JP2016134124A - Tag application method, tag application system, program, and query/response retrieval method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a tag application method making it possible to apply a tag, which is used to retrieve a response document that meets a predetermined standard concerning a user who performs retrieval or an object of retrieval, at the time of retrieving the response document relevant to a user's query.SOLUTION: A tag application method in accordance with the present invention is a tag application method to be implemented in a tag application system that applies a tag, which is used to retrieve a response, to a response document signifying a response to a user's query. A tag adjustment document for use in adjusting a topic, which is applied as a tag to the response document, according to a predetermined standard concerning a user who performs retrieval or an object of retrieval is selected. The topic indicating the response document and the selected tag adjustment document is extracted from the response document and selected tag adjustment document. The extracted topic is applied as the tag, which meets the predetermined standard, to the response document.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a tag assigning method for assigning a tag for searching for an answer to an answer document constituting an answer to a question from a user.

  In recent years, a question answering system based on FAQ (Frequently Asked Questions) has been developed. The question answering system is a computer system in which a user inputs a question about a specific field or type in a natural sentence, and the system returns an appropriate answer to the question. Generally, in a question answering system, a group of documents to be answered is stored in a database.

  In such a question answering system, it is necessary to appropriately search for an answer suitable for the user's intention from a large number of answer documents. As one means for realizing this, a technique is disclosed in which a large number of comment sentences relating to a question keyword are classified into potential topic groups, and an answer sentence similar to a sentence group included in the topic group is associated ( For example, see Patent Document 1).

  In addition, as another means for appropriately searching for an answer suitable for the user's intention from a large number of answer documents, a search tag may be assigned in advance to the answer document. The search tag is a word that appropriately represents the contents of the answer document. If the search tag is included in the user's question sentence, the corresponding answer is presented to the user. This improves the search accuracy for a huge document group.

  In adding search tags, it takes a lot of cost to manually tag a large number of documents. Therefore, automatic search tag addition that automatically extracts words characterizing a document is also being considered (for example, Non-Patent Document 1).

JP 2013-143066 A

Ryo Kato, Daihiro Yoshikawa, Takeshi Furuhashi “Study on Automatic Tagging of Documents Using Potential Dirichlet Allocation Method”, Proceedings of National Conference of the Japanese Society for Artificial Intelligence 28, 1-4, Japan Society for Artificial Intelligence May 1, 2014

  However, in the automatic search tagging, further improvements are required so that appropriate tagging can be performed according to the user's knowledge level with respect to the question target (that is, the answer search target) and the answer search target.

  In order to solve the above-mentioned problem, a tag assignment method according to the present invention is a tag assignment method performed by a tag assignment device for assigning a tag for searching for an answer to an answer document constituting an answer to a question from a user. And selecting a tag adjustment document for adjusting a topic to be assigned as a tag to the answer document according to a predetermined criterion related to a user who performs the search or the search target, and the answer document and the selected document A topic representing the answer document and the selected tag adjustment document is extracted from the tag adjustment document, and the extracted topic is assigned to the answer document as a predetermined reference tag.

  According to this configuration, for a reply document, a user who asks a question, a predetermined criterion (for example, a user's knowledge level for a home appliance, It is possible to attach a search tag according to whether it is a function installed in a new model of home appliances or a function installed in an old model. As a result, an appropriate answer document can be searched for a question from the user according to the knowledge level of the questioning user and the search target of the answer.

It is a figure which shows an example of the whole structure of the question answering system which concerns on embodiment of this invention. It is a block diagram which shows an example of a structure of the control server 1 which concerns on embodiment of this invention. It is a block diagram which shows an example of a structure of the tag provision server 2 which concerns on embodiment of this invention. It is a block diagram which shows an example of a structure of the question answering server 3 which concerns on embodiment of this invention. It is a block diagram which shows an example of a structure of the operating terminal 4 which concerns on embodiment of this invention. It is a sequence diagram which shows an example of a series of processes of the question answering system which concerns on embodiment of this invention. It is a flowchart which shows an example of the procedure of the server selection process which concerns on embodiment of this invention. It is a flowchart which shows an example of the procedure of the document selection process which concerns on embodiment of this invention. It is a flowchart which shows an example of the procedure of the tag extraction / linking process which concerns on embodiment of this invention. It is a figure which shows an example of the tag adjustment document determination table by this embodiment which concerns on embodiment of this invention. It is a figure which shows an example of the display screen which concerns on embodiment of this invention. It is a figure which shows the other example of the display screen which concerns on embodiment of this invention. It is a figure which shows an example of the data structure of the reply database 5 which concerns on embodiment of this invention. It is a figure which shows an example of the data structure of the tag adjustment document database 6 which concerns on embodiment of this invention. It is a figure which shows an example of the experiment result of tag provision by the topic model which concerns on embodiment of this invention.

