Information retrieval system

Abstract

An information retrieval system includes a database (12), an inquiry article receiving section (26) that receives an inquiry article, an input search keyword extracting section (44) that extracts an input search keyword from the inquiry article, a common keyword dictionary (30) that registers associated keywords in correspondence with each other, a local keyword dictionary (102) that registers a local keyword used in a specific region, a search candidate keyword reading section (32) that reads a keyword associated with the input search keyword, and a search executing section (40) that performs a search process on the database using the input search keyword and the keyword associated with the input search keyword. The search candidate keyword reading section reads the keyword associated with the input search keyword with reference to the common keyword dictionary in a case where the input search keyword is not registered in the local keyword dictionary.

Description

Technical Field

[0001] This invention relates to information retrieval systems. Background Technology

[0002] Previously, maintenance data in manufacturing and other industries was recorded on paper. In maintenance information retrieval systems, it is crucial to digitize, aggregate, and database this paper data. Advances in retrieval technology have improved the accuracy of search results, thus driving the digitization and database creation of paper data.

[0003] For example, in manufacturing, malfunctions, defects, causes, and solutions for equipment are constantly being databased. When the same malfunction or defect occurs, past examples can be referenced to prevent further problems and shorten the time to recovery. Additionally, the database also stores operating standards, manuals, and other information. Therefore, the amount of information stored in the database becomes enormous.

[0004] Conventional methods for obtaining information from such massive amounts of data involve text retrieval processing where a given string is used as a keyword, and content matching that keyword is output as the search result. For example, Patent Document 1 discloses an information retrieval system that allows even those with limited knowledge and experience, or those who struggle to input suitable keywords, to obtain good search results.

[0005] Patent Document 1: Japanese Patent Application Publication No. 2019-121392

[0006] In situations where factories and organizations located in multiple regions use a common information retrieval system, there are often terms that are only used in specific regions. If these regionally specific terms are also managed by a common dictionary, the dictionary size increases, raising concerns about the inability to perform high-speed searches. Summary of the Invention

[0007] The present invention was made in view of the above-mentioned problems, and its object is to provide an information retrieval system that can perform high-speed retrieval even when it is used in multiple regions.

[0008] The information retrieval system of the present invention comprises:

[0009] A database stores multiple pieces of information that can be retrieved via text.

[0010] The inquiry article handling department accepts inquiry articles in natural language format.

[0011] The input search keyword extraction unit extracts the input search keywords from the aforementioned query articles;

[0012] A common keyword dictionary that registers related keywords in a corresponding manner;

[0013] A local keyword dictionary that registers regional keywords used in specific regions;

[0014] The candidate keyword retrieval unit, referring to at least one of the aforementioned common keyword dictionary and the aforementioned local keyword dictionary, reads keywords associated with the input search keyword; and

[0015] The retrieval execution unit uses the input search keywords and keywords associated with them to perform retrieval processing on the database.

[0016] If the input search keyword is not registered in the local keyword dictionary, the search candidate keyword reading unit refers to the common keyword dictionary to read keywords associated with the input search keyword.

[0017] According to the present invention, an information retrieval system is provided that enables high-speed retrieval even when used in multiple regions. Attached Figure Description

[0018] Figure 1 This is a block diagram illustrating a simplified structure of an information retrieval system according to one embodiment of the present invention.

[0019] Figure 2 This is a diagram representing an example of a keyword stored in a common keyword dictionary.

[0020] Figure 3 This is a diagram representing an example of keywords used to determine the type of information stored in a dictionary for information type determination.

[0021] Figure 4 This is a flowchart used to explain the operation of the information management device of the information retrieval system implemented in this way.

[0022] Figure 5 This is a diagram illustrating the hierarchical structure of a database.

[0023] Figure 6 This is a diagram illustrating the search results.

[0024] Figure 7 This is a flowchart used to explain the information retrieval operation of the information retrieval system in the implementation method.

[0025] Figure 8 It is a functional block diagram of a computer. Detailed Implementation

[0026] In this embodiment, the following information retrieval system will be described as an example. This system pre-databases equipment maintenance operation information in the manufacturing industry as text-searchable information. When a malfunction or defect occurs, the user inputs the situation in the form of an article, thereby retrieving information from the database regarding past cases that can be used as references. The user can then perform recovery operations based on the retrieved information. Here, equipment maintenance operation information includes, for example, operating standards and manuals that record malfunctions, defects, causes, and solutions for equipment, etc. Furthermore, in the following, instances related to malfunctions or defects will be referred to as "malfunction cases."

