Information retrieval method and information retrieval system

Abstract

To provide an information searching method and an information searching system capable of efficiently searching desired information among an enormous amount of information.SOLUTION: An information searching method is executed by an information searching system in which an information managing device executes a searching process based on a search condition containing a keyword input by a searcher, and one or greater than or equal to two search keywords extracted by a correlation analysis on the keyword, and searches information matching the search condition from a database. The information managing device receives a selection operation for causing the searcher to select, in the search condition, whether the search keyword is set to be an AND condition to the keyword or not, and whether the search keyword is set to be a NOT condition to the keyword, and then executes the searching process.SELECTED DRAWING: Figure 5

Description

Technical Field

[0001] The present invention relates to an information retrieval method and an information retrieval system that can efficiently retrieve information required by a searcher from a database storing a plurality of pieces of information.

Background Art

[0002] In order to retrieve desired information from a vast amount of information stored in a database, it is necessary to input appropriate keywords into an information retrieval system. However, in the case of a person with little knowledge and experience, it may be difficult to input appropriate keywords, resulting in a problem that the desired information cannot be retrieved. Against this background, for example, in Patent Document 1, an information retrieval system has been proposed that can retrieve desired information even if the input keyword is inappropriate by performing a search including synonyms of the input keyword.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the information retrieval system proposed in Patent Document 1, since the search is performed including synonyms, the number of search results becomes enormous depending on the input keyword. Therefore, even if the desired information is included in the search results, it takes a lot of time to identify the desired information by checking one by one from the enormous search results.

[0005] The present invention has been made in view of the above, and an object thereof is to provide an information retrieval method and an information retrieval system capable of efficiently retrieving desired information from a vast amount of information. [Means for solving the problem]

[0006] To solve the above-mentioned problems and achieve the objective, the information retrieval method according to the present invention is an information retrieval method in which an information management device executes a search process based on search conditions including a keyword entered by a searcher and one or more narrowing keywords extracted by correlation analysis of the keyword, and retrieves information from a database that matches the search conditions, wherein the information management device accepts a selection operation that causes the searcher to select whether or not to make the narrowing keyword an AND condition with respect to the keyword, or whether or not to make the narrowing keyword a NOT condition with respect to the keyword, and then executes the search process.

[0007] Furthermore, in the information retrieval method according to the present invention, if the information management device extracts a plurality of filtering keywords, including a first filtering keyword and a second filtering keyword, in that order of increasing correlation with the keyword, the device accepts the selection operation in the order of the first filtering keyword and the second filtering keyword, which have relatively high correlations with the keyword.

[0008] Furthermore, in the information retrieval method according to the present invention, when the information management device has selected whether or not to use the first filtering keyword as an AND condition or as a NOT condition, the searcher has selected whether or not to use the first filtering keyword as an AND condition, and the search device has selected the search process, and if the number of data matching the search conditions falls below a predetermined threshold, the selection operation for the second filtering keyword is not accepted.

[0009] Furthermore, in the information retrieval method according to the present invention, the correlation analysis is performed based on classification information arbitrarily set from a plurality of classifications that show the relationship between the keyword and the narrowing keyword.

[0010] To solve the above-mentioned problems and achieve the objective, the information retrieval system according to the present invention comprises a database and an information management device that performs a search process based on search conditions including keywords entered by a searcher and one or more refining keywords extracted by correlation analysis of the keywords, and retrieves information from the database that matches the search conditions, wherein the information management device accepts a selection operation that causes the searcher to select whether or not to make the refining keywords an AND condition with respect to the keywords, or whether or not to make the refining keywords a NOT condition with respect to the keywords, and then performs the search process. [Effects of the Invention]

[0011] According to the present invention, desired information can be efficiently retrieved from a vast amount of information. [Brief explanation of the drawing]

