JP2017078976A - Product retrieval device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a product retrieval device capable of achieving visualization of quality information for each product and supporting its quality improvement.SOLUTION: In a product retrieval device 10, a user can view a relation between one keyword dependency pattern and another keyword dependency pattern in a co-occurrence information table for each product, and thus, can know what other kind of information (the co-occurrence information) should be searched for the keyword dependency pattern. At this time, a relation degree between the keyword dependency patterns can be grasped by a number representing their simultaneous generation ratio.SELECTED DRAWING: Figure 7A

Description

  The present invention relates to a product search apparatus, and more particularly to a product search apparatus that contributes to appropriate use of product-related information such as troubles in the product distribution market from document data.

  In recent years, tools for extracting a person and a keyword from document data and using information obtained by linking them are widely used. For example, a savvy person extraction device disclosed in Patent Document 1 (Japanese Patent Application Laid-Open No. 2002-56001) is a data in which a technical term is cut out from a document, and a degree of familiarity about a field to which the technical term belongs is stored in advance for each person. -Extracting an appropriate person from the base.

  However, although Patent Document 1 discloses a method for identifying a savvy person by associating a technical term with a person, it is necessary to extract a product and a keyword from document data and use information associated with them. Tools are not shown, and it is not possible to properly use information about troubles such as what kind of troubles occur in products in the market where the products are distributed and what causes the troubles.

  An object of the present invention is to provide a product search apparatus capable of “visualizing” quality information for each product and supporting quality improvement.

  A product search device according to a first aspect of the present invention includes a keyword extraction unit, a product information extraction unit, a product keyword extraction unit, and a co-occurrence information extraction unit. The keyword extracting unit extracts a keyword for each document data from a large number of accumulated document data including mail. The product information extraction unit extracts product specific information for each document data from the document data. The product keyword extracting unit creates and extracts association information between products and keywords for each document data. The co-occurrence information extracting unit creates and extracts co-occurrence information for each product by assigning a co-occurrence keyword to a keyword associated with the product for each document data.

  In this product search device, for example, when a searcher inputs "product name" and "symptom of failure", a related keyword and a co-occurrence keyword for the keyword are provided. be able to.

  A product search apparatus according to a second aspect of the present invention is the product search apparatus according to the first aspect, further comprising a keyword / dependency pattern extraction unit. The keyword / dependency pattern extraction unit creates and extracts a keyword / dependency pattern based on the extracted keyword.

  Even if keywords and co-occurrence keywords are provided, it is not easy to identify the cause of the problem when the number is very large. However, this product search device provides a problem by providing a keyword / dependency pattern. Factors can be easily identified.

  The product search device according to the third aspect of the present invention is the product search device according to the first aspect or the second aspect, and the co-occurrence information extraction unit assigns at least one co-occurrence keyword to the keyword.

  A product search device according to a fourth aspect of the present invention is the product search device according to any one of the first to third aspects, wherein the co-occurrence information extraction unit matrixes the keyword group in the co-occurrence information. Create a co-occurrence information table.

  In this product search apparatus, it is possible to list “ties” between one keyword in the co-occurrence information and another keyword.

  A product search device according to a fifth aspect of the present invention is the product search device according to the fourth aspect, wherein the co-occurrence information table is assigned to a keyword and a column assigned to a row at an intersection of the row and the column. Displays the rate of simultaneous occurrence of keywords.

  In this product search device, the “association degree (co-occurrence degree)” between keywords can be grasped by the number “the ratio of simultaneous occurrence (simultaneous occurrence ratio)” (that is, the characteristic value of “the degree of association”) The substitute characteristic value of “is the simultaneous occurrence rate”).

  A product search device according to a sixth aspect of the present invention is the product search device according to the fourth or fifth aspect, wherein the co-occurrence information extraction unit determines whether one keyword and another keyword in the co-occurrence information are present. The ratio of simultaneous occurrence is calculated, and the co-occurrence information with the ratio of simultaneous occurrence equal to or greater than a predetermined value is stored in the co-occurrence information table.

  In this product search device, whether or not to accumulate each keyword is determined based on the size of the concurrent occurrence with other co-occurrence keywords, thereby eliminating obsolete information and increasing storage capacity. Can be suppressed.

  A product search device according to a seventh aspect of the present invention is the product search device according to any one of the first to fifth aspects, wherein the co-occurrence information extraction unit includes a keyword in the co-occurrence information, The ratio of the simultaneous occurrence with other keywords is calculated, and the co-occurrence information with the simultaneous occurrence ratio being a predetermined value or more is displayed.

  In this product search device, narrowing down is facilitated by refraining from displaying keywords that occur at the same time but are not so frequently occurring.

  A product search device according to an eighth aspect of the present invention is the product search device according to any one of the first to sixth aspects, wherein the keyword extraction unit determines the number of keywords to be extracted for each document data. decide.

  In this product search apparatus, it is possible to prevent everything from becoming keywords. Further, if the number of keywords to be extracted is determined according to the size of the document data, the number of keywords proportional to the size of the document data can be secured.

