JP4699909B2 - Keyword correspondence analysis apparatus and analysis method - Google Patents

Description

  The present invention relates to a keyword correspondence analysis device and an analysis method for analyzing the correspondence between keywords included in document information such as patent documents and academic documents.

  For example, by investigating patent documents as document information, it is possible to find out the direction of research and development or to make business plans. For this purpose, patent documents that match a predetermined purpose are extracted from a large group of patent documents, and the extracted patent documents are analyzed. A representation of this analysis result on a two-dimensional plane is known as a so-called patent map, and a technique for automatically creating a patent map from a group of patent documents is known (Patent Document 1).

In addition, by performing morphological analysis and extracting terms that appear more than a certain number in the target document, even technical terms that are not registered in the dictionary or abbreviations that are used only in specific organizations are detected as terms A technique that enables this is also known (Patent Document 2).
JP 2005-149346 A JP 2002-342321 A

  In patent documents and the like, it is preferable to use terms as accurate as possible. However, since patent documents are directed to the latest technology, they have the property that they are easily created before exact terms are established. Since patent documents are created using terms used in the industry at the time of filing, terms indicating the same technical element may vary depending on the applicant.

  Even if the term is not the latest, the term commonly used by each applicant (company, etc.) may be different. For example, some applicants use formal names established in Japanese Industrial Standards, others use formal name abbreviations, and other applicants are used only within the organization. There are cases where special terms are used.

  Furthermore, terms have the property of changing with the times. Therefore, even a standard term in one era may be expressed in another term in another era.

  When terms indicating the same technical element are expressed by various different terms, the search accuracy by machine search generally decreases. Searchers who are familiar with both technical knowledge and search technology will be able to obtain search results with low search omissions and noise, but searchers who lack either technical knowledge or search technology. In this case, it is difficult to obtain a highly accurate search result.

  Moreover, even if it is an expert searcher, when the same technical vocabulary is expressed with a different term, the time and effort required for a search increase. This is because, for example, it is necessary for a searcher to read a large number of patent documents in advance and study the nicknames of terms before assembling a search expression.

  Furthermore, when reading a patent document in which a term different from the term that is normally used is read, it is difficult to grasp the correct contents in a short time, which may lead to confusion and misunderstanding.

  The present invention has been made in view of the above-described problems, and an object thereof is to provide a keyword correspondence analysis device and an analysis method capable of automatically detecting a keyword used only in specific document information. There is to do. Another object of the present invention is to automatically detect a correspondence relationship between a standard keyword used in a large amount of document information and a keyword used only in specific document information, and to support keyword search. To provide a correspondence analysis apparatus and an analysis method. Still another object of the present invention is to provide a keyword correspondence analysis device and an analysis method that can support grasping the contents of document information in which non-standard keywords are used. Further objects of the present invention will become clear from the description of the embodiments described later.

  In order to solve the above problem, a keyword correspondence analysis device according to one aspect of the present invention includes a document information storage unit that stores a plurality of digitized document information, and a document information storage unit based on a given analysis condition The keyword appearance frequency obtained by analyzing the application frequency for each keyword appearing in the document information based on the extracted document information and the target document extraction unit that extracts the document information to be analyzed by searching An appearance frequency analysis unit that generates analysis information, and a first keyword that detects a first keyword used in predetermined document information among keywords included in the extracted document information based on the keyword appearance frequency analysis information Based on the detection unit and the keyword appearance frequency analysis information, second keyword candidates that are candidates for the second keyword corresponding to the first keyword are extracted. A second keyword candidate detecting unit that detects from keywords included in other document information other than the predetermined document information, and a second corresponding to the first keyword from the detected second keyword candidates A second keyword detection unit for detecting a keyword.

  Examples of the document information include patent documents (including open gazettes, registered gazettes) and academic papers. The document information is digitized and stored in the document information storage unit. Of the plurality of document information stored in the document information storage unit, the document information to be analyzed is extracted based on the given analysis condition. More specifically, the literature information extracted as the analysis target is literature information related to a specific technical field.

  The appearance frequency analysis unit calculates each appearance frequency for each keyword included in each extracted document information, for example, assigns a ranking in descending order of appearance frequency to generate keyword appearance frequency analysis information To do. The first keyword detection unit detects a first keyword used in predetermined document information. The first keyword is, for example, a standard keyword used in predetermined document information of a predetermined number or more. The second keyword candidate detection unit detects a keyword that is a candidate for the second keyword corresponding to the first keyword. The second keyword is a special keyword used only in other document information other than the predetermined document information, and can be called a so-called dialect keyword. Then, the second keyword detecting unit detects a second keyword corresponding to the first keyword from the second keyword candidates. Thereby, it is possible to automatically detect the correspondence between keywords having different expressions.

  In one aspect of the present invention, the second keyword detection unit detects the second keyword candidate selected by the user from the second keyword candidates as the second keyword. For example, a candidate list of second keyword candidates can be presented to the user, and the second keyword can be selected by the user. For example, in the candidate list, the second keyword candidates are listed in order from the one that is most likely to correspond to the first keyword.

  In one aspect of the present invention, a first feature information detection unit that detects feature information of the detected first keyword, and a second feature information detection unit that detects feature information of the detected second keyword candidate are further included. The second keyword detection unit detects the second keyword from the second keyword candidates by comparing the detected feature information of the first keyword with the detected feature information of the second keyword candidate. As the feature information, for example, the coordinates of each keyword in the coordinate system generated based on the extracted document information, the ranking of the appearance frequency of each keyword, and the like can be used. The second keyword detection unit can detect, as the second keyword, a second keyword candidate that has the least difference in feature information from the first keyword.

  In one aspect of the present invention, a first feature information detector that detects feature information of a detected first keyword, a second feature information detector that detects feature information of a detected second keyword candidate, A feature information comparison unit that compares the feature information of the keyword with the feature information of the second keyword, and a comparison result output unit that outputs a comparison result by the feature information comparison unit, and the second keyword detection unit includes the comparison result Based on the above, the user specification mode for detecting the second keyword candidate selected by the user among the second keyword candidates as the second keyword, and the feature having the smallest difference between the feature information of the first keyword based on the comparison result And an automatic detection mode for detecting a second keyword candidate having information as a second keyword.

  In one aspect of the present invention, the first keyword detection unit extracts a document that is equal to or more than the second predetermined value and extracted from keywords having the appearance frequency rank up to the first predetermined value based on the keyword appearance frequency analysis information. A keyword used in predetermined document information less than the total number of information is detected as a first keyword.

  In one aspect of the present invention, the second keyword candidate detection unit uses the predetermined document information as a third predetermined keyword out of keywords included in other document information other than the predetermined document information based on the keyword appearance frequency analysis information. Keywords remaining after removal of keywords that are used in excess of the value are detected as second keyword candidates.

  In one aspect of the present invention, the first feature information detecting unit detects the average appearance frequency rank in the predetermined document information of the first keyword as the feature information of the first keyword, and the second feature information detecting unit is the predetermined document. The appearance frequency rank of the second keyword candidate in the document information other than the information is detected as the feature information of the second keyword.

  In one aspect of the present invention, the principal component analysis is performed based on the total number of extracted document information and the number of appearances of a plurality of predetermined keywords extracted from the extracted document information, so that it is included in the extracted document information The first feature information detection unit further detects the first keyword coordinates calculated by the keyword coordinate calculation unit as feature information of the first keyword, and calculates a second feature. The information detection unit detects the coordinates of the second keyword candidate calculated by the keyword coordinate calculation unit as the feature information of the second keyword candidate.