[0012] [Figure 1] Figure 1 is a block diagram showing the configuration of an information retrieval system according to an embodiment. [Figure 2] Figure 2 shows an example of criteria for folder and file names. [Figure 3A] Figure 3A shows an example of definition information stored in the classification definition information storage unit. [Figure 3B] Figure 3B shows an example of a system for creating (defining) words belonging to a category. [Figure 4A] Figure 4A shows an example of keywords used for determining the type of information, stored in the dictionary for determining the type of information. [Figure 4B] Figure 4B shows an example of keywords stored in a keyword dictionary. [Figure 5] Figure 5 is a flowchart showing the flow of an information retrieval process, which is one embodiment of the present invention. [Figure 6A] Figure 6A shows an example of a first screen displayed on a display device in the information retrieval method according to the embodiment. [Figure 6B] FIG. 6B is a diagram showing an example of a second screen displayed on a display device in the information search method according to the embodiment. [Figure 6C] FIG. 6C is a diagram showing an example of a third screen displayed on a display device in the information search method according to the embodiment. [Figure 6D] FIG. 6D is a diagram showing an example of the degree of correlation between a search keyword and each word (refinement keyword) belonging to a category selected by a searcher in the information search method according to the embodiment. [Figure 6E] FIG. 6E is a diagram showing an example of a fourth screen displayed on a display device in the information search method according to the embodiment. [Figure 6F] FIG. 6F is a diagram showing an example of a fifth screen displayed on a display device in the information search method according to the embodiment. [Figure 6G] FIG. 6G is a diagram showing an example of a sixth screen displayed on a display device in the information search method according to the embodiment. [Figure 6H] FIG. 6H is a diagram showing an example of a seventh screen displayed on a display device in the information search method according to the embodiment. [Figure 6I] FIG. 6I is a diagram showing an example of an eighth screen displayed on a display device in the information search method according to the embodiment. [Figure 7A] FIG. 7A is a diagram showing an example of a ninth screen displayed on a display device in the information search method according to the embodiment. [Figure 7B] FIG. 7B is a diagram showing an example of a tenth screen displayed on a display device in the information search method according to the embodiment. [Figure 7C] FIG. 7C is a diagram showing an example of an eleventh screen displayed on a display device in the information search method according to the embodiment.

Embodiments for Carrying Out the Invention

[0013] The following describes an information retrieval system and information retrieval method according to embodiments of the present invention. In the following description, equipment maintenance information in the manufacturing industry will be described as an example of the information to be retrieved. In the following description, when simply referred to as "information" or "data," it will refer to "equipment maintenance information."

[0014] [Information Retrieval System] The configuration of the information retrieval system according to the embodiment will be described with reference to Figures 1 to 4B. As shown in Figure 1, the information retrieval system according to the embodiment comprises an information management device 10 for managing information, a database 12 for storing multiple pieces of equipment maintenance work information to be searched, multiple file servers 14, and a searcher terminal 20.

[0015] The information management device 10, the database 12, the multiple file servers 14, and the user terminal 20 are all connected via a network 18. The network 18 is a communication network managed by the organization (company) to which the user belongs, such as a LAN (Local Area Network). However, the network 18 may be a public communication network such as the Internet, or a communication network that partially uses public lines such as a WAN (Wide Area Network) or VAN (Virtual Private Network).

[0016] The information management device 10 and the user terminal 20 are each implemented by a computer and a program. Examples of user terminals 20 include mobile devices such as smartphones, personal computers, and tablet computers. This user terminal 20 is connected to or integrated with an input device 22 such as a keyboard or mouse and a display device 24 such as a display.

[0017] Database 12 aggregates and stores documents such as reports and manuals, drawings, and photographs stored in each file server 14. The information stored in database 12 includes information such as equipment failures and malfunctions, their causes, and countermeasures, as well as work standards and manuals. In this embodiment, file servers 14 are installed in each region, assuming that a particular piece of equipment exists in multiple regions (e.g., multiple countries, multiple provinces).

[0018] The information management device 10 executes a search process based on search conditions that include the search keyword entered by the searcher and one or more refinement keywords extracted through correlation analysis of the search keyword, and retrieves information from the database 12 that matches the search conditions.

[0019] The information management device 10 includes an information registration unit 110, an information retrieval unit 120, and various storage units (keyword dictionary 50, classification definition information storage unit 52, search information storage unit 54, and information type determination dictionary 56). In this embodiment, the registration function and the search function are divided between the information registration unit 110 and the information retrieval unit 120, but they may also be a single configuration.

[0020] The information registration unit 110 is configured to perform the registration function described above, more specifically, the process of storing the information stored in each file server 14 in the database 12. The information registration unit 110 also includes an information linkage unit 112, an information processing unit 114, and a storage processing unit 116.

[0021] The information sharing unit 112 automatically performs periodic or irregular linking between the information stored in the database 12 and the information stored in each file server 14. For example, when new data not stored in the database 12 is stored in the file server 14, the information sharing unit 112 retrieves the new data via the network 18.

[0022] The information processing unit 114 performs predetermined data processing on the data acquired by the information linkage unit 112. As part of this data processing, the information processing unit 114 adds information type information to the data. Information type information indicates which information type the document belongs to.