  A product search device according to a ninth aspect of the present invention is the product search device according to any one of claims 2 to 6, wherein the keyword / dependency pattern extraction unit should extract each document data Determine the quantity of keywords and dependency patterns.

  In this product search device, if the number of keywords and dependency patterns to be extracted is determined according to the size of the document data, the number of keywords and dependency patterns that are proportional to the size of the document data is secured. You can also.

  A product search device according to a tenth aspect of the present invention is the product search device according to any one of the first to sixth aspects, wherein the keyword extraction unit extracts the number of keywords extracted from one document data. Is determined according to the size of the document data.

  In this product search apparatus, the number of keywords proportional to the size of document data can be secured.

  A product search device according to an eleventh aspect of the present invention is the product search device according to any one of the second to sixth aspects, wherein the keyword / dependency pattern extraction unit is extracted from one document data. The upper limit of the number of keywords and dependency patterns to be determined is determined according to the size of the document data.

  In this product retrieval apparatus, the number of keywords and dependency patterns proportional to the size of document data can be secured.

  A product search device according to a twelfth aspect of the present invention is the product search device according to any one of the first to eleventh aspects, further comprising a systematic information extraction unit. The systematized information extraction unit systematizes and extracts keywords or keywords / dependency patterns from the co-occurrence information for each product created by the co-occurrence information extraction unit.

  A product search device according to a thirteenth aspect of the present invention is the product search device according to any one of the first to twelfth aspects, further comprising a word map creation unit. The word map creation unit creates and extracts a hot keyword map from keywords associated with a product.

  In the product search device according to the first aspect or the third aspect of the present invention, for example, when the searcher inputs “product name” and “symptom of failure”, related keywords and co-occurrence keywords are provided. Therefore, it is possible to support the identification of the failure factor.

  In the product search device according to the second aspect of the present invention, the dependency factor can be easily specified by providing the dependency pattern.

  In the product search device according to the fourth aspect of the present invention, it is possible to list “ties” between one keyword in the co-occurrence information and another keyword.

  In the product search device according to the fifth aspect of the present invention, it is possible to grasp the “degree of association (degree of co-occurrence)” between keywords by the number “the rate of simultaneous occurrence (co-occurrence rate)” (that is, The substitute characteristic value of the characteristic value of “degree of association” is the “simultaneous occurrence ratio”).

  In the product search device according to the sixth aspect of the present invention, the determination as to whether or not to accumulate each keyword is made based on the size of the simultaneous occurrence ratio with other co-occurrence keywords, thereby making the information obsolete. And the increase in storage capacity can be suppressed.

  In the product search apparatus according to the seventh aspect of the present invention, narrowing down is facilitated by refraining from displaying keywords that occur at the same time but are not so frequently occurring.

  In the product search device according to the eighth aspect of the present invention, it is possible to prevent everything from becoming keywords. Further, if the number of keywords to be extracted is determined according to the size of the document data, the number of keywords proportional to the size of the document data can be secured.

  In the product search device according to the ninth aspect of the present invention, if the number of keywords and dependency patterns to be extracted is determined according to the size of the document data, the keywords and dependency proportional to the size of the document data are determined. It is also possible to secure the number of patterns.

  In the product search device according to the tenth aspect of the present invention, the number of keywords proportional to the size of the document data can be ensured.

  In the product search device according to the eleventh aspect of the present invention, the number of keywords and dependency patterns proportional to the size of document data can be ensured.

  In the product search device according to the twelfth aspect of the present invention, the co-occurrence information is grouped on the system side, thereby reducing the number of man-hours for grouping manually.

  In the product search apparatus according to the thirteenth aspect of the present invention, the hot keywords of the product can be shown by dividing the period.

The block diagram which shows the structure of the product search apparatus which concerns on one Embodiment of this invention. The flowchart showing the 1st operation | movement of a product search apparatus. A product keyword table for product ID “101” created by the product keyword extraction unit. A product keyword table for product ID “201” created by the product keyword extraction unit. The product keyword table about product ID "301" which the product keyword extraction part produced. The flowchart showing the 2nd operation | movement of a product search apparatus. Keyword / dependency pattern table related to one document data. Another keyword / dependency pattern table for document data. Keyword / dependency pattern table related to other document data. The flowchart showing the 3rd operation | movement of a product search apparatus. A keyword / dependency pattern co-occurrence information table created by the co-occurrence information extraction unit. A keyword / dependency pattern co-occurrence information table created by the co-occurrence information extraction unit. A keyword co-occurrence information table created by the co-occurrence information extraction unit. A co-occurrence information table of other keywords created by the co-occurrence information extraction unit. Keyword product map that the product search device displays on the screen of the search terminal. The block diagram which shows the structure of the product search apparatus which concerns on the modification of this invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. The following embodiments are specific examples of the present invention and do not limit the technical scope of the present invention.