[0027] The retrieval applications of the information retrieval system of the present invention are not limited to those described above. The information retrieval system of the present invention can also be used for purposes other than equipment maintenance, such as assisting with equipment fault recovery operations. The information retrieval system of the present invention can be applied to situations such as: in the sales field, retrieving past examples in response to customer inquiries and suggesting the best answer. The information retrieval system of the present invention can be applied to situations such as: in the field of product design, retrieving similar manufacturing achievements, technical papers, and other literature in response to customer inquiries and proposing the best product design.

[0028] Figure 1 This is a block diagram illustrating a simplified structure of an information retrieval system according to one embodiment of the present invention. The information retrieval system is configured to include: an information retrieval device 10 that performs retrieval processing; a database 12 that stores multiple pieces of information capable of text retrieval; an input device 14 such as a keyboard and mouse connected to the information retrieval device 10; and a display device 16 such as a display connected to the information retrieval device 10. Furthermore, the information retrieval system is configured to include a searcher-side terminal 20, a local management device 100, and an information management device 110 connected to the information retrieval device 10 via a network 18.

[0029] Network 18 is a communication network such as a LAN (Local Area Network). However, it is not limited to this; in addition to public communication networks such as the Internet, network 18 can also be a communication network that locally incorporates public lines, such as a WAN (Wide Area Network) or a VPN (Virtual Private Network). The information retrieval device 10, the retrieval-side terminal 20, the local management device 100, and the information management device 110 are implemented by a computer and a program, respectively. Examples of the retrieval-side terminal 20 include a personal computer (PC), a tablet computer, etc., which are connected to or integrated with an input device 22 and a display device 24. Examples of the local management device 100 include a personal computer, etc., which has a local database 104. The local management device 100 may also have an input device and a display device. Examples of the information management device 110 include a personal computer, etc., which has a temporary storage unit 112. The information management device 110 may also have an input device and a display device.

[0030] Here, Figure 1 The structure of the information retrieval system is just one example; it may also exclude some of the constituent elements. Furthermore, the information retrieval system may also have other constituent elements. For example, the information retrieval system may also lack the input device 14 and display device 16, and instead use the user-side terminal 20 for information input and output. Additionally, the information retrieval system may have multiple local management devices 100 located in each of multiple regions. Because the structure is the same, in... Figure 1 Only one local management device 100 is shown in the figure, but in this embodiment, the information retrieval system has multiple local management devices 100. In addition, the multiple local management devices 100 are physically configured in each of the multiple regions that commonly use the information retrieval system.

[0031] Database 12 stores articles, illustrations, photographs, etc. Articles can be generated not only from initial text input, but also from audio transcribed using a voice recognition engine, or from printed text transcribed using an OCR (Optical Character Reader). When illustrations or photographs are stored, the system can retrieve information from characters embedded in the illustrations or from filenames. The information stored in database 12 includes malfunctions, defects, and their causes in devices, as well as operational standards and manuals for troubleshooting. In this embodiment, database 12 includes multiple malfunction examples.

[0032] Database 12 periodically adds data to local databases 104 located in multiple regions to expand the retrieved information. As detailed later, the data stored in local databases 104 in each region is converted into a file format suitable for storage in database 12 by information management device 110 and stored as storage data in temporary storage unit 112. Information management device 110 outputs the storage data stored in temporary storage unit 112 when database 12 is periodically updated with new information. In other words, data stored in local databases 104 in each region is imported into database 12 through file format conversion performed by information management device 110.

[0033] The information retrieval device 10 is configured to include: a query article receiving unit 26, a natural language processing unit 28, a common keyword dictionary 30, a local keyword dictionary 102, a search candidate keyword reading unit 32, a search keyword storage unit 34, an information type determination unit 36, an information type determination dictionary 38, a retrieval execution unit 40, and a search candidate keyword change request receiving unit 42. The natural language processing unit 28 includes an input search keyword extraction unit 44 and an information type determination keyword extraction unit 46. Here, Figure 1 The structure of the information retrieval device 10 is just one example; it may also exclude some of the constituent elements. Furthermore, the information retrieval device 10 may also have other constituent elements. For example, the information retrieval device 10 may also be structured without a common keyword dictionary 30, a local keyword dictionary 102, and an information category determination dictionary 38, but instead accesses these dictionaries located externally to the information retrieval device 10 via a network 18.