[0023] The information processing unit 114 identifies the type of information in the data, for example, by the folder name where the data was stored, the file name of the data, etc. Figure 2 shows an example of the criteria for defining the folder names available on each file server 14 and the file names of the data stored in each folder.

[0024] When identifying information types by folder, the information processing unit 114, for example, refers to Figure 2 and identifies the information type of data stored in the "Failure Report" folder as "Failure Cases." The information processing unit 114 also identifies the information type of data stored in the "Work Standards," "Creation Procedures," and "Safety Documents" folders as "Manuals," and the information type of data stored in the "Maintenance Know-how" folder as "Maintenance Know-how." Furthermore, the information processing unit 114 identifies the information type of data stored in the "Drawings" folder as "Drawings," the information type of data stored in the "Maintenance Logbook" folder as "Maintenance Logbook," and the information type of data stored in the "Operational Standards" folder as "Operational Standards."

[0025] When the information processing unit 114 identifies the type of information by its file name, for example, the file name is created by the data creator based on the file creation criteria shown in Figure 2. Based on this file name, the unit identifies that the data belongs to a "failure case".

[0026] The information processing unit 114 further processes the data by assigning classification information to the data based on the classification information stored in the classification definition information storage unit 52. Figure 3A shows an example of classification information stored in the classification definition information storage unit 52. This classification information defines which of several classifications a particular word belongs to, based on the criteria shown in Figure 3B.

[0027] Figure 3A shows, for example, that the word "abc" belongs to the "state" category and is assigned the category information "D51". Therefore, if the acquired data contains the word "abc", the category information "D51" assigned to the word "abc" will be attached to that data. Returning to Figure 1, we will continue the explanation of the remaining components.

[0028] The storage processing unit 116 stores the data processed by the information processing unit 114 in the database 12. In doing so, the storage processing unit 16 stores the data in the database 12, separating it by information type (see Figure 1).

[0029] The information stored in database 12 is pre-divided into multiple groups A, B, C, etc., according to the type of information. For example, in database 12 shown in Figure 1, group A of the first information type stores information about "failure cases," group B of the second information type stores information about "manuals," and group C of the third information type stores information about "drawings." Although not shown in the figure, group D of the fourth information type, group E of the fifth information type, and group F of the sixth information type store information about "maintenance know-how," "maintenance logbook," and "operational standards," respectively.

[0030] The information retrieval unit 120 searches and extracts specific equipment maintenance information from the database 12 based on search instructions from the user terminal 20. The information retrieval unit 120 includes a question reception unit 121, a language processing unit 122, a search candidate keyword reading unit 123, an information type determination unit 124, a classification information reception unit 125, a correlation calculation unit 126, a narrowing keyword presentation unit 127, and a search execution unit 130.

[0031] The question reception unit 121 receives questions entered by the searcher via the input device 22 of the searcher terminal 20, through the network 18. The format of the question input is not particularly limited; for example, one or more words may be entered, or a sentence in natural language form may be entered. From the searcher terminal 20, for example, a sentence such as "What should be done when the BB device fails in the AA facility?" or a question consisting of multiple words separated by spaces, such as "AA device BB device".

[0032] The language processing unit 122 extracts words (independent words) from the questions received by the question receiving unit 121 by performing known language processing such as morphological analysis.

[0033] Specifically, the language processing unit 122 extracts from the input question the input search keywords to be used later for searching, and the information type determination keywords to be used to determine the type of information that is the search intent (purpose). For example, if the question sentence "What should be done when the BB device fails in the AA equipment?" is input, the language processing unit 122 extracts "AA equipment" and "BB device" as input search keywords.

[0034] Furthermore, the language processing unit 122 considers the context of the question received by the question receiving unit 121, determines whether it is affirmative or negative, and extracts the input search keyword. For example, if the above question contains the word "burnt," it determines whether it is "burnt" or "not burnt" and extracts it as the input search keyword.

[0035] Furthermore, the language processing unit 122 refers to the information type determination dictionary 56 (see Figure 4A) and extracts words that match words registered in the information type determination dictionary 56 as information type determination keywords. For example, if the question sentence "What should be done when the BB device fails in the AA equipment?" is entered, the language processing unit 122 extracts the terms "failure" and "dealing with" as information type determination keywords.

[0036] Next, before explaining the search candidate keyword reading unit 123, we will describe the keyword dictionary 50. This keyword dictionary 50 contains related keywords that are pre-registered and stored, with their correspondences to one another.