(1) Configuration of Product Search Device 10 FIG. 1 is a block diagram showing a configuration of a product search device 10 according to an embodiment of the present invention. In FIG. 1, the product search device 10 includes a document data storage unit 20, a keyword extraction unit 30, a product information extraction unit 40, a product keyword extraction unit 50, a keyword / dependency pattern extraction unit 55, and a co-occurrence information extraction unit 60. I have.

(2) Detailed description (2-1) Document data storage unit 20
The document data storage unit 20 has a mechanism for storing document data (including mail) created for each product or document data (including mail) created by the person for each person as needed. At least a document name, a product name, and a creator name or creator ID are attached to the document.

(2-2) Keyword extraction unit 30
The keyword extraction unit 30 extracts characteristic words / phrases in document data, that is, keywords, for each document data. The number of words to be extracted can be specified for each document data, preventing all from becoming keywords. Of course, the number of keywords can be determined according to the size of the document data.

(2-3) Product information extraction unit 40
The product information extraction unit 40 extracts product information for each document data by using property information of document data, login information at the time of document registration, or from field in the case of mail.

(2-4) Product keyword extraction unit 50
For each document data, the product keyword extraction unit 50 creates a product keyword table in which products are associated with keywords appearing in the document data created for the products, and stores them.

  For example, as shown in FIG. 1, a table listing product IDs, document IDs, document names, creation dates, and keywords in order from the left is created. The product ID is an ID for specifying a product. The document ID is an ID for specifying what kind of document (for example, a quality report) the document data describing the product is. As an example, if the person is a quality report created in 2012, an ID of 2012 QR is given.

  Since the document name is usually determined according to the contents of the document, there is a high possibility that the most important keyword is included. The creation date / time is the creation date / time of the document, but may be either the creation date / time or the update date / time.

  Further, in the product keyword table of FIG. 1, the information related to the other document is also continued below the document “2012QR”. It is assumed that such a product keyword table is created and stored for each product.

(2-5) Keyword / Dependency Pattern Extraction Unit 55
The keyword / dependency pattern extraction unit 55 creates a “keyword / dependency pattern table” for each product and each document data based on the “product keyword table” created by the product keyword extraction unit 50.

  As a specific example, for the keyword “motor”, a plurality of dependency patterns “drive by” and “drive by” are obtained.

(2-6) Co-occurrence information extraction unit 60
The co-occurrence information extraction unit 60 creates a “co-occurrence information table” for each product based on the “product keyword table” created by the product keyword extraction unit 50.

  For example, as shown in FIG. 1, the co-occurrence information table displays a rate at which a keyword assigned to a row and a keyword assigned to a column simultaneously occur at the intersection of the row and the column. Thus, the “degree of association (degree of co-occurrence)” between keywords can be grasped by the number “the rate of simultaneous occurrence (simultaneous occurrence rate)”.

  As a result, not only can one or a plurality of products related to the keyword be searched, but also what information (co-occurrence information) each product is related to can be known.

(2-7) General control unit 100
The overall control unit 100 is a so-called CPU, and executes necessary processing according to the OS and other programs.

  When the user inputs a keyword from the dedicated terminal, the overall control unit 100 sends the keyword extraction unit 30, product information extraction unit 40, product keyword extraction unit 50, keyword / dependency pattern extraction unit 55, and co-occurrence information extraction unit 60. Send commands as needed to extract necessary information.

(2-8) Others The keyword extraction unit 30, the product information extraction unit 40, the product keyword extraction unit 50, the keyword / dependency pattern extraction unit 55, and the co-occurrence information extraction unit 60 each store a storage area for storing extracted data Are stored individually or in a common storage device.

(3) Operation of Product Search Device 10 Here, three operations will be described. The first operation is an operation when document data is input from each terminal to the product search device 10, and the second operation is an operation when a document name is input from each terminal to the product search device 10, The third operation is an operation when a keyword for search is input from the terminal to the product search apparatus 10.

(3-1) First Operation of Product Search Device 10 FIG. 2 is a flowchart showing a first operation of product search device 10. Here, the first operation is an operation until the product search apparatus 10 creates a “product keyword table” from the document data.

  In FIG. 2, the overall control unit 100 determines whether or not document data has been input from any of the terminals connected in step S <b> 1. If document data has been input, the overall control unit 100 proceeds to step S <b> 2. If not, continue the determination.

  Next, the overall control unit 100 stores the document data input in step S2 in the document data storage unit 20, and proceeds to step S3.

  Next, the overall control unit 100 extracts a keyword for each document data from the document data stored in the document data storage unit 20 via the keyword extraction unit 30 in step S3, and proceeds to step S4.

  Next, the overall control unit 100 extracts a product for each document data from the document data stored in the document data storage unit 20 via the product information extraction unit 40 in step S4, and proceeds to step S5.

  Next, the overall control unit 100 creates and stores a product keyword table from the keywords extracted in step S3 and the products extracted in step S4 via the product keyword extraction unit 50 in step S5. Hereinafter, the product keyword table will be described with reference to the drawings.