[0034] These functional modules briefly have the following functions. The query article receiving unit 26 receives queries in natural language form input by the queryer via the input device 22 of the queryer-side terminal 20 through the network 18. For example, the queryer-side terminal 20 can input natural language queries such as "A fault has occurred in device ○○. Have there been similar cases before?", "What is the type of fault △△?", "What is the replacement sequence for component □□?", and "What is the reason for the inverter tripping?".

[0035] The Natural Language Processing (NLP) unit 28 performs natural language processing, such as known morpheme analysis, on the query articles received by the query article receiving unit 26, and extracts words. The input search keyword extraction unit 44 of the NLP unit 28 extracts input search keywords to be used in the search from the input query article. To perform this extraction, the input search keyword extraction unit 44 can, for example, refer to the common keyword dictionary 30 and the local keyword dictionary 102, which will be described in detail later, and extract words registered in the common keyword dictionary 30 or the local keyword dictionary 102 as input search keywords. For example, when a query article such as "A malfunction of a ×× device occurs in ○○ device. Are there similar cases in the past?" is input, the input search keyword extraction unit 44 extracts words such as "○○ device" and "×× device" as input search keywords. For example, when a query article such as "The type of malfunction is △△ malfunction?" is input, the input search keyword extraction unit 44 extracts words such as "△△ malfunction" as input search keywords. Input search keyword extraction unit 44, for example, extracts the word "□□ component" as input search keywords when the input is a query such as "What is the replacement order of □□ component?". Input search keyword extraction unit 44, for example, extracts the words "inverter" and "tripping" as input search keywords when the input is a query such as "What is the reason for the inverter tripping?".

[0036] The information type determination keyword extraction unit 46 of the natural language processing unit 28 extracts information type determination keywords from the input query text to determine the information type as the search intent. To perform this extraction, the information type determination keyword extraction unit 46 can refer to the information type determination dictionary 38, which will be described in detail later, and extract words registered in the information type determination dictionary 38 as information type determination keywords. For example, when the query text "A malfunction of the ×× device occurs in the ○○ device. Are there similar cases in the past?" is input, the information type determination keyword extraction unit 46 extracts words such as "malfunction" and "similar cases" as information type determination keywords. For example, when the query text "The type of malfunction is △△ malfunction?" is input, the information type determination keyword extraction unit 46 extracts words such as "type of malfunction" as information type determination keywords. For example, when the query text "What is the replacement order of the □□ component?" is input, the information type determination keyword extraction unit 46 extracts words such as "replacement order" as information type determination keywords. Information type judgment using keyword extraction unit 46: For example, when an input is a query such as "What is the reason for the inverter tripping?", the word "reason" is extracted as the input search keyword.

[0037] Related keywords are pre-registered and stored in a common keyword dictionary 30, corresponding to each other. In this embodiment, synonymous keywords are registered in the common keyword dictionary 30. Additionally, similar keywords are registered in the common keyword dictionary 30. However, the common keyword dictionary 30 may only register synonymous keywords or only similar keywords. Alternatively, or based on this, keywords considered highly related may be pre-registered in a corresponding manner.

[0038] Figure 2 This represents an example of a keyword pre-registered in the common keyword dictionary 30. Keywords belonging to the same row are related to each other and correspond to each other.

[0039] The candidate keyword reading unit 32, for example, when extracting the input search keyword from the input search keyword extraction unit 44, refers to the common keyword dictionary 30 and reads keywords associated with the input search keyword as candidate keywords. For example, when the input search keyword is "○○ device", the candidate keyword reading unit 32 reads "AA device", "A′A′ device", and "A"A" as synonyms from the common keyword dictionary 30, and reads "aaa" and "a′a′a′" as related terms. In other words, the candidate keyword reading unit 32 reads "AA device", "A′A′ device", "A"A", "aaa", and "a′a′a′" as candidate keywords. Here, the candidate keyword reading unit 32 may also read only a portion, rather than all, of the synonyms and related terms from the common keyword dictionary 30 as candidate keywords. As an example, the candidate keyword reading unit 32 may also read only synonyms.