[0037] The keyword dictionary 50 contains keywords that are synonymous with each other, for example, and is registered with each other. It also contains keywords that are similar in meaning, and is registered with each other. However, the keyword dictionary 50 may contain only synonymous keywords, or only similar keywords.

[0038] Figure 4B shows an example of keywords pre-registered in the keyword dictionary 50. In this figure, keywords belonging to the same row (horizontal column) are related to each other and are associated with one another.

[0039] The search candidate keyword reading unit 123 refers to the keyword dictionary 50 each time the language processing unit 122 extracts an input search keyword and reads out keywords related to the input search keyword as search candidate keywords.

[0040] For example, as described above, if the input search keyword is "AA equipment", the search candidate keyword reading unit 123 performs a match search on the keyword dictionary 50 shown in Figure 4B. It then reads out all other keywords belonging to the row containing "AA equipment" (the first row in the figure), or the specified keywords if specified (for example, only synonyms), as search candidate keywords.

[0041] In this example, "○○ equipment," "A'A' equipment," and "A"A"" are retrieved as synonyms for "AA equipment," and "aaa" and "a'a'a'" are retrieved as related terms for "AA equipment." Similarly, for "BB equipment," the second row of Figure 4B is referenced to retrieve related keywords as search candidate keywords.

[0042] Next, before describing the information type determination unit 124, the search information storage unit 54 will be explained. The search information storage unit 54 stores the input search keywords and information type determination keywords extracted by the language processing unit 122, and the search candidate keywords read out by the search candidate keyword reading unit 123, as search keywords. The search information storage unit 54 stores these search keywords cumulatively each time, for example, an additional question is input and additional search keywords are extracted (read out).

[0043] The information type determination unit 124 determines the type of information that indicates the searcher's search intent based on the information type determination keywords extracted from the question. Specifically, the information type determination unit 124 refers to the information type determination dictionary 56 shown in Figure 4A and determines the type of information to be searched based on the information type determination keywords.

[0044] The information type determination dictionary 56 is pre-registered with information type determination keywords that are expected to be extracted from the question, and the corresponding information types. In the information type determination dictionary 56 in Figure 4A, the information type "failure case" is pre-registered and associated with the information type determination keywords "failure," "similar examples," and "solution." In the same information type determination dictionary 56, the information type "manual" is pre-registered and associated with the information type determination keywords "type of failure" and "replacement procedure." Although not shown in the figure, the information types "maintenance know-how," "drawings," "maintenance ledger," and "operational standards" shown in Figure 2 also have information type determination keywords registered.

[0045] The classification information receiving unit 125 displays the classification to which the filtering keywords belong, making it selectable. For example, in the above example, the search would be conducted using the search keywords "AA equipment" and "BB device" (including synonyms and related terms) for information belonging to group A "failure cases," but the search results could be enormous. Therefore, the classification information receiving unit 125 (and the correlation calculation unit 126 and filtering keyword display unit described later) assist the searcher in adding or removing filtering keywords in order to narrow down the search results.

[0046] The classification information receiving unit 125 displays multiple classifications for selection. Specifically, the classification information receiving unit 125 displays multiple classifications, including "line," "equipment," "device," "part," "status," and "cause," for example, by a pull-down menu (see Figure 6C below).

[0047] The correlation calculation unit 126 calculates the correlation between the search keyword and each word belonging to the category selected by the searcher in the category information receiving unit 125. The correlation calculation unit 126 can calculate the correlation based on the frequency with which the search keyword and each word belonging to the category appear together in the same report or manual, for example. The correlation calculation unit 126 can also calculate the correlation based on the number of times the search keyword and each word belonging to the category are combined as a search keyword stored in the search information storage unit 54.

[0048] If there are multiple search keywords, the correlation will be calculated for some or all of them, depending on the searcher's selection. Specifically, the "search keywords" mentioned above refer to the input search keywords extracted by the language processing unit 122 and the search candidate keywords read by the search candidate keyword reading unit 123.

[0049] The filtering keyword presentation unit 127 presents to the searcher whether or not to add each word belonging to the category selected by the searcher in the category information receiving unit 125 (hereinafter referred to as "filtering keyword") to the search conditions. In other words, the filtering keyword presentation unit 127 accepts a selection operation that allows the searcher to choose whether or not to make the filtering keyword an AND condition with respect to the search keyword, or whether or not to make the filtering keyword a NOT condition with respect to the search keyword in the search conditions of the database 12 (see Figures 6E to 6G below). Specifically, this selection operation includes three patterns: (1) making the filtering keyword an AND condition with respect to the search keyword, (2) making it a NOT condition, and (3) not making it an AND condition or a NOT condition (skipping).