  FIG. 3A is a product keyword table for the room air conditioner with the product ID “101” created by the product keyword extraction unit 50. In FIG. 3A, it is found that there are 100 documents related to “room air conditioner” from the product keyword table created for all 190 quality reports registered for the air conditioner.

  In the keyword column, keywords with a high simultaneous occurrence rate are displayed from the left. However, in order to know whether or not it overlaps with a keyword extracted in another quality report, it is not intentionally left-aligned.

  For example, in the “2014r” document, the product name itself becomes a keyword, and the keywords “filter”, “cleaning function”, and “biting” co-occur under the keyword “room air conditioner”.

  FIG. 3B is a product keyword table for the commercial air conditioner with the product ID “201” created by the product keyword extraction unit 50. In FIG. 3B, it is found that there are 50 documents including “commercial air conditioner” from the product keyword table created for all 190 quality reports registered for the air conditioner.

  For example, in the “2014g” document, keywords “compressor”, “motor”, “rotor”, “magnet”, and the like co-occur under the keyword “commercial air conditioner”.

  FIG. 3C is a product keyword table for the personal air conditioner with the product ID “301” created by the product keyword extraction unit 50. In FIG. 3C, it is found that there are 40 documents including “personal air conditioner” from the product keyword table created for all 190 documents registered for the air conditioner.

  For example, in the “2013p” document, the keywords “refrigerant”, “leak”, “pinhole”, and the like co-occur under the keyword “personal air conditioner”.

(3-2) Second Operation of Product Search Device 10 For example, when a searcher who looks at the product keyword table in FIG. 3B “Business Air Conditioner” focuses on the co-occurrence keyword of specific document data, further investigation is performed. To assist, the product search apparatus 10 can create a keyword / dependency pattern table of the document data and display it on the screen. That is, the product search apparatus 10 can create a keyword / dependency pattern table for each document data.

  FIG. 4 is a flowchart showing the second operation of the product search apparatus 10. Here, the second operation is an operation in which the product search apparatus 10 creates a keyword / dependency pattern table from the “product keyword table”.

  In FIG. 4, the overall control unit 100 determines whether or not a specific document ID or document name has been input in step S11. If either of the document ID or document name has been input, the overall control unit 100 proceeds to step S12. When there is no input, the determination is continued.

  Next, in step S12, the overall control unit 100 extracts a keyword / dependency pattern from the document data of the input document ID or document name via the keyword / dependency pattern extraction unit 55, and proceeds to step S13.

  Next, the overall control unit 100 creates a keyword / dependency pattern table (see FIGS. 5A to 5C) in step S13, and proceeds to step S14.

  Then, in step S14, the overall control unit 100 displays “keyword / dependency pattern table” on the search screen. The “keyword / dependency pattern table” will be described below with reference to the drawings.

  FIG. 5A is a keyword / dependency pattern table related to one document data. In FIG. 5A, the keyword / dependency pattern table includes a product ID, document ID, document data creation date / time, keyword, keyword / dependency pattern.

  From FIG. 5A, it can be seen that the document data is a quality report of a commercial air conditioner from the document ID “2012QR” and the document name “2012g”. In addition, from the description of the keyword / dependency pattern display column, keywords such as “commercial air conditioner / does not operate”, “compressor / does not start”, “motor / does not drive”, and “coil / is shorted”. / It can be seen that dependency patterns are accumulated.

  Then, the searcher can infer from the keyword / dependency pattern an event that “the wire is short-circuited, the motor does not drive, and the compressor does not start”.

  FIG. 5B is a keyword / dependency pattern table related to another document data. In FIG. 5B, the keyword / dependency pattern table includes a product ID, document ID, document data creation date / time, keyword, keyword / dependency pattern.

  From FIG. 5B, it can be seen that the document data is a quality report of a commercial air conditioner from the document ID “2013QR” and the document name “2013g”. In addition, from the keyword / dependency pattern display column, keywords such as “commercial air conditioner / does not operate”, “compressor / does not start”, “motor / does not drive”, and “coil / is disconnected”. / It can be seen that dependency patterns are accumulated.

  The searcher can then infer from the keyword / dependency pattern that “the coil is disconnected, the motor does not drive, and the compressor does not start”.

  As described above, since the co-occurrence keyword “coil” is the same in the product keyword table of FIG. 3B alone, there is a possibility that the content is the same as the quality information. By extracting the dependency pattern, it is possible to grasp even more detailed events such as “the coil / is short-circuited” and “the coil / is disconnected”.

  FIG. 5C is a keyword / dependency pattern table relating to other document data. In FIG. 5C, it can be seen that the document data is a quality report of a commercial air conditioner from the document ID “2014QR” and the document name “2014g”. In addition, from the description of the keyword / dependency pattern display column, “commercial air conditioner / does not operate”, “compressor / does not start”, “motor / does not drive”, “rotor / does not rotate”, and “magnet” It can be seen that the keyword / dependency pattern of “/ is off” is accumulated.