[0040] The local keyword dictionary 102 registers keywords used in specific regions, i.e., regional keywords. Regional keywords can also be registered in the local keyword dictionary 102 by the administrator. The local keyword dictionary 102 can also be structured similarly to the common keyword dictionary 30, where keywords belonging to the same row are related and correspond to each other. The local keyword dictionary 102 can also register synonymous keywords corresponding to each other, similar to the common keyword dictionary 30. Furthermore, the local keyword dictionary 102 can also register keywords with similar meanings corresponding to the common keyword dictionary 30. The local keyword dictionary 102 registers regional keywords, which are preferentially referenced by the search candidate keyword reading unit 32, thus eliminating the need to manage all terms in the common keyword dictionary 30. Therefore, it avoids the problem of the common keyword dictionary 30 becoming too large and reducing search speed. There can also be multiple local keyword dictionaries 102, each registering regional keywords for a different region. Alternatively, the local keyword dictionary 102 can also be configured to differentiate each registered regional keyword in which region it is used.

[0041] In this embodiment, when the input search keyword is not registered in the local keyword dictionary 102, the candidate keyword reading unit 32 refers to the common keyword dictionary 30 and outputs keywords associated with the input search keyword. As described above, regional keywords are keywords used in specific regions. Regional keywords may be words used only in factories or units in a limited region, or words that exist as common words but are commonly used in a limited region. For example, if there are multiple local keyword dictionaries 102, the regional keyword for region X is registered in the local keyword dictionary 102 used in region X. For example, if there is only one local keyword dictionary 102, the regional keyword for region X is registered in the local keyword dictionary 102 after being associated with region X. When the input search keyword is registered in the local keyword dictionary 102, the candidate keyword reading unit 32 reads the associated keywords from the local keyword dictionary 102 as candidate search keywords. When the input search keyword is not registered in the local keyword dictionary 102, the candidate keyword reading unit 32 refers to the common keyword dictionary 30. The candidate keyword reading unit 32 can, for example, read "EE device" as a synonym and "X′X′ device" as a classifier from the local keyword dictionary 102 when the input search keyword is "XX device" as a regional keyword, and use them as search candidate keywords.

[0042] Here, the local keyword dictionary 102 can also be used to match regional keywords with keywords registered in the common keyword dictionary 30. In this case, the search candidate keyword reading unit 32 can also, when the input search keyword is a regional keyword, refer to both the common keyword dictionary 30 and the local keyword dictionary 102 to read keywords associated with the input search keyword. For example, in the local keyword dictionary 102, "XX device" as a regional keyword can be matched with "○○ device" registered as a synonym in the common keyword dictionary 30. The search candidate keyword reading unit 32 can, for example, when the input search keyword is "XX device", refer to the local keyword dictionary 102 to read "○○ device", and further refer to the common keyword dictionary 30 to read "AA device", "A′A′ device", "A"A", "aaa", and "a′a′a′", and use them as search candidate keywords. By referring to both the common keyword dictionary 30 and the local keyword dictionary 102, the search candidate keyword reading unit 32 can set search candidate keywords without omission.

[0043] The search keyword storage unit 34 stores the input search keywords and search candidate keywords obtained as described above as search keywords. For example, if an additional query article is input, the search keyword storage unit 34 may also store the additional input search keywords and additional search candidate keywords as search keywords and accumulate them.

[0044] The information type determination unit 36 ​​determines the search intent based on information type determination keywords extracted from the query article. Referring to the information type determination dictionary 38, the information type determination unit 36 ​​determines the type of information to be retrieved based on the information type determination keywords. Figure 3 This is an example of information category judgment keywords stored in the information category judgment dictionary 38. The information category judgment dictionary 38 records information category judgment keywords that can be extracted from query articles, and the information categories corresponding to these keywords. For example, as an information category, "fault case" is recorded in relation to information category judgment keywords such as "fault" and "similar example". Similarly, as an information category, "manual" is recorded in relation to information category judgment keywords such as "type of fault" and "change order". Furthermore, the information stored in the database 12 is divided into multiple groups and managed according to each information category. For example, in the database 12, "fault cases" can be stored in a folder such as "past cases", and "manual" can be stored in other folders.