(Knowledge that became the basis of the present invention)
The inventor has studied in detail a question answering system that answers a user's question by searching for tagged answer documents. As a result, in the above-described conventional technology, when searching for an answer to a certain question, an appropriate answer sentence is not returned according to the knowledge level of the user for the question target (that is, the answer search target) or the answer search target. I found a problem that there was a case. This will be described by taking the difference in the user's knowledge level (skill level) with respect to the search target as an example.

  For example, if a certain answer document has a search tag of “access, LAN, proxy server”, a question about “I want to set a proxy server” that a user with a high level of knowledge about network technology can ask Because the word “proxy server” included in the URL matches one of the search tags, users who have a low level of knowledge about network technology can search for this answer document and respond based on the contents of the answer document. Cannot answer this question because the search tag does not match any words included in the question. There arises a problem that a reply based on the content of the document cannot be made.

  In order to solve such a problem, the tag assignment method according to the present invention is a tag performed by a tag assignment device that assigns a tag for searching for an answer to an answer document that constitutes an answer to a question from a user. A method for providing a tag adjustment document for adjusting a topic to be added as a tag to the answer document according to a predetermined criterion related to a user who performs the search or the search target, and the answer document and the answer document A topic representing the answer document and the selected tag adjustment document is extracted from the selected tag adjustment document, and the extracted topic is assigned to the answer document as a predetermined reference tag.

  As a result, it is possible to attach an appropriate search tag to the answer document according to the user's knowledge level with respect to the question target (that is, the answer search target) and the predetermined criteria regarding the answer search target.

  The predetermined criterion may be whether the user's proficiency level or knowledge level with respect to the search target is high or low.

  This makes it possible to assign an appropriate search tag to the answer document based on the user's proficiency level or knowledge level for the search target.

  The search target may be a product, and the answer document may be at least a part of a document explaining handling of the product.

  Thereby, it becomes possible to assign an appropriate search tag with respect to the handling of the product based on the user's proficiency level or knowledge level.

  Further, a document relating to general knowledge may be selected as the tag adjustment document for assigning a tag used for a search by a user having a low proficiency level or knowledge level with respect to the search target.

  The general knowledge document may be at least one of a dictionary, textbook, news, blog, or short text posting service.

  Thereby, it is possible to attach a tag suitable for a user with a low knowledge level to a response document using a document in which simpler words are described than a document explaining the function of the electronic device.

  Moreover, it is good also as selecting the document regarding expert knowledge as the said tag adjustment document for providing the tag used for the search by the user with a high skill level or knowledge level with respect to the said object of search.

  Further, the document related to the specialized knowledge may be at least one of a construction installation technical manual, an instruction manual, an academic technical book, an in-house document, or sales information.

  Thereby, it is possible to give a tag suitable for a user having a low knowledge level to a response document using a document in which a higher level of words is described than a document explaining the function of the electronic device.

  The search target may be a product, and the predetermined criterion may be information on a new function or information on an old function of the product.

  This makes it possible to attach an appropriate search tag to the answer document related to the product function based on either the information related to the new function of the product or the information related to the old function.

  The answer document is at least a part of a document explaining the function of the product, and the tag adjustment document is at least a part of a document explaining a function of a new model or an old model from the product. Also good.

  As a result, based on the document on the function installed in the new model or the document on the function installed in the old model, it is based on either the information on the new function of the product or the information on the old function. As a result, an appropriate search tag can be assigned.

  The search target may be a product, and the predetermined reference may be information related to a higher model or a lower model of the product.