[0050] Note that the above filtering keywords are words that distinguish between affirmative and negative forms. For example, the filtering keyword for "burnt" would be either "burnt" or "not burnt."

[0051] Here, through processing in the correlation calculation unit 126, multiple filtering keywords may be extracted, including a first filtering keyword and a second filtering keyword, in that order of increasing correlation with the search keyword (see Figure 6D below). For example, in the example in Figure 6D, the first filtering keyword is "abc" and the second filtering keyword is "def".

[0052] In this case, the filtering keyword presentation unit 127 accepts a selection operation regarding whether to use the first filtering keyword and the second filtering keyword, which have a relatively high correlation with the search keyword, as AND conditions or NOT conditions with respect to the search keyword (see Figures 6E and 6F below). Specifically, this selection operation includes three patterns: (1) using the first and second filtering keywords as AND conditions with respect to the search keyword, (2) using NOT conditions, and (3) not using either AND or NOT conditions (skipping).

[0053] Furthermore, the filtering keyword presentation unit 127 executes a preliminary search process (preliminary search process) when the searcher selects whether to use the first filtering keyword as an AND condition or a NOT condition. If the number of data matching the search conditions falls below a predetermined threshold, the filtering keyword presentation unit 127 does not accept a selection operation for the second filtering keyword. In other words, if the search results become sufficiently small as a result of the preliminary search by adding or removing the first filtering keyword, the filtering keyword presentation unit 127 does not display a screen (Figure 6F described later) that prompts the user to select whether to add the second filtering keyword to the search conditions.

[0054] The search execution unit 130 performs a text-based matching search on the information stored in the database 12, comparing the search keywords stored in the search information storage unit 54 with the filtering keywords selected by the searcher. The search execution unit 130 also performs the same matching search on the information within the group corresponding to the information type determined by the information type determination unit 124, from among multiple groups A, B, C, etc., in the database 12.

[0055] [Information retrieval methods] The information retrieval method performed by the information retrieval system according to the embodiment will be described with reference to Figures 5 to 7C. Figure 5 is a flowchart showing the flow of the information retrieval method according to the embodiment. Figures 6A to 7C are screens displayed on the display device 24 of the searcher terminal 20, and are examples of application screens for using the information retrieval system according to the embodiment. The information retrieval method according to the embodiment is started, for example, when the searcher launches the above application on the searcher terminal 20.

[0056] First, the question receiving unit 121 determines whether or not it has received a question (step S101). For example, when a user launches the application, the display device 24 displays a screen as shown in Figure 6A. Then, when the user enters a question into the search space 200 via the input device 22 and selects (for example clicks) the send icon 201, the question receiving unit 121 accepts the question.

[0057] In this example, we will explain using the case where the question reception unit 121 receives the question, "What should be done if the BB device fails in the AA equipment?". A history display area 300 is provided at the bottom of the screen, and all or part of the questions previously entered into the search space 200 can be duplicated.

[0058] In step S101, if the question receiving unit 121 receives a question (Yes in step S101), the process proceeds to step S102; otherwise, if the question is not received (No in step S101), the process returns to step S101.

[0059] Next, the language processing unit 122 extracts words (independent words) from the question text received by the question receiving unit 121 by performing natural language processing such as morphological analysis (step S102). In this example, the language processing unit 122 extracts "AA equipment," "BB equipment," "failure," and "solution" from the question text, for example, "What should be done when the BB device fails in the AA equipment?"

[0060] Next, the language processing unit 122 extracts (selects) input search keywords and information type determination keywords from the words extracted by language processing (step S103). In this example, the language processing unit 122 extracts, for example, "AA equipment" and "BB device" as input search keywords, and extracts "failure" and "solution" as information type determination keywords.

[0061] Next, the search candidate keyword reading unit 123 refers to the keyword dictionary 50 (see Figure 4B) and reads out keywords related to the input search keyword as search candidate keywords (step S104). In this example, the search candidate keyword reading unit 123 reads out synonyms and related terms to which the input search keyword "AA equipment" extracted in step S103 belongs, and synonyms and related terms to which "BB device" belongs, as search candidate keywords.

[0062] Next, the search information storage unit 54 stores the input search keyword and the search candidate keyword read out by the search candidate keyword reading unit 123 (step S105). In this example, the search information storage unit 54 stores "AA equipment" and "BB device," as well as their synonyms and related terms.