In one aspect of the present invention, the first feature information detection unit detects a ranking of words that are dependent on the first keyword in the predetermined document information as the first feature information,
The second feature information detection unit is configured to detect a ranking of words that are dependent on the second keyword candidate in other document information other than the predetermined document information as the second feature information.

  In one aspect of the present invention, by performing principal component analysis based on a combination of a plurality of predetermined keywords extracted from the extracted document information and the number of appearances, the coordinates of each document information of the extracted document information group are obtained. Keyword coordinates for calculating the coordinates of each predetermined keyword by performing a principal component analysis based on the total number of document information including each predetermined keyword and the number of occurrences of each predetermined keyword. The distribution density of each document information is calculated based on the coordinates of each document information calculated by the calculation unit and the document coordinate calculation unit, and a map figure having an outline based on the calculated distribution density and a predetermined keyword are visualized respectively. And a map generation unit for generating map information, wherein the first keyword and the second keyword are visualized in the map information. That.

  Here, the document coordinate calculation unit performs principal component analysis based on the combination of a plurality of predetermined keywords and the number of appearances of the predetermined plurality of keywords, and calculates the coordinates of the extracted document information, respectively. Principal component analysis is a method of multivariate analysis. To put it simply, each sample information that contains many variables is represented by several comprehensive indicators (principal components) that most directly represent the differences. This is an analysis method that reduces the number of dimensions.

  Similarly, the keyword coordinate calculation unit calculates the coordinates of each predetermined keyword by principal component analysis. The map generation unit calculates the distribution density of the document information based on the coordinates of each document information. For example, the map generation unit can detect the density of the document information for each block by dividing the entire area of the map into a plurality of blocks and calculating the number of document information located in each block. For example, the map generation unit creates a map figure having an outline by associating display elements (for example, contour lines) with the distribution density of the document information. The map generation unit generates map information by arranging the map graphic and a display element indicating the presence of the keyword (for example, a keyword character itself or a symbol) on the map area. This map information is expressed, for example, on a two-dimensional plane, but is not limited thereto, and can be expressed in a three-dimensional space.

  The second keyword detection unit can detect the second keyword from the second keyword candidates by comparing the coordinates of the first keyword and the coordinates of the second keyword candidate respectively calculated by the keyword coordinate calculation unit. .

  The first keyword and the second keyword are visualized in map information in different display forms, and the map information can include a display element indicating a correspondence relationship between the first keyword and the second keyword.

  In one aspect of the present invention, a keyword replacement unit that replaces the second keyword with the first keyword is provided. For example, the keyword replacement unit can output the other document information in a state where the second keyword included in the other document information is replaced with the first keyword.

  In one aspect of the present invention, a relevance registration unit is provided that stores the second keyword in association with the first keyword.

  A program according to another aspect of the present invention includes a computer, a document information storage unit that stores a plurality of digitized document information, and a target document extraction unit that extracts document information to be analyzed based on a given analysis condition And, based on the extracted document information, for each keyword appearing in the document information, based on the keyword appearance frequency analysis information, the appearance frequency analysis means for generating keyword appearance frequency analysis information obtained by analyzing the application frequency The first keyword detection means for detecting the first keyword used in the predetermined document information among the keywords included in the extracted document information and the first keyword based on the keyword appearance frequency analysis information The second keyword candidate that is a candidate for the second keyword to be used is the document information other than the predetermined document information among the extracted document information. A second keyword candidate detection means for detecting from the Murrell keyword, second keyword detection means for detecting a second keyword corresponding from the second keyword candidates detected in the first keyword, thereby to function.

  A keyword correspondence analysis method according to still another aspect of the present invention includes a step of acquiring analysis conditions, and a step of extracting document information to be analyzed by searching a document information storage unit based on the acquired analysis conditions And, based on the extracted document information, for each keyword appearing in the document information, generating a keyword appearance frequency analysis information obtained by analyzing the application frequency, and extracting based on the keyword appearance frequency analysis information The second keyword corresponding to the first keyword is selected based on the step of detecting the first keyword used in the predetermined document information from the keywords included in the document information and the keyword appearance frequency analysis information. The second keyword candidate is a keyword included in other document information other than the predetermined document information among the extracted document information. Includes a step of detecting, the step of detecting a second keyword corresponding to the first keyword from among the second keyword candidates detected, from among the.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. First, a case where the keyword correspondence analysis device is configured alone will be described, and then a case where the keyword correspondence analysis device is incorporated in the document information analysis device will be described.

  FIG. 1 is an explanatory diagram showing the overall configuration of a keyword correspondence analysis device (hereinafter, sometimes abbreviated as “correspondence analysis device”). As will be described later, this correspondence analysis device includes, for example, an analysis condition setting unit 1, a target document extraction unit 2, an information storage unit 3, a keyword ranking generation unit 4, and a reference keyword detection unit 6. Keyword candidate detection unit 7, reference keyword feature information detection unit 8, corresponding keyword candidate feature information detection unit 9, feature information comparison unit 10, corresponding keyword detection unit 11, keyword correspondence output unit 12, and thesaurus registration A unit 14 and a keyword replacement unit 15 can be provided. The correspondence analysis device is configured as a computer device or a program that causes a computer to realize a predetermined function.

  The analysis condition setting unit 1 sets analysis conditions specified by the user. The user can specify various analysis conditions. For example, the user can analyze various conditions alone or in combination, such as a specific company, a specific technical field, a specific document creation or publication period, a specific inventor, a specific technical term, etc. Can be specified. For example, the user can specify an analysis condition such as “patent document filed within a specific period for a specific technical field by a specific company”. By setting analysis conditions, a population to be analyzed is specified.

  The target document extraction unit 2 extracts document information that matches the analysis conditions from the information storage unit 3. The information storage unit 3 is provided with a document information storage unit 3A, a storage unit 3B for storing keywords, and the like. In the document information storage unit 3A, for example, various documents such as patent documents, scientific and technical documents, and academic papers are stored in an electronic state. The storage unit 3B that stores keywords and the like stores, for example, main keywords used in each document information and information such as a dictionary.

  The keyword ranking generation unit 4 generates a keyword ranking table 5 by analyzing the extracted document information. The keyword ranking generation unit 4 corresponds to “appearance frequency analysis unit”, and the keyword ranking table 5 corresponds to “keyword appearance frequency analysis information”. A keyword ranking generation unit (hereinafter also abbreviated as a ranking generation unit) 4 extracts keywords used in each extracted document information, and counts the appearance frequency for each keyword. Then, the keyword ranking table 5 is generated by ranking in the order of appearance frequency.

  Here, as a keyword extraction method, various known methods can be employed. Further, for example, keywords such as “glass” and “crow” that are expressed by replacing the dakuten with other symbols or characters can be detected as a group of keywords. In this case, for example, a katakana character block set included in the sentence may be extracted by morphological analysis.

  In addition, when an unnecessary space code or line feed code is included in a keyword, it can be extracted as a keyword by deleting the space code or line feed code. Therefore, even if a keyword is described in a different expression from the standard expression due to the addition of a cloud point or extra code, it can be detected as a keyword, and the correspondence with the standard keyword can be detected. Can do.

  In the keyword ranking table 5 shown in FIG. 1, the keywords used in the literature information group relating to the specific technical field respectively created by the three applicants A company, B company, and C company are ranked first to third. Shown up to.