  Accordingly, it is possible to attach an appropriate search tag to the answer document regarding the function of the product on the basis of either the information regarding the higher model of the product or the information regarding the lower model.

  The answer document is at least a part of a document explaining the function of the product, and the tag adjustment document is at least a part of a document explaining a function of a higher model or a lower model than the product. Also good.

  As a result, based on the document on the function installed in the higher model or the document on the function installed in the lower model, it is based on either the information on the higher model of the product or the information on the lower model. As a result, an appropriate search tag can be assigned.

  In addition, the extraction of keywords assigned as tags may be performed by extracting features of each document using a topic model for automatically extracting topics that constitute the document.

  In addition, the topic model may be at least one of LDA (Lent Dielectric Allocation), LSI (Lent Semantic Indexing), or PLSA (Probabilistic Latin Semantic Analysis).

  Thereby, it is possible to extract a keyword used as a search tag from the answer document and the tag adjustment document using the topic model.

  The tag may be assigned to the answer document by selecting a tag group including the most answer documents from among a plurality of tag groups obtained from the topic model.

  This makes it possible to assign the most suitable keyword extracted from the answer document and the tag adjustment document as a search tag.

  The tag adjustment document may be selected based on the user's action history.

  As a result, the user's knowledge level and the like can be determined based on the user's behavior history, and a tag adjustment document corresponding to the knowledge level can be selected.

  The question search method according to the present invention is a question search method performed by a question answering system for searching for an answer document that matches a user's question from a group of documents to be searched, the user performing the search or the search target A predetermined criterion for determining a tag adjustment document for adjusting a topic to be assigned as a tag to the answer document based on the predetermined criterion, and selecting from the answer document and the selected tag adjustment document , Extracting a topic representing the answer document and the selected tag adjustment document, assigning the extracted topic to the answer document as a predetermined reference tag, extracting a keyword from the user's question, The answer document to which a tag matching the keyword is assigned is searched.

  As a result, an appropriate search tag is given to the answer document according to the user's knowledge level for the question target (that is, the answer search target) and the predetermined criteria regarding the answer search target, and then the user's question It is possible to search for an answer document suitable for.

  Note that each of the embodiments described below shows a specific example of the present invention. Numerical values, shapes, components, steps, order of steps and the like shown in the following embodiments are merely examples, and are not intended to limit the present invention. In addition, among the constituent elements in the following embodiments, constituent elements that are not described in the independent claims indicating the highest concept are described as optional constituent elements. In all the embodiments, the contents can be combined.

(Embodiment)
Hereinafter, embodiments of the present invention will be described with reference to the drawings. The embodiments of the present invention are not limited to the following examples, and can take various forms as long as they belong to the technical scope of the present invention.

FIG. 1 is a diagram showing an example of the overall configuration of the question answering system in the present embodiment.
The question answering system includes a control server 1, a tag assignment server 2, a question answering server 3, an operation terminal 4, an answer database 5, a tag adjustment document database 6, and a network 7.

  The control server 1 has a function of transmitting / receiving data to / from the tag assignment server 2, the question answering server 3, and the operation terminal 4. For example, for the input obtained from the operation terminal 4, the server to be processed is selected from the tag assignment server 2 or the question answering server 3. The detailed configuration of the control server 1 will be described later.

  The tag assignment server 2 has a function of transmitting / receiving data to / from each of the control server 1, the answer database 5, and the tag adjustment document database 6. For example, the information of the knowledge level with respect to the search target of the user input from the operation terminal 4 is received, and processing such as selecting a tag adjustment document taking into account the user's knowledge level from the tag adjustment document database 6 is performed. The search target here is, for example, usage of a product, system or application program, information on a failure, or the like. The search target may be tangible and intangible merchandise including services, financial products, documents, and the like. Here, the product refers to articles that require explanation for handling, such as household electrical equipment, information communication equipment, automobiles, and furniture. The knowledge level represents the amount of knowledge, proficiency, etc. that the user has for the search target. Anything other than those listed here is not limited as long as an answer to a user's question can be prepared. Hereinafter, the “knowledge level for the search target of the user” is referred to as “knowledge level”. The detailed configuration of the tag assignment server 2 will be described later.