[0063] Next, the information type determination unit 124 determines the information type (search intent) based on the information type determination keywords extracted in step S103 and the information type determination dictionary 56 (see Figure 4A) (step S106). In this example, the information type determination unit 124 determines that "failure cases" associated with "failure" and "solution" are the information type.

[0064] Next, the information retrieval unit 120 displays the search period, information type, search keywords, and classification information on the display device 24 via the searcher terminal 20, for example, as shown in Figure 6B (step S107). Specifically, the "search keywords" mentioned above refer to the input search keywords extracted by the language processing unit 122 and the search candidate keywords read by the search candidate keyword reading unit 123.

[0065] In the screen shown in Figure 6B, the information type, search keywords, and category information are displayed as selectable options, allowing the searcher to change them as needed. Furthermore, it is possible to set an arbitrary search period; if no period is specified, the search will cover the entire period. Search keywords can also be excluded from the search by unchecking the checkboxes next to them.

[0066] In the search keyword display area, a word request icon 203 is available for selection, allowing users to request the addition or deletion of other keywords similar to the currently displayed search keyword. Furthermore, while "Not Specified" is initially selected for the category information, manipulating this area using the input device 22 displays a pull-down menu, as shown in Figure 6C, allowing the user to select the desired category. This example explains the case where, for example, the information type and input search keyword remain unchanged, but the category information is changed from "Not Specified" to "Status."

[0067] In step S107, if the classification information is changed from "not specified" to another classification, and the classification information receiving unit 125 accepts a classification other than "not specified" (Yes in step S108), the process proceeds to step S109. On the other hand, if the classification information is "not specified" (No in step S108), the process proceeds to step S112.

[0068] Next, the correlation calculation unit 126 calculates the correlation between the search keyword selected by the searcher in step S107 and each word belonging to the category received by the category information receiving unit 125 (step S109). In this example, the correlation calculation unit 126 calculates the correlation between both the keywords "AA equipment" and "BB device" and each word (abc, def, ...) belonging to the category "state," as shown in Figure 6D, for example.

[0069] Next, the refinement keyword presentation unit 127 displays the refinement keyword with the Nth highest correlation (N=1 for the first time) with the search keyword on the display device 24 via the searcher's terminal 20, along with the search condition setting candidates (step S110). The above-mentioned "refinement keyword" refers to each word (abc, def, ...) belonging to the category "state," as shown in Figure 6D.

[0070] Next, the refinement keyword presentation unit 127 performs a preliminary search based on the search conditions selected by the searcher (step S111). Subsequently, the refinement keyword presentation unit 127 determines whether or not predetermined conditions are met (step S112). If predetermined conditions are met, the system proceeds to step S113; otherwise, it returns to step S110 and sets "N=N+1". The above "predetermined conditions" are the conditions for whether or not to present refinement keywords to the searcher. Examples of predetermined conditions include when there are no correlated refinement keywords, or when the number of search hits during the preliminary search is tens of results (e.g., 50 results) or less.

[0071] The following describes the details of the processing in steps S110 to S112. The filtering keyword presentation unit 127, as shown in Figure 6E for example, presents the filtering keyword "abc", which has the highest correlation with the search keyword, as an option to choose whether to set it as an exclusion condition (NOT condition) or an addition condition (AND condition). The screen in the same figure also displays the current number of search hits when a preliminary search is performed using only the search keyword, as well as icons for "Exclude Keyword Setting", "Additional Keyword Setting", and "Skip". The "Skip" icon is a button to select when you do not want to exclude or add any filtering keywords. In this example, we will explain using the case where the filtering keyword "abc" is set as an additional keyword.

[0072] Next, the refinement keyword presentation unit 127 performs a preliminary search based on the search keyword and the additional keyword "abc". Then, as shown in Figure 6F, for example, the refinement keyword presentation unit 127 presents the option to choose whether to set "def", the second-highest-correlated refinement keyword with the search keyword, as an exclusion condition (NOT condition) or an additional condition (AND condition). In this example, we will explain the case where the refinement keyword "def" is set as an exclusion keyword.

[0073] Next, the refinement keyword presentation unit 127 performs a preliminary search based on the search keyword, the additional keyword "abc", and the exclusion keyword "def". Then, as shown in Figure 6G, for example, the refinement keyword presentation unit 127 presents the option to select whether to set "ghi", the refinement keyword with the third highest correlation to the search keyword, as an exclusion condition (NOT condition) or an additional condition (AND condition). In this example, we will explain the case where the refinement keyword "ghi" is set as an exclusion keyword. The refinement keyword presentation unit 127 repeatedly presents refinement keywords as shown in Figures 6E to 6G until the above predetermined conditions are met.