The reference keyword detection unit 6 detects a keyword as a reference, and corresponds to a “first keyword detection unit”. Here, the reference keyword means a keyword that is used in a lot of document information but is not used in some document information. That is, the reference keyword can be expressed as a keyword that is used as standard in the literature information group that constitutes the analysis population, or a mainstream keyword that is used by the majority. Of the document information group constituting the analysis population, document information including the reference keyword corresponds to “predetermined document information”, and document information not including the reference keyword is referred to as “other document information other than the predetermined document information”. Applicable.
Strictly speaking, even a keyword used in a large number of document information is not necessarily an official name. Abbreviations may be used more frequently than formal names. In addition, as is known as a common name or a common name of a registered trademark, a registered trademark may be used rather than an official name as a result of a registered trademark of a specific company becoming famous.

  In the example shown in the keyword ranking table 5 in FIG. 1, the keyword W2 corresponds to the reference keyword. The keyword W2 has the second highest appearance frequency in the company A and the first appearance frequency in the company B, but does not appear in the company C. Although it appears in many pieces of literature information in the analysis population, it does not appear in some literature information, so the keyword W2 is detected as a reference keyword.

  The corresponding keyword candidate detection unit 7 detects a keyword that can be a candidate for the corresponding keyword, and corresponds to a “second keyword candidate detection unit”. The corresponding keyword is a keyword that is considered to correspond to the reference keyword in the literature information that does not include the reference keyword. Corresponding keywords can be thought of as so-called dialect keywords that deviate from the standard expression. Corresponding keywords in this embodiment, that is, dialectal keywords, include abbreviations and different notations in addition to minor expressions that are commonly used within a specific organization.

  An analysis population composed of specific document information is considered to include one or more common keywords that match the analysis conditions. In particular, keywords whose appearance frequency rank is within a predetermined range are likely to be common to each applicant. However, in some applicants (Company C), the keyword W2 is not used. This means that some applicants may describe specific technical elements using different expressions different from the standard expressions. Therefore, the corresponding keyword candidate detection unit 7 detects a keyword corresponding to the reference keyword from the literature information group of some applicants. In the example shown in FIG. 1, among the keywords appearing in the document information of company C, the second-ranked W4 and the third-ranked W5 correspond to corresponding keyword candidates. Since the keyword W1 is used by all applicants, it is excluded from the corresponding keyword candidates.

  The reference keyword feature information detection unit 8 detects feature information that is considered useful for examining the relevance with the corresponding keyword among the attribute information of the reference keyword, and is referred to as a “first feature information detection unit”. Applicable.

  Similarly, the corresponding keyword candidate feature information detecting unit 9 detects feature information of the corresponding keyword candidate, and corresponds to a “second feature information detecting unit”.

  Here, as the attribute information, for example, the character type of the keyword (hiragana, katakana, alphanumeric characters, symbols, kanji characters), the length of the character string, and the like are known. Focus on the position (coordinates) on the patent map. The feature information can also be expressed as, for example, relevance determination information for determining relevance between keywords, proximity determination information for determining closeness between keywords, and the like.

  Although the details will be described later, in the example of FIG. 1, the ranking of the appearance frequency of the reference keyword W1 is second in the company A and first in the company B, so the average rank is 1.5. . The corresponding keyword candidate W4 is ranked second, and another corresponding keyword candidate W5 is ranked third.

  The feature information comparison unit 10 compares the feature information detected by the reference keyword feature information detection unit 8 with the feature information detected by the corresponding keyword candidate feature information detection unit 9. The corresponding keyword detection unit 11 detects a keyword corresponding to the reference keyword from the corresponding keyword candidates based on the comparison result of the feature information. The corresponding keyword detection unit 11 corresponds to a “second keyword detection unit”. In the above example, since the average appearance rank “1.5” of the reference keyword W1 is, the keyword W4 having the closest rank (second rank) to this is detected as the corresponding keyword.

The keyword correspondence output unit 12 outputs the reference keyword and the corresponding keyword in association with each other. For example, the correspondence relationship between keywords is presented to the user via the display unit 13 such as a display. As will be apparent from the examples described later, it is also possible to ask the user's judgment when detecting the corresponding keyword.
Further, the keyword ranking table 5 itself can be output to the display unit 13. When the keyword ranking table 5 is output on the screen, for example, the font type, font size, character color, etc. are appropriately set so that the correspondence between the reference keyword and the corresponding keyword (or corresponding keyword candidate) can be clearly distinguished. You can also

  The thesaurus registration unit 14 associates the detected corresponding keyword with the reference keyword and registers it in the dictionary, for example. Thereby, in the next search, a more improved dictionary can be used, and the search accuracy is improved.

  The keyword replacement unit 15 replaces the corresponding keyword with the reference keyword. Conversely, the reference keyword can be replaced with the corresponding keyword. Thereby, literature information can be read carefully by a standard expression, and a user's convenience improves.

  FIG. 2 is an explanatory diagram schematically showing the flow of data processing. The user specifies the analysis condition 1A via the user interface. In this example, document information created for a specific technical field is specified by specific applicants (creators) A to E. The technical field can be specified by, for example, a patent classification code, a technical term, or the like. Further, for example, when analyzing the research contents involving a predetermined limited researcher, the technical field can be specified by the name of the researcher.

  A document information group that matches the analysis condition 1A is extracted from the information storage unit 3. This extracted document information group constitutes a population to be analyzed. The keyword appearance frequency is analyzed for this analysis population, and the keyword ranking table 5 is generated.

  By analyzing the keyword ranking table 5, a reference keyword table 6A is generated. In the example shown in FIG. 2, the keyword “personal computer” shown in bold is detected as a reference keyword and stored in the reference keyword table 6A.

  The frequency of appearance of the keyword “personal computer” is 3rd in Company A, 4th in Company B, 2nd in Company C, and 3rd in Company D. In Company C, the official name “Personal Computer” is used, but in this example, the official name “Personal Computer” and the abbreviation “Personal Computer” have already been found to be synonymous, It is assumed that the information has been registered in the information storage unit 3.

  The keyword “personal computer (including personal computer)” appears in a high ranking in Company A to Company D, while Company E is within the ranking of the judgment target (in the example shown, first to sixth). Not used in. Therefore, the keyword “PC” is detected as the reference keyword.

  Then, keyword candidates that can correspond to the reference keyword are detected from the keyword group of Company E recorded in the keyword ranking table 5. The detected keywords are stored in the corresponding keyword candidate table 7A.

  In the example shown in FIG. 2, “software”, “hardware”, and “system” are common keywords used by all applicants A company to E company. Therefore, “software”, “hardware”, and “system” do not correspond to any of the reference keyword or the corresponding keyword.

  The keyword “program” is used by A company, B company, C company, and E company except for D company, and is not a common keyword, but is used by many applicants (ie, many). ) Is not a keyword corresponding to the reference keyword “computer”. In Company A, Company B, Company C, and Company E, “program” is considered to be used in the same meaning. Therefore, the keyword “program” used by a predetermined number or more of applicants is not a candidate for the corresponding keyword, and is excluded.

However, focusing on the keyword “program” for Company D, this “program” corresponds to another reference keyword. For convenience of explanation, the following description focuses on one reference keyword “personal computer”. However, the keyword correspondence analysis apparatus of the present embodiment can detect a plurality of reference keywords from the keyword ranking table 5, and Each keyword corresponding to the reference keyword can be detected.
In other words, even when multiple applicants use different dialect expressions (keywords different from the standard keyword) in the literature information group related to a specific technical field, each dialect expression is detected separately and supported. It is possible to detect a relationship with a standard expression (reference keyword).

  The keywords “information processing apparatus” and “speech recognition” used by company E are not used more than a predetermined number by other applicants A to D. Therefore, these “information processing apparatus” and “voice recognition” are detected as keywords that can correspond to the reference keyword “personal computer”, and are stored in the corresponding keyword candidate table 7A. In other words, Company E may represent the technical element “PC” as another keyword “Information processing device” or “Voice recognition”.