[0045] The retrieval execution unit 40 performs retrieval processing on the information stored in the database 12 using a text library containing retrieval keywords stored in the retrieval keyword storage unit 34. In other words, the retrieval execution unit 40 performs retrieval processing on the database 12 using input retrieval keywords and keywords associated with the input retrieval keywords. The retrieval execution unit 40 performs retrieval processing on information belonging to at least one group of multiple groups in the database 12 that corresponds to the information type determined by the information type determination unit 36.

[0046] The request receiving unit 42 can add or delete words registered in the common keyword dictionary 30 and the local keyword dictionary 102 by the searcher. For example, the searcher can add new synonyms, thereby enabling the search execution unit 40 to further perform a search process without omissions.

[0047] Local management device 100 is located in each of multiple regions that commonly use the information retrieval system, and manages the local database 104 for that region. Local management device 100 may also include, for example, an input device, which is operated by the region's administrator.

[0048] Local database 104 stores data created in factories or units within the region. The created data can be automatically stored in local database 104 by the regional system, including local management device 100, or stored in local database 104 by the administrator. Data stored in local database 104 includes equipment malfunction examples, manuals, and other information. As described above, the data stored in local database 104 is converted into a file format suitable for database 12 by information management device 110 and stored in database 12 periodically.

[0049] The information management device 110 converts the data stored in the local database 104 located in multiple regions into file format and stores it as storage data in the temporary storage unit 112. Then, the information management device 110 outputs the storage data when the database 12 is periodically updated with new information.

[0050] Figure 4 This is a flowchart illustrating the operation of the information management device 110. For example, the information management device 110 retrieves new data stored in a local database 104 located in multiple regions at regular intervals. As another example, the information management device 110 may also retrieve new data immediately after it has been stored in the local database 104.

[0051] If the size of the acquired data is below a threshold (as in step S1), the information management device 110 proceeds to step S2. For example, the threshold is 1 MB.

[0052] The information management device 110 performs character extraction processing (step S2) to extract only character information from the acquired data. The character information is stored in the database 12 to speed up the text database retrieval processing performed by the retrieval execution unit 40.

[0053] Information management device 110 converts data smaller than a threshold into a first file format. The first file format compresses the entire data so that it can be directly viewed by the search user on the search screen (see reference). Figure 6 The first file format can be downloaded as a file. For example, the first file format can be PDF (Portable Document Format). Then, the information management device 110 stores the extracted character information and the data converted into the first file format as storage data in the temporary storage unit 112 (step S3).

[0054] If the size of the acquired data is not below the threshold (No in step S1), the information management device 110 proceeds to step S4.

[0055] The information management device 110 performs character extraction processing (step S4) to extract only character information from the acquired data. The processing in step S4 is the same as that in step S2.

[0056] Information management device 110 converts data larger than a threshold into a second file format. The second file format is a file containing links to the entire data or links to parts other than character information. The link destination is the original storage location of the data obtained by information management device 110, i.e., a local database 104 for a specific region. The searcher can access the data from the search screen (see reference). Figure 6 Access the local database 104 of the linked destination. The second file format can also be HTML (HyperText Markup Language). Then, the information management device 110 stores the extracted character information and the data converted into the second file format as storage data in the temporary storage unit 112 (step S5).

[0057] The information management device 110 outputs the stored data in the temporary storage unit 112 to the database 12 at regular intervals (step S6).

[0058] Here, if multiple data generated in multiple regions are stored unchanged in database 12, there is a concern that the size of database 12 will increase, making high-speed retrieval impossible. Furthermore, the capacity of database 12 and temporary storage unit 112 needs to be increased. In this embodiment, the information management device 110 performs the above-described processing, thus character information, compressed data, or link information are stored in database 12 and temporary storage unit 112. In other words, the size of database 12 can be prevented from increasing, therefore the information retrieval system of this embodiment can perform high-speed retrieval. Furthermore, the capacity of database 12 and temporary storage unit 112 can be prevented from increasing, thus enabling a streamlined system architecture. As an example, by converting data into HTML as a second file format, information management device 110 can compress data containing images larger than 3MB to approximately 5kB.