  The question / answer server 3 has a function of transmitting / receiving data to / from the control server 1 and the answer database 5. For example, the question answering server 3 receives a user question input from the control server 1 and performs processing such as searching the contents of the answer database 5. The detailed configuration of the question answering server 3 will be described later.

  The operation terminal 4 is a smartphone, a tablet terminal, a dedicated display device terminal, a personal computer (PC), a television, or the like. Any terminal other than those listed here can be used as long as it can exchange data with the user. The operation of the terminal is assumed to be input using a touch panel, input using an image or moving image, input using sound, input using a keyboard, and the like.

  The answer database 5 exchanges data with the tag assignment server 2 and the question answer server 3. Here, an answer document which is a document constituting an answer to the question is held. Examples of questions to be asked include how to use a product, system, or application program, and trouble consultation, and answer a user's question by presenting a QA. This is only an example, and the present invention is not limited to this as long as a question and an answer to the question can be prepared. An example of the data storage format is an XML (Extensible Markup Language) format, but is not limited to this as long as the data can be structurally held. FIG. 12 shows an example of an XML format, in which tags extracted automatically together with a document to be answered are stored.

  In the following description, the answer document is sometimes referred to as a tagging target document, but it means the same thing.

  The tag adjustment document database 6 exchanges data with the tag assignment server 2. Here, a tag adjustment document, which is a document for adjusting a tag word according to a predetermined standard when a tag is added to a tagging target document, is held. Here, the predetermined standard refers to a standard regarding a user who asks a question (that is, searches for an answer to the question) and a question target (that is, a search target of the answer). When the predetermined standard is the user's knowledge level, the tag adjustment document has, for example, general knowledge (a dictionary, textbook, news, blog (abbreviation of Weblog), short text posting service such as Twitter (registered trademark), etc.). Documents and documents having high expertise (construction installation manuals, instruction manuals, academic technical books, in-house documents, sales information, etc.) are retained. The dictionary refers to both a dictionary on the Internet such as Wikipedia and a dictionary as a book. These are only examples, and are not limited to this as long as they are documents. An example of the data storage format is an XML (Extensible Markup Language) format, but is not limited to this as long as the data can be structurally held.

  The network 7 is connected to the control server 1 and the operation terminal 4. An example of the connection method is wired or wireless LAN connection, but is not limited to this as long as each component is connected to be communicable.

  FIG. 2 is a block diagram showing an example of the configuration of the control server 1 in the present embodiment.

  As illustrated in FIG. 2, the control server 1 includes an input detection unit 101, a process selection unit 102, a control unit 103, and a communication unit 104.

  The input detection unit 101 detects an input from the user through the operation terminal 4.

  The process selection unit 102 is a part that selects a server to be processed from the tag assignment server 2 or the question answering server 3 in response to a user input. For example, when information indicating a standard such as “knowledge level: 1” is acquired, processing in the tagging server 2 is selected, and information indicating a question such as “ineffectiveness of air conditioner” is acquired. A process in the question answering server 3 is selected.

  The control unit 103 transmits a command for processing to the server selected by the processing selection unit 102.

  The communication unit 104 communicates with the tag assignment server 2, the question answering server 3, and the operation terminal 4.

  FIG. 3 is a block diagram showing an example of the configuration of the tag assignment server 2 in the present embodiment.

  As shown in FIG. 3, the tag assignment server 2 includes a reference acquisition unit 201, a document selection unit 202, a morpheme analysis unit 203, a vector generation unit 204, a tag automatic generation unit 205, a tag association unit 206, a communication unit 207, A morphological analyzer 208, an answer database 5, and a tag adjustment document database 6 are provided.

  The reference acquisition unit 201 acquires information serving as a reference for selecting a tag adjustment document such as a user knowledge level. The reference information may be acquired by an input of a user who makes a question, or may be acquired automatically. For example, when the reference information is the user's knowledge level, the user's knowledge level can be automatically determined and acquired from the user's question history or the like.