  Next, attention is paid to the feature information of the reference keyword and the corresponding keyword candidate. Here, keyword appearance frequency ranking is adopted as feature information. The ranking of the reference keyword “PC” is No. 3 for Company A, No. 4 for Company B, No. 2 for Company C, and No. 3 for Company D, so the average value is “3”. On the other hand, the ranking of the corresponding keyword candidate “information processing apparatus” is third, and the ranking of another corresponding keyword candidate “speech recognition” is sixth. Therefore, the “information processing apparatus” having a lower rank is detected as a keyword corresponding to the reference keyword “PC”.

  Instead of selecting only the corresponding keyword candidates that are closest in rank to the reference keyword, all the corresponding keyword candidates that have ranks within a predetermined range from the average rank of the reference keywords are selected to generate the keyword correspondence table 12A. be able to.

  In the keyword correspondence table 12A, the reference keyword “PC” and the corresponding keyword “information processing apparatus” are stored in association with each other. This correspondence is presented to the user via the display unit 13. This correspondence is reflected in the information storage unit 3 to improve the dictionary included therein.

  FIG. 3 is a flowchart showing the keyword correspondence analysis process. Each flowchart described below shows an outline of processing to the extent necessary for understanding and implementing the invention, and is different from an actual computer program. In the following description, step is abbreviated as “S”.

  The correspondence analysis device acquires the analysis condition 1A designated by the user (S11), and extracts a document information group that matches the analysis condition 1A from the information storage unit 3 (S12). Then, the correspondence analysis device generates the keyword ranking table 5 based on the extracted document information group (S13). The step of generating the keyword ranking table 5 will be described in detail with reference to FIG.

  Next, the correspondence analysis device generates a reference keyword table 6A based on the keyword ranking table 5 (S14), and further calculates the average rank of the reference keywords (S15). The step of generating the reference keyword table 6A will be described in detail with reference to FIG.

  The correspondence analysis device generates a correspondence keyword candidate table 7A based on the keyword ranking table 5 (S16), and acquires the ranks of the detected correspondence keyword candidates, respectively (S17). The step of generating the corresponding keyword candidate table 7A will be described in detail with reference to FIG.

  Then, the correspondence relationship analysis apparatus compares the average ranking of the reference keywords with the ranking of each corresponding keyword candidate (S18), associates the correspondence keyword candidates having similar ranks with the reference keyword, and generates the keyword correspondence table 12A. (S19).

  As described above, only one corresponding keyword candidate can be registered in the keyword correspondence table 12A, or a plurality of corresponding keyword candidates can be registered. First, a method of registering only the keywords having the ranking closest to the average ranking of the reference keywords in the table 12A is conceivable. Second, a method of registering a plurality of keywords having a rank within a predetermined range from the average rank of the reference keywords in the table 12A can be considered. Thirdly, a method of switching between the first method and the second method by a user's prior or subsequent selection is also conceivable.

  Here, a case will be described in which a plurality of corresponding keywords (candidates) are presented to the user by the second or third method. The user can select any one of the presented keywords (S21).

  In the example of FIG. 2, the user can manually select which of “information processing apparatus” and “voice recognition” is a keyword corresponding to “personal computer”. When selected by the user (S21: YES), the correspondence analysis device associates the selected keyword as a keyword corresponding to the reference keyword (S22). When the manual selection by the user is not performed (S21: NO), the correspondence analysis device automatically selects the keyword having the closest rank and associates it with the reference keyword (S23). By this association (S22, S23), the corresponding keyword is determined.

  The correspondence analysis device can use the correspondence between keywords in various ways. For example, when the user wishes to register a thesaurus (S24: YES), the correspondence relationship analysis apparatus registers the corresponding keyword as a thesaurus for the reference keyword (S25). Further, for example, when the user desires to replace the corresponding keyword (S26: YES), the corresponding relationship analysis device is selected by the user from all the literature information created by the company E included in the analysis population, or by the user. For a part of the document information, the corresponding keyword is converted into the reference keyword (S27). On the other hand, the reference keyword can be converted into a corresponding keyword.

  Next, FIG. 4 is a flowchart showing details of the keyword ranking table generation process shown in S13 of FIG. First, the correspondence analysis device selects one applicant from the applicants in the analysis condition 1A (S130), and acquires document information created by the selected applicant (S131).

  Then, the correspondence analysis device removes unnecessary words such as particles, fixed phrases, and headings from the document information (S132), and extracts only keywords (S133). Then, the correspondence analysis device calculates the number of appearances for each extracted keyword (S134), rearranges the keywords in descending order of appearance frequency (S135), and registers them in the keyword ranking table 5 (S136). ).

  The correspondence analysis device determines whether or not the keyword appearance frequency has been analyzed for all applicants in the analysis condition 1A (S137). If there are unanalyzed applicants remaining (S137: NO), the next The applicant is selected (S138), and the process returns to S131. In this manner, the keywords used in the document information and their appearance frequencies are detected for each applicant.

  FIG. 5 is a flowchart showing the reference keyword table generation process shown in S14 of FIG. The correspondence analysis device refers to the keyword ranking table 5 (S140), and acquires keywords up to the determination target rank N1 (S141). The order N1 of this determination target can be specified by the user. In the example shown in FIG. 2, “6” corresponds to N1.

  Next, the correspondence analysis device detects the usage status by each applicant for the keywords up to rank N1 (S142), and excludes the common keywords used by all applicants from the reference keyword (S143). .

  Next, the correspondence analysis device detects keywords used by a predetermined number N2 or more of applicants (S144), and registers the detected keywords as reference keywords in the reference keyword table 6A (S145). The value of N2 can be specified by the user. N2 is set to a value of 2 or more and less than the total number of document information of the analysis population.

  Then, the correspondence analysis device stores the name or identification code of the applicant who uses the reference keyword (S146). An applicant who uses a reference keyword can be referred to as a reference applicant. Similarly, the correspondence analysis apparatus stores the name or identification code of the applicant who does not use the reference keyword (S147). An applicant who does not use the reference keyword can be called a target applicant. The name or identification code of the reference applicant (information that can identify the applicant in the apparatus) is registered in the reference keyword table 6A. The name of the subject applicant is used to generate the corresponding keyword candidate table 7A.

  FIG. 6 is a flowchart showing the corresponding keyword candidate table generation process shown in S16 of FIG. The correspondence analysis device refers to the keyword ranking table 5 (S160), and acquires keywords up to rank N1 used in the literature information group created by the subject applicant (S161).

  For each acquired keyword, the correspondence analysis device detects the usage status by the reference applicant (S162), and excludes keywords used by the reference applicant of a predetermined number N3 or more from the corresponding keyword candidates (S163). ). In this way, unnecessary keywords are removed from the keywords up to the rank N1 used by the subject applicant. Then, the correspondence relationship analysis apparatus registers the remaining keywords as correspondence keyword candidates in the correspondence keyword candidate table 7A (S164).

Since this embodiment is configured as described above, the following effects can be obtained. In the present embodiment, even when the same technical element is expressed by a different term for the applicant, it is possible to analyze the correspondence relationship between the keywords and detect the keyword corresponding to the reference keyword. Therefore, it is possible to search documents in a technical field in which terms are not unified more efficiently and with higher accuracy, and user convenience is improved.
That is, according to the present embodiment, the correspondence relationship between the document information group using the reference keyword (standard expression) and the document information group using the corresponding keyword (dialectual expression) deviating from the standard expression is grasped. be able to. In other words, dialectal expressions are extracted from the literature information group that uses expressions different from standard expressions, and the correspondence between the extracted dialect expressions and standard expressions is visualized. Can show.