[0059] Here, the local databases 104 located in multiple regions are preferably configured to have a common hierarchical structure with the database 12. Figure 5 This is a diagram illustrating the hierarchical structure of a database. For example, the hierarchical structure, starting from the top, includes regions, distinctions, item 1, item 2, etc. As a specific example, the local database 104 for region X has... Figure 5 The hierarchical structure of the folders is shown. This hierarchical structure is preferably the same as that of database 12, and is common to both regions Y and Z. In this way, local database 104 has the same hierarchical structure as database 12, so that for data displayed with links in the search results, that is, data whose main body is only stored in local database 104, the searcher can easily determine the storage destination.

[0060] Figure 6 This is a diagram illustrating, for example, the search results displayed on display device 24 after search processing. Figure 6 In the example, the retrieval execution unit 40 performs a database retrieval process on the input of an inquiry such as "What is the reason for the inverter tripping?", and obtains multiple pieces of information. Hereinafter, these multiple pieces of information will often be referred to as "detection information." The detection information includes multiple file formats of information converted by the information management device 110 according to the size of the original data. Figure 6For example, the example shows first information including a downloadable file and second information including a link to data stored in local database 104. The retrieval execution unit 40 retrieves the first and second information, which are stored as text in database 12 through character extraction processing performed by information management device 110. The retrieval execution unit 40 displays the registration date, organization number, registration region, and summary in text form, and displays the downloadable file or link as the search result to the user. Regardless of the file format converted by information management device 110, the user can see the full text of the data as the search result.

[0061] Furthermore, in this embodiment, the retrieval execution unit 40 can filter and display the retrieval results by comparing the hierarchical structure of the database 12 and the local database 104 described above. This makes it easier to see the retrieval results.

[0062] Figure 7 It is a flowchart used to explain the processing actions of an information retrieval system.

[0063] First, the query article receiving unit 26 of the information retrieval device 10 receives the query article input by the retrieval user through the input device 22 via the retrieval user side terminal 20 (step S101).

[0064] Next, the Natural Language Processing Unit 28 performs natural language processing such as morpheme analysis on the query articles received by the Query Article Receiving Unit 26 to extract words (independent words) (step S102).

[0065] Next, the input search keyword extraction unit 44 extracts (selects) input search keywords from the words extracted by natural language processing (step S103).

[0066] Next, the candidate keyword reading unit 32 refers to the keyword dictionary (common keyword dictionary 30 and local keyword dictionary 102) and reads the keywords associated with the input search keyword as search candidate keywords (step S104). Here, if the input search keyword is not registered in the local keyword dictionary 102, the candidate keyword reading unit 32 refers to the common keyword dictionary 30.

[0067] The search keyword storage unit 34 stores the input search keywords and the search candidate keywords read by the search candidate keyword reading unit 32 (step S105).

[0068] The information retrieval device 10 displays the search keywords stored in the search keyword storage unit 34 on the display device 24, and can remove unwanted search keywords from the multiple displayed search keywords as needed (step S106).

[0069] The information type determination unit 36 ​​refers to the information type determination dictionary 38 and determines the information type (search intent) based on the information type determination keywords extracted by the information type determination keyword extraction unit 46 (step S107).

[0070] The retrieval execution unit 40 performs a text database retrieval process on information stored in the database 12 that corresponds to the information type determined by the information type determination unit 36 ​​or the information type manually selected by the retrieval user, using the selected retrieval keywords (step S108).

[0071] The information retrieval device 10 displays the retrieval results on the display device 24 on the user's side (step S109).

[0072] Here, the information retrieval device 10, the retrieval-side terminal 20, the local management device 100, and the information management device 110 can also be respectively Figure 8 Such a computer. For a computer, the memory (storage device), CPU (processing device), hard disk drive (HDD), display control unit that controls the display of display device 16 or display device 24, and communication control unit for connecting to network 18 are connected via a bus. The operating system (OS) and the application program for implementing the processing in the above embodiment can be stored in the HDD and read from the HDD to the memory when executed by the CPU. The CPU controls the display control unit, communication control unit, etc., to perform the required actions as needed. Furthermore, data generated during processing is stored in the memory and, if necessary, in the HDD. All of the above functions are achieved by enabling the CPU, memory, and other hardware, the OS, and the required application programs to work together organically.

[0073] As described above, the information retrieval system according to this embodiment, by having the above-described structure, can perform high-speed retrieval even when it is used in multiple regions.