[0074] Thus, in the information retrieval system according to this embodiment, the searcher can easily and visually narrow down the information stored in the database 12 by suggesting the addition or removal of filtering keywords until the number of search results falls below a predetermined number.

[0075] Next, the keyword refinement display unit 127 displays the information type, search keyword, classification information, and refinement keywords (additional keywords, exclusion keywords) on the display device 24 via the searcher's terminal 20, for example as shown in Figure 6H (step S113). At this stage as well, additional keywords and exclusion keywords can be excluded from the search by unchecking the checkboxes in front of them.

[0076] Next, the search execution unit 130 executes a search process on the information stored in the database 12 when the search icon 202 (see Figure 6H) is selected on the screen displayed on the display device 24 (step S114). Specifically, the search execution unit 130 executes a text-based matching search process on the information belonging to the group corresponding to the selected information type in the database 12, based on the input search keyword, search candidate keyword, and filtering keyword.

[0077] Next, the search execution unit 130 displays the search results on the user's display device 24, for example, as shown in Figure 6I (step S115), and completes the process.

[0078] In Figures 6E to 6G, the searcher was presented with options to exclude or add refinement keywords in order of their correlation with the search keyword, but the method of presentation is not limited to this. For example, as shown in Figures 7A to 7C, multiple refinement keywords, their correlation with the search keyword, and checkboxes for adding or excluding these refinement keywords may be presented.

[0079] In this case, when a user adds or removes any of the filtering keywords, the filtering keyword display unit 127 performs a preliminary search with those conditions and displays the current number of search hits on the screen. For example, in the example in Figure 7A, when the filtering keyword "abc" is added, the number of search hits is "xxxxx," but as shown in Figure 7B, by further removing the filtering keyword "def," the number of search hits decreases to "xxxxx." Then, as shown in Figure 7C, by further removing the filtering keyword "ghi," the number of search hits decreases to "xxx."

[0080] Furthermore, the "Add to Keywords" icon on each screen in Figures 7A to 7C is used to register the word selected with the checkbox as an additional keyword or an excluded keyword. For example, as in Figure 7A, if the "Add to Keywords" icon is pressed while "abc" is selected with the add checkbox, "abc" will be officially registered as an additional keyword, and the screen will transition to Figure 6H. Similarly, as in Figures 7B and 7C, if the "Add to Keywords" icon is pressed while "def" and "ghi" are selected with the exclude checkboxes, "def" and "ghi" will be officially registered as excluded keywords, and the screen will transition to Figure 6H.

[0081] Furthermore, the "Skip" icon in each screen from Figures 7A to 7C is used to skip the selection of each word. For example, if the "Skip" icon is pressed when "abc" can be added or removed (the line containing "abc" is highlighted), as in Figure 7A, the addition or removal of "abc" is skipped. Then, the line containing "def" below it is highlighted, and the screen transitions to a state where "def" can be added or removed. Similarly, if the "Skip" icon is pressed when "def" can be added or removed (the line containing "def" is highlighted), as in Figure 7B, the addition or removal of "def" is skipped. Then, the line containing "ghi" below it is highlighted, and the screen transitions to a state where "ghi" can be added or removed. Similarly, if the "Skip" icon is pressed when "ghi" can be added or removed (the line containing "ghi" is highlighted), as in Figure 7C, the addition or removal of "ghi" is skipped. The screen then transitions to Figure 6H.

[0082] Thus, the information retrieval system according to this embodiment presents the number of search results when filtering keywords are added or removed, and suggests to the searcher the addition or removal of filtering keywords, thereby making it easy and visually clear to filter the information stored in the database 12.

[0083] In the information retrieval method and information retrieval system according to the embodiments described above, input search keywords are extracted from the question entered by the searcher, and the searcher is further given the option to choose whether or not to use the refinement keywords extracted by correlation analysis as an AND condition or a NOT condition. This makes it possible to efficiently search for desired information from the vast amount of information stored in the database 12. As a result, for example, when a malfunction or failure occurs, the searcher can input the situation as text, efficiently search the database 12 for reference information such as past cases, and based on the retrieved information, the searcher can quickly perform equipment repair work.

[0084] Furthermore, the information retrieval method and information retrieval system according to the embodiment can efficiently narrow down the number of search results by excluding highly correlated filtering keywords. Therefore, even in cases where there have been many similar cases in the past, the desired information can be accurately found.