  Then, the searcher can infer from the keyword / dependency pattern that “the magnet is removed, the motor does not drive, and the compressor does not start”.

  As described above, even if the keyword / dependency pattern of the same phenomenon “business-use air conditioner / does not operate”, the keyword / dependency pattern of the factor “magnet / disengaged” co-occurs depending on the document. In some cases, the keyword / dependency pattern of the factor “coil / short-circuited” may co-occur.

  Therefore, in this embodiment, instead of extracting keywords for all document data, keywords are extracted for each document data, associated with products, and co-occurrence keywords are assigned to create keywords and dependency patterns. doing.

(3-3) Third Operation of Product Search Device 10 Here, the searcher inputs the keyword “commercial air conditioner” from the dedicated terminal of the product search device 10 for the market complaint performance survey of “air conditioner”. This will be explained under the assumption.

  FIG. 6 is a flowchart showing a third operation of the product search apparatus 10. In FIG. 6, the overall control unit 100 determines whether or not a keyword has been input in step S21. When there is a keyword input, the overall control unit 100 proceeds to step S22, and when there is no keyword input, continues the determination.

  Next, the overall control unit 100 scans all data in the product keyword table in step S22, extracts the product keyword table that hits the input keyword, and proceeds to step S23.

  Next, in step S23, the overall control unit 100 extracts a keyword / dependency pattern table for the keywords on the product keyword table, and proceeds to step S24.

  Next, the overall control unit 100, based on the “product keyword table” extracted in the previous step S22 via the co-occurrence information extraction unit 60 in step S24, the “co-occurrence information table” (FIG. 7) of the commercial air conditioner. Reference) is created, and the process proceeds to step S25.

  Next, the overall control unit 100 displays the “co-occurrence information table” of the commercial air conditioner extracted in the previous step S24 via the co-occurrence information extraction unit 60 in step S25 on the screen of the search terminal. Hereinafter, the “co-occurrence information table” will be described with reference to the drawings.

(3-3-1) Detailed Description of Co-occurrence Information Table The co-occurrence information extraction unit 60 is based on the “product keyword table” described in FIG. 3A to FIG. Create a table. Here, a “co-occurrence information table” based on the “product keyword table” illustrated in FIG. 3B will be described as an example.

  The “product keyword table” in FIG. 3B is created by assigning a co-occurrence keyword to a keyword associated with a commercial air conditioner for each document data related to the commercial air conditioner.

  Further, the keyword / dependency pattern tables in FIGS. 5A, 5B, and 5C are created based on the keywords on the “product keyword table” in FIG. 3B.

  Then, the co-occurrence information extraction unit 60 creates a matrix-like co-occurrence information table from these keywords and dependency patterns, and at the intersections of the rows and columns, the keywords and dependency patterns and columns assigned to the rows are displayed. Displays the rate at which assigned keywords and dependency patterns occur simultaneously.

  For example, for a keyword / dependency pattern “compressor / does not start”, a co-occurrence keyword / dependency pattern “rotor / does not rotate” with a high degree of co-occurrence exists in one document data. In the co-occurrence information table focused on commercial air conditioners, the co-occurrence degree of the co-occurrence keyword / dependency pattern “rotor / does not rotate” may be low.

  FIG. 7A is a co-occurrence information table created by the co-occurrence information extraction unit 60. In FIG. 7A, in the leftmost column of the co-occurrence information table, “commercial air conditioner / does not operate”, “compressor / does not start”, “motor / does not drive”, “coil / is shorted”, “ “The coil / is disconnected”, “the rotor / does not rotate”, and “the magnet / is disconnected” are listed in order from top to bottom.

  Similarly, keywords and dependency patterns are listed in order from left to right in the top row of the co-occurrence information table.

  At the intersection between the row and the column, the ratio of the keyword / dependency pattern assigned to the row and the keyword / dependency pattern assigned to the column is displayed in%.

(3-3-1-1) First aspect represented by co-occurrence information table This co-occurrence information table represents two aspects. The first aspect is the co-occurrence degree (cooccurrence ratio) viewed from the keyword / dependency pattern, and the second aspect is the co-occurrence degree viewed from the co-occurrence keyword / dependency pattern. Hereinafter, it demonstrates concretely, referring drawings.

  First, “a co-occurrence keyword / dependency pattern occurrence ratio viewed from a keyword / dependency pattern”, which is a first aspect, will be described. In FIG. 7, the area where the background is shaded displays “co-occurrence keyword / dependency pattern simultaneous occurrence ratio viewed from keyword / dependency pattern”.

  A numerical value of 80 is displayed at the intersection of the row “commercial air conditioner / does not operate” and the column “compressor / does not start”. This is because there is a keyword / dependency pattern of “commercial air conditioner / does not work” in 100 documents, and “compressor / does not start” is “operating air conditioner / does work” in 80 documents. It means that it was appearing with "No". In this case, “compressor / does not start” versus “commercial air conditioner / does not operate” is a co-occurrence keyword / dependency pattern, and both appear in 80 out of 100 documents. Therefore, the rate of simultaneous occurrence is 80%.