[0074] The above description is based on the illustrated examples, but the present invention is not limited to the above embodiments and various improvements and changes can be made without departing from the spirit of the invention.

[0075] For example, in the above embodiment, the local keyword dictionary 102 registers regional keywords. Then, keywords other than regional keywords can be registered in the common keyword dictionary 30. Here, the keyword dictionaries (common keyword dictionary 30 and local keyword dictionary 102) can also register keywords corresponding to categories. In other words, keywords registered in the keyword dictionary can also be classified (distinguished) by category. For example, keywords can also be classified as "equipment", "component", "cause", "status", "manufacturer", "production line", and "model". For example, "AA equipment" corresponds to the category of equipment. For example, "×× device" corresponds to the category of model and also to the category of manufacturer if it is a specific model. For example, "CC failure" corresponds to the category of status. For example, "DD component" corresponds to the category of component. For example, "board defect" corresponds to the category of cause. For example, "factory 1" corresponds to the category of production line.

[0076] When keywords are categorized, the keyword dictionary to be registered can be determined based on the category. For example, keywords corresponding to the category of equipment or production line can be registered in the local keyword dictionary 102, while keywords corresponding to components, reasons, statuses, manufacturers, or models can be registered in the common keyword dictionary 30. By registering and managing keywords related to equipment and production lines that are highly relevant to the region using the local keyword dictionary 102, efficient keyword setting and retrieval can be achieved. As another example, the common keyword dictionary 30 can also repeatedly register at least a portion of the keywords corresponding to the categories registered in the local keyword dictionary 102. In this case, the management of keywords repeatedly registered in the common keyword dictionary 30 and the local keyword dictionary 102 can also be performed by the local keyword dictionary 102.

[0077] Industrial availability

[0078] According to the present invention, an information retrieval system is provided that enables high-speed retrieval even when used in multiple regions.

[0079] Explanation of reference numerals in the attached figures

[0080] 10...Information retrieval device; 12...Database; 14...Input device; 16...Display device; 18...Network; 20...Retriever-side terminal; 22...Input device; 24...Display device; 26...Query article receiving unit; 28...Natural language processing unit; 30...Common keyword dictionary; 32...Search candidate keyword reading unit; 34...Search keyword storage unit; 36...Information type judgment unit; 38...Information type judgment dictionary; 40...Retrieval execution unit; 42...Search candidate keyword change request receiving unit; 44...Input search keyword extraction unit; 46...Information type judgment keyword extraction unit; 100...Local management device; 102...Local keyword dictionary; 104...Local database; 110...Information management device; 112...Temporary storage unit.

Claims

1. An information retrieval system, characterized in that, have: A database that stores multiple pieces of information capable of text retrieval; The Inquiry Article Receiving Department accepts inquiry articles in natural language format. An input search keyword extraction unit extracts input search keywords from the query article; A common keyword dictionary, which registers related keywords in a corresponding manner; A local keyword dictionary that registers regional keywords used in a specific region and is configured to distinguish which region each regional keyword is used in. The candidate keyword reading unit reads keywords associated with the input search keyword by referring to at least one of the common keyword dictionary and the local keyword dictionary; as well as The retrieval execution unit uses the input search keywords and keywords associated with the input search keywords to perform retrieval processing on the database. When the input search keyword is not registered in the local keyword dictionary, the search candidate keyword reading unit refers to the common keyword dictionary to read keywords associated with the input search keyword. When the input search keyword is registered in the local keyword dictionary, the unit reads synonyms and related terms from the local keyword dictionary and adds them to the search candidate keywords. The unit also reads synonyms and related terms from the common keyword dictionary and adds them to the search candidate keywords.

2. The information retrieval system according to claim 1, characterized in that, The local keyword dictionary registers the regional keywords in correspondence with the keywords registered in the common keyword dictionary. The search candidate keyword reading unit refers to the common keyword dictionary and the local keyword dictionary to read keywords associated with the input search keyword.

3. The information retrieval system according to claim 1 or 2, characterized in that, The common keyword dictionary and the local keyword dictionary will register synonymous keywords in a corresponding manner.

4. The information retrieval system according to claim 1 or 2, characterized in that, The common keyword dictionary and the local keyword dictionary will register keywords of similar meanings in a corresponding manner.

Images