selection reference | standard storage part 123 Similar category storage part 124 Text document storage part 440 Category reliability list 700 Constraint category selection criteria 702 Category selection criteria 703 Minimum criteria 704 Additional category selection criteria

Claims (10)

Input character string receiving means for receiving an input character string serving as a search condition, a plurality of categories corresponding to the input character string, and specifying means for specifying the certainty of correspondence between each category and the input character string An information processing apparatus,
Selection means for selecting a plurality of categories for searching for a search object based on the certainty information of the plurality of categories specified by the specifying means;
Search control means for controlling to perform search processing using the category selected by the selection means as a constraint;
An information processing apparatus comprising: When the category selected by the selection unit does not satisfy a criterion that is a restriction condition for search processing, the selection unit adds another category identified by the identification unit until the criterion is satisfied. Select as a constraint,
The information processing apparatus according to claim 1, wherein the search control unit controls the search process to be performed using a category selected by the selection unit and satisfying the criterion as a constraint.   The information processing apparatus according to claim 2, wherein the criterion is that a total of certainty factors of a plurality of categories selected by the selection unit reaches a specified value.   The information processing apparatus according to claim 2, wherein the selection unit selects a category in descending order of certainty from a plurality of categories specified by the specification unit until the criterion is satisfied.   The selection unit resets a category already selected when it is determined that the category selected by the selection unit performed by the selection unit a specified number of times does not satisfy the criterion. 5. The information processing apparatus according to claim 2, wherein the control is performed so as to release the restriction condition. The information processing apparatus further stores a similar relationship between categories,
The selection unit selects another category similar to the category as the constraint condition based on the similar relationship of the category having the highest certainty among the plurality of categories specified by the specifying unit. The information processing apparatus according to claim 1, wherein:   The selection means does not select a category as a constraint condition when the category having the highest certainty factor is stored in association with information indicating that the category should not be selected as a constraint condition. The information processing apparatus according to claim 6. Input character string receiving means for receiving an input character string serving as a search condition, a plurality of categories corresponding to the input character string, and specifying means for specifying the certainty of correspondence between each category and the input character string A method for controlling an information processing apparatus,
A selection step of selecting a plurality of categories for searching for a search object based on the certainty information of the plurality of categories specified by the specifying means;
A search control step for controlling to perform a search process using the category selected in the selection step as a constraint;
A method for controlling an information processing apparatus, comprising: Input character string receiving means for receiving an input character string serving as a search condition, a plurality of categories corresponding to the input character string, and specifying means for specifying the certainty of correspondence between each category and the input character string A program that can be executed by an information processing device,
The information processing apparatus;
Selection means for selecting a plurality of categories for searching for a search object based on the certainty information of the plurality of categories specified by the specifying means;
A program for an information processing apparatus that functions as search control means for controlling to perform search processing using a category selected by the selection means as a constraint. A client device including an input character string receiving unit that receives an input character string as a search condition, a plurality of categories to which the input character string corresponds, and a specifying unit that identifies the certainty of correspondence between each category and the input character string An information processing apparatus system comprising: an information processing apparatus comprising:
Selection means for selecting a plurality of categories for searching for a search object based on the certainty information of the plurality of categories specified by the specifying means;
Search control means for controlling to perform search processing using the category selected by the selection means as a constraint;
An information processing system comprising:

Images