  In addition, the keyword / dependency pattern of “Motor / not drive” appeared in 65 documents out of 100 documents in which the keyword / dependency pattern of “business air conditioner / does not operate” appeared. In this case, “Professional air conditioner / does not work” and “Motor / does not drive” are co-occurrence keywords and dependency patterns, and both appear in 65 out of 100 documents. The coincidence rate is 65%.

  In addition, the keyword / dependency pattern “coil / short-circuited” appeared in 50 documents out of 100 documents where the keyword / dependency pattern “business air conditioner / does not operate” appeared. , “Coil / Short-circuited” is a co-occurrence keyword / dependency pattern with respect to “Professional air conditioner / does not operate”, and both appear in 25 out of 100 documents. Therefore, the simultaneous occurrence rate is 25%.

  In addition, the keyword / dependency pattern “coil / disconnected” appears in 50 documents out of 100 documents in which the keyword / dependency pattern “business air conditioner / does not operate” has appeared. In the case of “commercial air conditioner / does not operate”, “coil / disconnected” is a co-occurrence keyword / dependency pattern, and both appear in 25 out of 100 documents. Therefore, the simultaneous occurrence rate is 25%.

  In addition, when the keyword / dependency pattern of “rotor / does not rotate” appears in 35 out of 100 documents in which the keyword / dependency pattern of “business air conditioner / does not operate” has appeared , “Rotor / does not rotate” against “commercial air conditioner / does not operate” is a co-occurrence keyword / dependency pattern, both appearing in 35 out of 100 documents at the same time The occurrence rate is 35%.

  In addition, the keyword / dependency pattern “magnet / disconnected” appears in 30 documents out of 100 documents where the keyword / dependency pattern “commercial air conditioner / does not operate” has appeared. In this case, “magnet / disengaged” is “co-occurrence keyword / dependency pattern” for “commercial air conditioner / does not operate”, and both appear in 30 out of 100 documents. Therefore, the rate of simultaneous occurrence is 30%.

(3-3-1-2) Second aspect represented by co-occurrence information table The following second aspect is “the ratio of simultaneous occurrence of other keywords / dependency patterns viewed from co-occurrence keywords / dependency patterns” Is explained.

  First, “compressor / does not start”, “motor / does not drive”, “coil / is shorted”, “coil / is disconnected”, “rotor / does not rotate” and “magnet / The numerical value of 100 is displayed at all the intersections between each row of “I am off” and the column of “Commercial air conditioner / not working”. Since this is a keyword / dependency pattern extracted from all 100 documents related to commercial air conditioners, the rate of simultaneous occurrence is inevitably 100%.

  Here, a description will be given by taking the column of “compressor / not start” as an example. If the keyword / dependency pattern “Motor / not drive” appears in 65 documents out of 80 documents where the keyword / dependency pattern “compressor / does not start” appears, “Compressor / does not start” versus “motor / does not drive” is a co-occurrence keyword / dependency pattern, and both of them appeared in 65 documents out of 80. 81%.

  In addition, the keyword / dependency pattern “coil / short-circuited” appears in 25 of the 80 documents in which the keyword / dependency pattern “compressor / does not start” has appeared. When “compressor / does not start”, “coil / shorted” is a co-occurrence keyword / dependency pattern, and both appear in 25 out of 80 documents. Therefore, the simultaneous occurrence rate is 31.5%.

  In addition, the keyword / dependency pattern “coil / short-circuited” appears in 25 of the 80 documents in which the keyword / dependency pattern “compressor / does not start” has appeared. When “compressor / does not start”, “coil / shorted” is a co-occurrence keyword / dependency pattern, and both appear in 25 out of 80 documents. Therefore, the simultaneous occurrence rate is 31.5%.

  In addition, when the keyword / dependency pattern “rotor / does not rotate” appears in 35 of the 80 documents in which the keyword / dependency pattern “compressor / does not start” appears , “Rotor / does not rotate” for “compressor / does not start” is a co-occurrence keyword / dependency pattern, both appearing together in 35 out of 80 documents. The percentage is 44%.

  In addition, the keyword / dependency pattern “magnet / disconnected” appeared in 30 documents out of 80 documents where the keyword / dependency pattern “compressor / does not start” appeared. In this case, “compressor / does not start” and “magnet / does not come off” are co-occurrence keywords and dependency patterns, and both of them appeared in 30 out of 80 documents at the same time. The generation rate is 38%.

  In FIG. 7A, “commercial air conditioner / does not operate” → “compressor / does not start” → “motor / does not drive” → “coil is shorted” → “coil is disconnected / broken. The dependency part is expressed as a verb such as “is”, but is not limited to this, and as shown in FIG. 7B, “commercial air conditioner / operation” → “compressor / startup” → “motor / drive” The dependency portion may be expressed as a noun such as “→ coil / short circuit” → “coil / disconnection”.