  In this embodiment, the keyword corresponding to the reference keyword is automatically or semi-automatically detected based on the keyword ranking table 5. Accordingly, it is possible to analyze the correspondence between keywords while having a relatively simple configuration.

  In the present embodiment, the combination of corresponding keywords is detected by comparing the appearance frequency ranks of the keywords. Therefore, the correspondence relationship between keywords can be analyzed with a relatively simple control configuration using the keyword ranking table 5.

  In this embodiment, the correspondence relationship between keywords can be registered as a thesaurus. Therefore, the dictionary in the information storage unit 3 can be improved, which can be used for the next search, and the convenience for the user is improved.

  In this embodiment, the corresponding keyword can be replaced with the reference keyword. Therefore, by translating unique expressions used by some applicants into standard expressions, it can be used for grasping the contents, and user convenience is improved. In other words, by replacing the detected dialect keyword with the standard keyword for the entire original “naked” specification, it is possible to make it easier to understand the “naked” specification.

  A second embodiment will be described with reference to FIGS. In the second embodiment, the keyword correspondence analysis apparatus described in the first embodiment is incorporated in the document information analysis apparatus 100 for automatically creating a map.

  The document information analysis apparatus 100 can be configured as a computer apparatus including a map control unit 110, a keyword correspondence analysis unit 120, and an information storage unit 300. The document information analyzing apparatus 100 is connected to the client terminal 200 so as to be capable of bidirectional communication via a communication network such as the Internet or a LAN (Local Area Network). The client terminal 200 can be configured as, for example, a personal computer or a portable information terminal (including a mobile phone).

  The map control unit 110 generates and outputs a technical map 150 based on a large number of document information stored in the information storage unit 300. The generated map 150 is transmitted to the client terminal 200 via the communication network. The map 150 may be storable in the client terminal 200, or may not be storable in the client terminal 200. Details of the map control unit 110 will be described later.

  The keyword correspondence analysis unit 120 is mapped on the map 150 based on the analysis conditions designated by the user and analyzes the correspondence between the keywords. The keyword correspondence analysis unit 120 has the same function as the keyword correspondence analysis device described above.

  The information storage unit 300 stores a large number of document information such as patent publications, registration bulletins, and scientific and technical papers.

  The configuration of the map 150 will be described. The map 150 is generated according to the purpose instructed by the user. For example, the user can select only a predetermined document information group from among a large number of document information groups stored in the information storage unit 300 by designating a technical field desired to be investigated, a specific search word, or the like. it can. The content of the document information group selected by the user is analyzed by the map control unit 110, and a map 150 is generated.

  The contour lines 151A, 151B, 151C, and 151D are display elements indicating the distribution density of the document information included in the map 150. For example, the contour line 151A indicates that the number of document information existing therein is 1 or more and less than B1, and the contour line 151B indicates that the number of document information existing therein is B1 or more and less than B2 (B1, B2 is a natural number). In this way, the number of document information included therein increases step by step from 151A toward 151B, 151C, and 151D.

  A plurality of keywords 152 are also displayed on the map 150. These keywords 152 are main keywords that characterize each piece of document information of the document information group constituting the map 150, and may be referred to as the main keywords 152 in the following description.

Of the main keywords 152, one or more keywords become reference keywords 152A, and one or more keywords become corresponding keywords 152B. The reference keyword 152A and the corresponding keyword 152B are connected by a connection line 153 for indicating the correspondence between them. The reference keyword 152A and the corresponding keyword 152B can be made conspicuous on the map 150, for example, by changing the color, line type, line thickness, and the like from the other keywords 152. Alternatively, the corresponding keywords 152A and 152B can be blinked to distinguish them from other keywords 152.
It should be noted that the content disclosed in Japanese Patent Application Laid-Open No. 2005-149346 can be used within the scope necessary for carrying out the present invention.

  FIG. 8 is a block diagram illustrating a functional configuration of the document information analysis apparatus 100. For example, the map control unit 110 includes a map generation condition input reception unit 111, a document extraction unit 112, a document coordinate calculation unit 113, a keyword extraction unit 114, a keyword coordinate calculation unit 115, a map generation unit 116, and a map display. It can comprise including the part 117. FIG.

  Further, as an example of the information storage unit 300, a document database (in the figure, database is abbreviated as “DB”) 31, a word database 32, and an index database 33 can be provided.

  The document database 31 stores a plurality of document information. The word database 32 is used as a dictionary. The index database 33 is for managing what keywords are included in each document.

  The map generation condition input receiving unit 111 is for receiving an input of a map generation condition designated by the user. Accepting means, for example, receiving information indicating map generation conditions designated by the user as electronic information via a communication interface and storing the information in a memory or the like.

  The document extraction unit 112 extracts document information within a predetermined range by searching the document database 31 and the index database 33 based on conditions designated by the user. The document coordinate calculation unit 113 calculates coordinates on the map 150 by analyzing the extracted document information. For example, the document coordinate calculation unit 113 calculates the coordinates of each document on the two-dimensional plane by performing principal component analysis on the keyword combination and the number of appearances included in each extracted document information.

  The keyword extraction unit 114 extracts a plurality of keywords included therein by analyzing the extracted document information. The keyword coordinate calculation unit 115 calculates the coordinates of the extracted keywords. For example, the keyword coordinate calculation unit 115 calculates the coordinates of each keyword on the two-dimensional plane by performing principal component analysis on the total number and the total number of appearances of each piece of document information including the keyword. Principal component analysis is a technique for multivariate analysis, and is a known technique, so details are omitted. The calculated keyword coordinates are also used by the keyword correspondence analysis unit 120.

  The map generation unit 116 generates the map 150 based on the calculated coordinates of each document information and the coordinates of each keyword. For example, the map generation unit 116 sets a large number of block areas by finely dividing a finite two-dimensional plane vertically and horizontally, and calculates the number of document information existing in each block area. Thereby, the map generation part 116 calculates | requires the distribution density of literature information, sets the contour line 151A etc. corresponding to this distribution density, and produces a map figure.

  The map display unit 117 generates a map 150 that can be visually recognized by the user by assigning predetermined display elements according to the configuration of the generated map 150, and provides the map 150 to the client terminal 200. Examples of the predetermined display element include an outline for indicating each contour line 151A and the like, characters for indicating each keyword, and the like. In addition, a menu display unit for performing operations on the map 150 is also added.

  The keyword correspondence analysis unit 120 includes, for example, a target document extraction unit 121, a keyword ranking generation unit 122, a reference keyword detection unit 123, a corresponding keyword candidate detection unit 124, a keyword coordinate comparison unit 125, and a corresponding keyword detection unit. 126 and a keyword correspondence output unit 127.

  The target document extraction unit 121 extracts document information that matches the analysis conditions specified by the user, and corresponds to the target document extraction unit 2 in FIG. The target document extraction unit 121 extracts all document information or a part of document information constituting the map 150.

  The keyword ranking generation unit 122 generates the keyword ranking table 5 and corresponds to the keyword ranking generation unit 4 in FIG. The reference keyword detection unit 123 detects a reference keyword, and corresponds to the reference keyword detection unit 6 in FIG. The corresponding keyword candidate detection unit 124 detects keyword candidates that can correspond to the reference keyword, and corresponds to the corresponding keyword candidate detection unit 7 in FIG. The keyword coordinate comparison unit 125 compares the coordinates in the reference keyword map 150 with the coordinates in the corresponding keyword candidate map 150, and corresponds to the feature information comparison unit 10 in FIG. The keyword coordinate calculation unit 115 corresponds to the feature information detection units 8 and 9 in FIG.