[0085] The information retrieval method and information retrieval system according to the present invention have been specifically described above with reference to embodiments and examples for carrying out the invention. However, the spirit of the present invention is not limited to these descriptions and must be interpreted broadly based on the claims. Furthermore, it goes without saying that various modifications and alterations based on these descriptions are also included in the spirit of the present invention.

[0086] For example, the above embodiment described an example of searching for equipment maintenance information in the manufacturing industry, but the present invention is not limited thereto. The information retrieval method and information retrieval system according to the present invention can be applied to all kinds of Q&A (Question & Answer) information, such as information indicating countermeasures and disease names according to the symptoms of a disease, or information indicating countermeasures according to the error names and conditions displayed on machinery such as cars.

[0087] Furthermore, in the above embodiment, the correlation with the search keyword is calculated only once (see step S109 in Figure 5), and the filtering keywords with the 1st to 3rd highest correlations are sequentially questioned (see Figures 6E to 6G). However, the correlation may be calculated multiple times. In this case, the filtering keyword presentation unit 127 calculates the correlation of multiple filtering keywords with respect to the search keyword (see Figure 6D) and indicates via the display device 24 whether or not to add or remove the filtering keyword with the highest correlation.

[0088] When a searcher selects to add or remove a filtering keyword, the filtering keyword presentation unit 127 recalculates the correlation under those conditions and indicates via the display device 24 whether or not to add or remove the filtering keyword with the highest correlation. The filtering keyword presentation unit 127 then repeats the calculation of correlation and the presentation of filtering keywords until the above predetermined conditions are met. By calculating the correlation each time a filtering keyword is presented in this way, search accuracy can be improved. [Explanation of Symbols]

[0089] 10 Information management device 12 Databases 14 File Servers 18 Network 20. Searcher's terminal 22 Input devices 24 Display device 50 Keyword Dictionary 52 Classification definition information storage section 54 Search Information Storage Unit 56 Dictionary for determining information type 110 Information Registration Department 112 Information Sharing Department 114 Information Processing Section 116 Storage Processing Unit 120 Information Retrieval Department 121 Question Reception Department 122 Language Processing Unit 123 Search candidate keyword reading section 124 Information type judgment unit 125 Classification Information Reception Department 126 Correlation Calculation Unit 127 Keyword display section for filtering 130 Search Execution Unit

Claims

1. An information retrieval method comprising an information management device that performs a search process based on search conditions including keywords entered by a searcher and one or more filtering keywords extracted by correlation analysis of those keywords, and retrieves information from a database that matches the search conditions, The aforementioned information management device is The search process is executed after accepting a selection operation from the searcher to choose whether or not to use the filtering keyword as an AND condition for the keyword, or whether or not to use the filtering keyword as a NOT condition for the keyword. Prior to executing the aforementioned search process, If multiple filtering keywords are extracted, including a first filtering keyword and a second filtering keyword, in that order of increasing correlation with the keyword, the selection operation will be accepted in the order of the first filtering keyword and the second filtering keyword, which have relatively high correlations with the keyword. When accepting the selection operation, the number of hits in the provisional search at that time is displayed in the order of the first filtering keyword and the second filtering keyword. Information retrieval methods.

2. The aforementioned information management device is When the searcher selects whether or not to use the first filtering keyword as an AND condition, or whether or not to use the first filtering keyword as a NOT condition, the search process is executed. The information retrieval method according to claim 1, wherein if the number of data matching the search conditions falls below a predetermined threshold, the selection operation for the second filtering keyword is not accepted.

3. The information retrieval method according to claim 1, wherein the correlation analysis is performed based on classification information arbitrarily set from a plurality of classifications indicating the relationship between the keyword and the narrowing keyword.

4. An information retrieval system comprising: a database; an information management device that performs a search process based on search conditions including keywords entered by a searcher and one or more filtering keywords extracted by correlation analysis of the keywords, and retrieves information from the database that matches the search conditions, The aforementioned information management device is The search process is executed after accepting a selection operation from the searcher to choose whether or not to use the filtering keyword as an AND condition for the keyword, or whether or not to use the filtering keyword as a NOT condition for the keyword. Prior to executing the aforementioned search process, If multiple filtering keywords are extracted, including a first filtering keyword and a second filtering keyword, in that order of increasing correlation with the keyword, the selection operation will be accepted in the order of the first filtering keyword and the second filtering keyword, which have relatively high correlations with the keyword. When accepting the selection operation, the number of hits in the provisional search at that time is displayed in the order of the first filtering keyword and the second filtering keyword. Information retrieval system.

Images