(3-3-2)
The co-occurrence information extracting unit 60 creates a matrix-like co-occurrence information table from keywords and dependency patterns, but of course, the present invention is not limited to this, and only the keywords of the “product keyword table” are shown in FIG. A matrix-like co-occurrence information table as shown may be created.

  However, as is clear from the comparison of the keyword “coil” in the co-occurrence information tables of FIGS. 7A, 7B, and 8, the co-occurrence information table using only the keywords has different dependency parts and only the keywords are the same. It should be noted that they will be integrated.

  In order to eliminate this problem, for example, by treating the keyword / dependency pattern “coil / short-circuited” as the keywords “coil” and “short-circuit”, the keyword “short-circuit” is displayed in the co-occurrence information table. The The co-occurrence information table shown in FIG.

  As shown in FIG. 9, the keywords and dependency patterns of the co-occurrence information table of FIG. 8 are “coil / short circuited” and “coil / disconnected” are “coil”, “short circuit”, “ It will be treated as a keyword “disconnection”.

  As a result, it is possible to prevent the “short circuit” and the “break” that co-occur on the keyword “coil” from being overlooked.

(3-3-3) Keyword Set Taking the quality reports “2012g” and “2013g” that appeared in the section of [(3-2) Second Operation of Product Search Device 10] as described above as an example, “Keyword Set” Will be described.

  For example, the quality report “2012g” states that “commercial air conditioner does not operate .... compressor does not start .... motor does not drive .... coil is shorted.” It is assumed that the report “2013g” states that “the commercial air conditioner does not operate .... the compressor does not start .... the motor does not drive .... the coil is disconnected.”

  According to the second operation of the product search device 10, the quality report “2012g” indicates that “commercial air conditioner / does not operate”, “compressor / does not start”, “motor / does not drive”, “coil / “Short-circuited” keyword / dependency pattern is extracted, and from the quality report “2013g”, “commercial air conditioner / does not operate”, “compressor / does not start”, “motor / does not drive”, “ The keyword / dependency pattern “coil / disconnected” is extracted.

  In the present embodiment, “commercial air conditioner / operation”, “compressor / startup”, “motor / drive”, “coil / short circuit”, and “coil / disconnection” are extracted as [keyword set].

  Further, if any of the quality reports “2012g” and “2013g” has a description that “the operation of the commercial air conditioner does not necessarily mean that the compressor is activated”, “the commercial air conditioner / The keywords and dependency patterns “operating” and “compressor / activate” are extracted, and “commercial air conditioner / operation / compressor / activation” becomes a set of keywords.

  In terms of the difficulty of establishment, the above keyword combinations are difficult to establish in the order of co-occurrence keyword group of the same document → keyword group of the same sentence → keyword / dependency pattern group.

  That is, when the searcher inputs the keyword “coil” for the commercial air conditioner at the initial stage of the search, it is known that there are 50 quality reports as shown in FIG. On the other hand, if you enter the keyword set “Coil / Short” for commercial air conditioners, you can narrow down to 25 quality reports from the beginning, as shown in FIG. 7B, without knowing the dependency pattern. You can get to the core of quality issues faster.

  Therefore, when searching by keyword, a search by “keyword pair” using the co-occurrence of sentences between the keyword and the dependency is more effective than a search by a single keyword having many co-occurrence sources. This is useful when the search target parameter is large.

(3-3-4) Description of Keyword Product Map FIG. 10 is a “keyword product map” displayed on the screen of the search terminal by the product search device 10. In FIG. 10, the keyword product map displays that product information related to the upper keyword “commercial air conditioner” exists for three types of type A, type B, and type C. At the same time, it can be understood that the co-occurrence keyword “compressor” frequently appears in all product information.

(4) Other operations The product search apparatus 10 is configured so that the user performs a predetermined operation from the search terminal, whereby a “product keyword table” for each product, a “co-occurrence information table” for each product, and a “keyword product map”. Can be displayed individually on the screen.

(5) Features (5-1)
The product search apparatus 10 can list “associations” between one keyword / dependency pattern and another keyword / dependency pattern in the co-occurrence information table for each product. It is possible to know whether to investigate with such information (co-occurrence information).

(5-2)
At that time, the “degree of connection” between the keywords and the dependency patterns can be grasped by the number of the “simultaneous occurrence ratio”.

(6) Modification (6-1)
The above embodiment has been described on the assumption that the keyword / dependency pattern and the co-occurrence keyword / dependency pattern are stored in the “product keyword table” or the “co-occurrence information table” regardless of the size of the simultaneous occurrence ratio. However, the present invention is not limited to this.

  For example, by not accumulating keywords and co-occurrence keywords that have a co-occurrence ratio less than a threshold value, it is possible to eliminate obsolete information and to suppress an increase in storage capacity.

(6-2)
In the above embodiment, it is assumed that the keyword / dependency pattern and the co-occurrence keyword / dependency pattern are displayed in the “product keyword table” or the “co-occurrence information table” regardless of the size of the simultaneous occurrence ratio. Although described, the present invention is not limited to this.