  The corresponding keyword detection unit 126 detects a keyword corresponding to the reference keyword, and corresponds to the corresponding keyword detection unit 11 in FIG. The keyword correspondence result output unit 127 outputs the relationship between the reference keyword and the corresponding keyword, and corresponds to the keyword correspondence relationship output unit 12 in FIG.

  FIG. 9 is a configuration explanatory diagram showing an outline of the hardware configuration of the document information analysis apparatus 100 and the client terminal 200. The literature information analysis apparatus 100 can be configured as a server computer or the like as described above.

  The document information analyzing apparatus 100 includes, for example, a communication interface (in the drawing, the interface is abbreviated as “I / F”) 1001, a CPU (Central Processing Unit) 1002, a ROM (Read Only Memory) 1003, and a RAM (Random Access). Memory) 1004 and an auxiliary storage device 1005 can be provided.

  In the auxiliary storage device 1005, for example, in addition to the OS (Operating System), the document database 31, the word database 32, the index database 33, the keyword correspondence analysis program 1100, the display control program 1110, and the principal component analysis A program 1120, a structural analysis program 1130, a search program 1140, and a web server program 1150 can be stored.

  As described above, the document database 31 stores document information such as patent publications in advance. In the word database 32, for example, words, synonyms, synonyms, and the like inappropriate for keywords such as particles and conjunctions are stored in advance. The keywords included in each document information can be extracted by using the document database 31 and the word database 32, respectively. The index database 33 manages the location of keywords for each piece of document information extracted in this way. Therefore, by using the index database 33, necessary document information can be quickly retrieved from the document database 31 storing a large number of document information.

  The keyword correspondence analysis program 1100 analyzes the correspondence between the keywords included in the map 150 and displays the analysis result on the map 150. The display control program 1110 is a program that performs drawing processing of the map 150 and the like. The principal component analysis program 1120 is a program for performing principal component analysis. The structure analysis program 1130 is a program for analyzing the structure of text data based on a technique such as text mining. The search program 1140 is a program that searches the literature database 31 and the like based on the input search conditions. The web server program 1150 is a program that realizes a web server function.

  The configuration of the client terminal 200 will be described. The client terminal 200 is connected to the document information analysis apparatus 100 via a communication network CN such as the Internet. For example, the client terminal 200 can be configured to include a communication interface 2001, a CPU 2002, a ROM 2003, a RAM 2004, and an auxiliary storage device 2005. is there.

  In addition, the above structure is an illustration and this invention is not limited to this. For example, it may be possible to realize at least a part of each function by using a hardware circuit such as a programmable logic device instead of the program.

  A document information analysis method using the document information analysis apparatus 100 (hereinafter also referred to as the analysis apparatus 100) will be described. FIG. 10 is a flowchart showing an overview of map control processing for generating and outputting the map 150.

  First, the user inputs search conditions via the user interface of the client terminal 200 (S30). The search condition can be performed, for example, by specifying a keyword such as “photocatalyst” or by specifying a patent classification code, a publication date of a document, or the like.

  When acquiring the search condition from the client terminal 200 (S31), the analysis apparatus 100 searches the index database 33 and the document database 31 based on the search condition, and extracts all the document information that matches the search condition (S32). More specifically, the index database 33 is used to extract document information having a keyword designated by the user. When document information having the specified keyword is specified, the document information is read from the document database 31. Then, the extracted result is transmitted from the analysis apparatus 100 to the client terminal 200 (S33) and displayed on the screen of the client terminal 200 (S34).

  The user confirms the total number of extracted documents, the document names, and the like via the screen of the client terminal 200, and gives approval to the extraction result (S35). If the user is not satisfied with the extraction result, the search condition can be changed and the search can be instructed again.

  When the analysis apparatus 100 confirms the user's approval, the analysis apparatus 100 extracts a keyword from each extracted document information (S36). The keyword extraction result is transmitted from the analysis apparatus 100 to the client terminal 200 (S37) and displayed on the screen of the client terminal 200 (S38). The user confirms whether or not the desired keyword is included in the keyword extraction result and gives approval (S39). If the user is not satisfied with the keyword extraction result, it is possible to request another keyword extraction.

  After confirming the user's approval, the analysis apparatus 100 calculates the coordinates of each extracted document information by performing principal component analysis (S40). Subsequently, the analysis apparatus 100 calculates the distribution density of the document information based on the coordinates of each document information (S41).

  The analysis apparatus 100 calculates the coordinates of each extracted keyword by performing principal component analysis (S42). Then, the analysis apparatus 100 generates a map 150 and transmits it to the client terminal 200 (S43). The user confirms the map 150 displayed on the screen of the client terminal 200 (S44).

  FIG. 11 is an explanatory diagram schematically showing how the map 150 is generated. The analyzer 100 calculates the coordinates of each document information and stores them in the document coordinate management table T1. In the document coordinate management table T1, for example, the document number for specifying document information is associated with the X-axis coordinate and the Y-axis coordinate of the document information.

  Further, the analysis apparatus 100 calculates the coordinates of each extracted keyword (main keyword) and stores it in the keyword coordinate management table T2. For example, the keyword coordinate management table T2 stores a keyword, a document number including the keyword, and the X-axis coordinate and the Y-axis coordinate of the keyword. In the figure, only one document number including a keyword is shown, but a list of all document numbers including the keyword can be included.

  As shown in the lower side of FIG. 11, for example, a plurality of block regions 130 can be set on the map 150 by dividing each of the X-axis method and the Y-axis direction of the map 150 by a plurality. The analysis apparatus 100 obtains the distribution density of the document information by calculating the number of document information located in each block region 130. A contour line 151 is set according to the calculated distribution density.

  FIG. 12 is a flowchart showing the keyword correspondence analysis process. This flowchart includes steps common to the flowchart shown in FIG. Therefore, a duplicate description will be omitted, and the steps characteristic of this embodiment will be mainly described.

  In this embodiment, since the function for analyzing the keyword correspondence is incorporated in the document information analyzing apparatus 100 that generates the map 150, the coordinates of the keyword are used as the keyword feature information (S15A, S17A). This is because keywords that mean the same technical element but differ in their expression form are considered to be arranged at relatively close positions on the map 150.

  The analysis apparatus 100 compares the coordinates of the reference keyword with the coordinates of the corresponding keyword candidate (S18A), selects the corresponding keyword candidate having a close coordinate, and generates the keyword correspondence table 12A (S19A). The analysis apparatus 100 displays the analysis result of the keyword correspondence on the map 150. When the user does not select the corresponding keyword candidate (S21: NO), the analysis apparatus 100 associates the keyword having the coordinates closest to the coordinates of the reference keyword with the reference keyword (S23A).

  According to the present embodiment, the same operational effects as those of the first embodiment described above can be obtained. In the present embodiment, as shown in FIG. 13, a plurality of keywords 152 are displayed on the map 150, and the reference keyword 152 </ b> A and the corresponding keyword 152 </ b> B are connected and displayed through a connection line 153. Therefore, the user can easily grasp on the map 150 that some terms are inconsistent and the standard expression of the terms in all or part of the document information group constituting the map 150. Can do.

  A third embodiment will be described with reference to FIG. In this embodiment, when the user does not select the keyword corresponding to the reference keyword from the corresponding keyword candidates, the appearance frequency rank (may be coordinates) as the feature information is calculated, and the ranks between the keywords are compared.

  First, the analysis apparatus 100 acquires the analysis condition 1A designated by the user (S51), and extracts the document information group that matches the analysis condition 1A from the document information group constituting the map 150 (S12). Then, the analysis apparatus 100 generates the keyword ranking table 5 based on the extracted document information group (S53), and generates the reference keyword table 6A and the corresponding keyword candidate table 7A based on the keyword ranking table 5, respectively. (S54, S55).

  Then, the analysis apparatus 100 generates and outputs the keyword correspondence table 12A (S56, S57), and waits for selection from the user (S58). When the user manually selects any one keyword from the corresponding keyword candidates (S58: YES), the selected keyword is associated with the reference keyword (S62).

  On the other hand, when the user does not select any one keyword from the corresponding keyword candidates (S58: NO), the analysis apparatus 100 calculates the average ranking of the reference keywords (S59) and ranks the corresponding keyword candidates. Is acquired (S60). Then, the analysis apparatus 100 selects a corresponding keyword candidate having a rank closest to the average rank of the reference keywords, and associates the selected corresponding keyword with the reference keyword (S61).

  Hereinafter, as described in FIG. 3, when the user desires to register a thesaurus (S63: YES), the analysis apparatus 100 registers the corresponding keyword as the reference keyword thesaurus (S64), and the user replaces the corresponding keyword. (S65: YES), the corresponding keyword is converted into a reference keyword (S66).

  Next, a fourth embodiment will be described with reference to FIGS. In this embodiment, a word group that is modified with the reference keyword and a word group that is modified with each corresponding keyword candidate are obtained, and the corresponding keyword candidate that most matches each of these modified keyword groups is used as the corresponding keyword. To detect.

  FIG. 15 is an explanatory diagram schematically showing the flow of data by the keyword correspondence analysis apparatus according to this embodiment. For the sake of space, the analysis conditions 1A and the information storage unit 3 shown in FIG. 2 are not shown in FIG.

  In the present embodiment, the reference keyword feature information detection unit 8 generates a ranking table 8A of words that are dependent on the reference keyword. For example, it is possible to generate a dependency word ranking table 8A by extracting words that are dependent on the reference keyword by performing morphological analysis or dependency analysis, and rearranging the extracted words in order of appearance frequency. it can. For example, when the reference keyword is “personal computer”, the words to be modified include “install”, “freeze”, “purchase”, and the like.

  In the same manner as described above, the corresponding keyword candidate feature information detection unit 9 generates, for each corresponding keyword candidate, a word ranking table 9A that is dependent on each corresponding keyword candidate. For example, when the corresponding keyword candidate is “information processing apparatus”, the words to be modified include “freeze”, “install”, “control”, and the like. In the case where the corresponding keyword candidate is “mouse”, the words to be modified include “drag”, “click”, “purchase”, and the like.

  Then, the corresponding keyword detection unit 11 compares the dependency word ranking table 8A of the reference keyword with the dependency word ranking table 9A of each corresponding keyword candidate. Thereby, the corresponding keyword detection unit 11 detects a plurality of corresponding keyword candidates in which the number of matching dependency words within a predetermined order is equal to or greater than a predetermined value. Alternatively, the corresponding keyword detection unit 11 can also detect only one corresponding keyword candidate whose dependency word in the predetermined order most closely matches.

  FIG. 16 is a flowchart showing the keyword correspondence analysis processing. Similar to FIG. 12, this flowchart includes steps common to the flowchart shown in FIG. Therefore, a duplicate description will be omitted, and the steps characteristic of this embodiment will be mainly described.

  In this embodiment, as described above, instead of directly comparing the attributes of the reference keyword and the corresponding keyword candidate, the reference keyword and the corresponding keyword candidate are determined based on the ranking of words that are dependent on these keywords. Guess the relationship.

  Therefore, in S15B, a word ranking table 8A that is dependent on the reference keyword is generated, and in S17B, a word ranking table 9A that is dependent on each corresponding keyword candidate is generated for each corresponding keyword candidate. In S18B, the dependency word ranking tables 8A and 9A are compared, and in subsequent S19B, corresponding keyword candidates whose dependency words match by a predetermined value or more are extracted. If the user does not select manual operation (S21: NO), one corresponding keyword candidate having the largest number of matching words (ie, matching rate) is selected (S23B).

  The present invention is not limited to the above-described embodiment. A person skilled in the art can make various additions and changes within the scope of the present invention. For example, those skilled in the art can appropriately combine the above embodiments.

  For example, in the above embodiment, the relationship between the reference keyword and the corresponding keyword is analyzed in units of applicants, but the present invention is not limited to this, and the relationship between keywords can be analyzed in units of documents. It is also possible to analyze between keywords on a period basis. For example, PHS (Personal Handyphone System) was previously called PHP (Personal Handy Phone), but even when the term changes with the times, it can be detected by the present invention. Furthermore, by calculating the keyword ranking based on the name of the inventor and the name of the applicant, even if the name of the inventor changes due to marriage, etc., the relationship with the maiden name is grasped and integrated into the current name. Can be managed. Similarly, even when the name of the applicant changes due to merger or division of companies, the correspondence between the old name and the latest name can be easily grasped.

  A plurality of types of algorithms for detecting the correspondence between keywords can be used in combination. For example, the correspondence detection accuracy can be improved by combining any of a method based on the keyword appearance rank, a method based on the keyword coordinates, and a method based on the matching rate of the words that are dependent on the keyword. It can also be increased.

It is explanatory drawing which shows the whole keyword corresponding relationship analyzer which concerns on embodiment of this invention. It is explanatory drawing which shows typically the flow of the data in a keyword corresponding relationship analysis process. It is a flowchart which shows a keyword correspondence analysis process. It is a flowchart which shows the keyword ranking table production | generation process in FIG. It is a flowchart which shows the reference | standard keyword table production | generation process in FIG. It is a flowchart which shows a corresponding keyword candidate table production | generation process in FIG. It is explanatory drawing which concerns on 2nd Example of this invention, and shows the whole literature information analysis apparatus provided with the keyword corresponding relationship analysis function. It is a functional block diagram of a literature information analysis device. It is explanatory drawing which shows the hardware and software structure outline | summary of a literature information analyzer. It is a flowchart which shows the process which produces | generates and displays a map. It is explanatory drawing which shows the relationship between a literature coordinate management table and a keyword coordinate management table, and a map. It is a flowchart which shows a keyword correspondence analysis process. It is explanatory drawing which shows a mode that the correspondence of a keyword is displayed on the map. It is a flowchart of the keyword corresponding relationship analysis process which concerns on 3rd Example of this invention. It is explanatory drawing which shows typically the flow of the data in the keyword corresponding relationship analysis process which concerns on 4th Example of this invention. It is a flowchart which shows a keyword correspondence analysis process.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Analysis condition setting part, 1A ... Analysis condition, 2 ... Target literature extraction part, 3 ... Information storage part, 3A ... Document information storage part, 3B ... Keyword storage part, 4 ... Keyword ranking production | generation part, 5 ... Keyword ranking Table 6: Reference keyword detection unit 6A ... Reference keyword table 7 ... Corresponding keyword candidate detection unit 7A ... Corresponding keyword candidate table 8 ... Reference keyword feature information detection unit 8A ... Words that depend on the reference keyword Ranking table, 9 ... corresponding keyword candidate feature information detection unit, 9A ... ranking table of words that depend on the corresponding keyword candidate, 10 ... feature information comparison unit, 11 ... corresponding keyword detection unit, 12 ... keyword correspondence output unit, 12A ... Keyword correspondence table, 13 ... display unit, 14 ... thesaurus registration unit, 15 Keyword substitution unit, 31 ... literature database, 32 ... word database, 33 ... index database, 100 ... literature information analyzer, 110 ... map control unit, 111 ... map generation condition input acceptance unit, 112 ... literature extraction unit, 113 ... literature Coordinate calculator, 114 ... keyword extractor, 115 ... keyword coordinate calculator, 116 ... map generator, 117 ... map display, 120 ... keyword correspondence analysis unit, 121 ... target document extractor, 122 ... keyword ranking generator , 123 ... reference keyword detection unit, 124 ... corresponding keyword candidate detection unit, 125 ... keyword coordinate comparison unit, 126 ... corresponding keyword detection unit, 127 ... keyword correspondence result output unit, 127 ... keyword correspondence output unit, 150 ... map, 151A, 151B, 151C ... contour lines, 52 ... keyword, 152A ... reference keyword, 152B ... corresponding keywords, 153 ... connection line, 200 ... client terminal