  For example, the product search apparatus 10 can refrain from displaying keyword / dependency patterns that have a co-occurrence rate that is too low to be co-occurrence. Thereby, narrowing down can be facilitated.

(6-3)
By grouping co-occurrence information on the system side, it is possible to reduce man-hours for grouping manually. Therefore, as shown in FIG. 11, systematic information extraction in which the product search apparatus 10 systematizes and extracts keywords or keyword / dependency patterns from the co-occurrence information for each product created by the co-occurrence information extraction unit 60. It is possible to further include a unit 71.

(6-4)
The hot keywords of the product may be indicated by dividing the period. Therefore, as shown in FIG. 11, it is possible to further include a word map creation unit 73 that creates and extracts a hot keyword map from keywords associated with a product.

(7) Others (7-1)
The creator may be designated from the outside. Examples of the external include “UI input”, “corresponding file between file and creator”, and “document specifying the owner of the file”.

(7-2)
The “product keyword table” and the “co-occurrence information table” are created in the CPU regardless of whether or not they are displayed on the screen, and are stored in a predetermined storage area for a short period or a long period.
(8) Other Embodiments In the above embodiments, in principle, keywords that appear simultaneously in the same document / one file are treated as “co-occurrence keywords”.

  However, some documents to be searched are briefly described in short sentences with an emphasis on communication such as support mail. In this case, it becomes a problem whether data for searching to form some output is obtained from one document data. This problem can be solved by the following method.

  For example, in the case of a support mail when making an inquiry to the support center, it is possible to adopt a method of collecting related events in the scene that occurs. As a specific example, the support email sent to the PC support center is categorized according to the occurrence scene, such as when saving, starting, and when power is restored, and a co-occurrence information table is created for the classified units. As a result, useful data is formed.

DESCRIPTION OF SYMBOLS 10 Product search apparatus 30 Keyword extraction part 40 Product information extraction part 50 Product keyword extraction part 55 Keyword and dependency pattern extraction part 60 Co-occurrence information extraction part 71 Systematization information extraction part 73 Word map creation part

JP 2002-560001 A

Claims (13)

A keyword extraction unit (30) for extracting a keyword for each document data from a large number of accumulated document data including mail;
A product information extraction unit (40) for extracting product specific information for each document data from the document data;
A product keyword extraction unit (50) for creating and extracting association information between the product and the keyword for each document data;
A co-occurrence information extracting unit (60) that creates a co-occurrence information for each product by adding a co-occurrence keyword to the keyword associated with the product for each document data; and
Comprising
Product search device (10). A keyword / dependency pattern extraction unit (55) for creating and extracting a dependency pattern based on the extracted keyword;
The product search device (10) according to claim 1. The co-occurrence information extraction unit (60) assigns at least one co-occurrence keyword to the keyword.
The product search device (10) according to claim 1 or 2. The co-occurrence information extraction unit (60) creates a co-occurrence information table that represents a keyword group in the co-occurrence information in a matrix.
The product search device (10) according to any one of claims 1 to 3. The co-occurrence information table displays the rate at which the keyword assigned to the row and the keyword assigned to the column occur simultaneously at the intersection of the row and the column.
The product search device (10) according to claim 4. The co-occurrence information extraction unit (60) calculates a rate at which one keyword and another keyword in the co-occurrence information are simultaneously generated, and calculates the co-occurrence information in which the rate at which the co-occurrence occurs is a predetermined value or more. Accumulate in the co-occurrence information table;
The product search device (10) according to claim 4 or 5. The co-occurrence information extraction unit (60) calculates a rate at which one keyword and another keyword in the co-occurrence information are simultaneously generated, and calculates the co-occurrence information in which the rate at which the co-occurrence occurs is a predetermined value or more. indicate,
The product search device (10) according to any one of claims 1 to 6. The keyword extraction unit (30) determines the quantity of the keywords to be extracted for each document data.
The product search device (10) according to any one of claims 1 to 6. The keyword / dependency pattern extraction unit (55) determines the quantity of the keyword / dependency pattern to be extracted for each document data.
The product search device (10) according to any one of claims 2 to 6. The keyword extraction unit (30) determines an upper limit of the quantity of the keywords extracted from one document data according to the size of the document data.
The product search device (10) according to any one of claims 1 to 6. The keyword / dependency pattern extraction unit (55) determines the upper limit of the quantity of the keyword / dependency pattern extracted from one document data according to the size of the document data.
The product search device (10) according to any one of claims 2 to 6. A systematic information extracting unit (71) for systematizing and extracting the keyword or the keyword / dependency pattern from the co-occurrence information for each product created by the co-occurrence information extracting unit (60);
The product search device (10) according to any one of claims 2 to 11. A word map creation unit (73) for creating and extracting a hot keyword map from the keywords associated with the product;
The product search device (10) according to any one of claims 1 to 12.

Images