Claims (16)

A document information storage unit for storing a plurality of digitized document information;
By searching the document information storage unit based on the given analysis conditions, a target document extraction unit that extracts the document information to be analyzed,
For each keyword that appears in the document information based on the extracted document information, an appearance frequency analysis unit that generates keyword appearance frequency analysis information obtained by analyzing the application frequency;
A first keyword detecting unit that detects a first keyword used in predetermined document information among keywords included in the extracted document information based on the keyword appearance frequency analysis information;
Based on the keyword appearance frequency analysis information, a second keyword candidate that is a candidate for the second keyword corresponding to the first keyword is used as document information other than the predetermined document information in the extracted document information. A second keyword candidate detection unit for detecting from included keywords;
A second keyword detection unit for detecting the second keyword corresponding to the first keyword from the detected second keyword candidates;
Keyword correspondence analysis device with   The keyword correspondence analysis device according to claim 1, wherein the second keyword detection unit detects a second keyword candidate selected by a user from the second keyword candidates as the second keyword. A first feature information detector for detecting feature information of the detected first keyword;
A second feature information detection unit for detecting feature information of the detected second keyword candidate,
The second keyword detecting unit compares the feature information of the detected first keyword with the feature information of the detected second keyword candidate, thereby determining the second keyword from the second keyword candidates. The keyword correspondence analysis device according to claim 1 to detect. A first feature information detector for detecting feature information of the detected first keyword;
A second feature information detector for detecting feature information of the detected second keyword candidate;
A feature information comparison unit for comparing feature information of the first keyword and feature information of the second keyword;
A comparison result output unit that outputs a comparison result by the feature information comparison unit, and
The second keyword detection unit
A user designation mode for detecting, as the second keyword, a second keyword candidate selected by a user from the second keyword candidates based on the comparison result;
The automatic detection mode which detects the 2nd keyword candidate which has the feature information with the least difference with the feature information of the 1st keyword based on the comparison result as the 2nd keyword. Keyword correspondence analysis device.   The first keyword detection unit is based on the keyword appearance frequency analysis information, and is a second predetermined value or more and less than the total number of the extracted document information among keywords having appearance frequency ranks up to a first predetermined value. The keyword correspondence analysis apparatus according to claim 1, wherein a keyword used in the predetermined document information is detected as the first keyword.   Based on the keyword appearance frequency analysis information, the second keyword candidate detection unit has a third predetermined value or more in the predetermined document information among keywords included in the other document information other than the predetermined document information. The keyword correspondence analysis apparatus according to claim 1, wherein a keyword remaining after removing a used keyword is detected as the second keyword candidate. The first feature information detection unit detects an average appearance frequency rank in the predetermined document information of the first keyword as feature information of the first keyword,
The said 2nd feature information detection part detects the appearance frequency rank of the said 2nd keyword candidate in said other document information other than the said predetermined document information as the feature information of the said 2nd keyword. The keyword correspondence analysis device according to any one of the above. Coordinates of keywords included in the extracted document information by performing principal component analysis based on the total number of the extracted document information and the number of appearances of a plurality of predetermined keywords extracted from the extracted document information A keyword coordinate calculation unit for calculating
The first feature information detection unit detects the coordinates of the first keyword calculated by the keyword coordinate calculation unit as feature information of the first keyword,
The said 2nd feature information detection part detects the coordinate of the said 2nd keyword candidate calculated by the said keyword coordinate calculation part as the feature information of the said 2nd keyword candidate, either of Claim 3 or Claim 4 Keyword correspondence analysis device. The first feature information detection unit detects a ranking of words that are dependent on the first keyword in the predetermined document information as the first feature information,
The said 2nd feature information detection part detects the ranking of the word which makes a dependency with the said 2nd keyword candidate in document information other than the said predetermined document information as said 2nd feature information. The keyword correspondence analysis apparatus according to any one of the above. Document coordinates for calculating the coordinates of each document information of the extracted document information group by performing principal component analysis based on the combination and the number of appearances of a plurality of predetermined keywords extracted from the extracted document information A calculation unit;
A keyword coordinate calculation unit that calculates the coordinates of each of the predetermined keywords by performing principal component analysis based on the total number of document information including the predetermined keywords and the number of appearances of the predetermined keywords;
A distribution density of each document information is calculated based on the coordinates of each document information calculated by the document coordinate calculation unit, and a map figure having an outline based on the calculated distribution density and the predetermined keyword are visualized respectively. And a map generation unit for generating map information.
The keyword correspondence analysis device according to claim 1, wherein the first keyword and the second keyword are visualized in the map information.   The second keyword detection unit compares the coordinates of the first keyword and the coordinates of the second keyword candidate calculated by the keyword coordinate calculation unit, respectively, so that the second keyword candidate is selected from the second keyword candidates. The keyword correspondence analysis device according to claim 10, wherein   The first keyword and the second keyword are visualized in the map information in different display forms, and the map information includes a display element indicating a correspondence relationship between the first keyword and the second keyword. The keyword correspondence analysis device according to claim 10, which is included.   The keyword correspondence analysis device according to claim 1, further comprising a keyword replacement unit that replaces the second keyword with the first keyword.   The keyword correspondence analysis device according to claim 1, further comprising a relevance registration unit that stores the second keyword in association with the first keyword. Computer
Document information storage means for storing a plurality of digitized document information;
A target document extracting means for extracting the document information of the analysis target based on the given analysis condition;
Based on the extracted document information, for each keyword appearing in the document information, appearance frequency analysis means for generating keyword appearance frequency analysis information obtained by analyzing the application frequency;
First keyword detecting means for detecting a first keyword used in predetermined document information among keywords included in the extracted document information based on the keyword appearance frequency analysis information;
Based on the keyword appearance frequency analysis information, a second keyword candidate that is a candidate for the second keyword corresponding to the first keyword is used as document information other than the predetermined document information in the extracted document information. Second keyword candidate detection means for detecting from among the included keywords;
A program that functions as a second keyword detection unit that detects the second keyword corresponding to the first keyword from the detected second keyword candidates. Obtaining analysis conditions;
Extracting document information to be analyzed by searching a document information storage unit based on the acquired analysis conditions;
For each keyword appearing in the document information based on the extracted document information, generating keyword appearance frequency analysis information obtained by analyzing the application frequency;
Detecting a first keyword used in predetermined document information among keywords included in the extracted document information based on the keyword appearance frequency analysis information;
Based on the keyword appearance frequency analysis information, a second keyword candidate that is a candidate for the second keyword corresponding to the first keyword is used as document information other than the predetermined document information in the extracted document information. A step of detecting from included keywords;
Detecting the second keyword corresponding to the first keyword from the detected second keyword candidates;
Keyword correspondence analysis method including

Images