  The document selection unit 202 selects a tag adjustment document suitable for reference information from the tag adjustment document database 6 and a tag target document from the response database 5. For example, when the user knowledge level is used as a reference, a tag adjustment document is selected with reference to a table as shown in FIG. 10 that associates the user knowledge level with the difficulty level of the tag adjustment document. Note that the numerical values and the tag adjustment document illustrated in FIG. 10 are examples, and the present invention is not limited thereto. Further, these tag adjustment document determination methods are merely examples, and are not limited thereto. For example, the tag adjustment document may be selected based on some mathematical formula, or may be selected statistically using an algorithm such as machine learning.

  The morpheme analysis unit 203 is a part that receives a document from the document selection unit 202 and performs morpheme analysis of the document via the morpheme analyzer 208. Morphological analysis is a technique for dividing a document into meaningful words and discriminating parts of speech and contents using a dictionary. A “morpheme” is the smallest meaningful unit of document elements. In the morphological analysis unit 203, the document is data having part-of-speech information for each word.

  The vector generation unit 204 is a part that receives the document subjected to the morphological analysis by the morphological analysis unit 203 and expresses the document as a vector. At this time, since information with a part of speech is obtained from the morphological analysis unit 203, only a specific part of speech may be employed. Further, words having a high appearance frequency may be weighted only for a specific document by TF-IDF (Term Frequency-Inverse Document Frequency). The weighting is an example, and methods other than those listed here may be used.

  The tag automatic generation unit 205 generates a tag by extracting topics representing these documents from the answer document and the tag adjustment document. A topic model can be used to extract a topic representing a document. A topic model is a model that statistically estimates potential topics hidden in data. Specifically, there are LDA (Lent Dirichlet Allocation), LSI (Lentent Semantic Indexing), PLSA (Probabilistic Latin Semantic Analysis), etc., but not limited to this as long as the topic representing the document can be extracted. .

  The tag association unit 206 associates, as a tag, a topic that characterizes the answer document most among a plurality of topics (tag group) extracted from the topic model.

  The morpheme analyzer 208 exchanges data with the morpheme analyzer 203. For example, an open source morphological analysis engine “MeCab” can be used. Furthermore, according to this engine, the part of speech of a morpheme can also be analyzed. For example, “noun”, “verb”, etc. for each morpheme. Note that it is not always necessary to perform morpheme analysis, and a mechanism for extracting words in a document by referring to a prepared keyword dictionary or the like without passing through the morpheme analysis unit 203 and the morpheme analyzer 208 may be used. For example, when extracting words for a document “air conditioner temperature setting method”, if “air conditioner”, “temperature”, “setting”, and “method” are registered in the keyword dictionary, those words are extracted. Is. These are merely examples, and the method is not limited to this as long as the method can extract words.

  The communication unit 207 communicates with the control server 1, the answer database 5, and the tag adjustment document database 6.

  FIG. 4 is a block diagram showing an example of the configuration of the question answering server 3 in the present embodiment.

  As shown in FIG. 4, the question / answer server 3 includes a question input unit 301, a morpheme analysis unit 302, a keyword extraction unit 303, a question search unit 304, a communication unit 305, a morpheme analyzer 306, and an answer database 5.

  The question input unit 301 receives a user question input from the operation terminal 4.

  The morpheme analysis unit 302 receives a document from the question input unit 301 and performs morpheme analysis of the document via the morpheme analyzer 306. In addition, it is desirable if it has a function equivalent to the morphological analysis part 203 in FIG.

  The keyword extraction unit 303 extracts keywords from the result obtained from the morphological analysis unit 302. At this time, since information with a part of speech is obtained from the morphological analysis unit 302, a word having a specific part of speech is selected as a keyword. For example, only “noun” and “verb” are acquired as keywords.

  The question search unit 304 is a part that searches the answer database 5 for answers using the keywords obtained from the keyword extraction unit 303.

  The morpheme analyzer 306 exchanges data with the morpheme analyzer 302. In addition, what has the same function as the morphological analyzer 208 in FIG.

  The communication unit 305 communicates with the control server 1 and the answer database 5.

  FIG. 5 is a block diagram showing an example of the configuration of the operation terminal 4 in the present embodiment.

  As illustrated in FIG. 5, the operation terminal 4 includes an input detection unit 401, a control unit 402, a display unit 403, and a communication unit 404.

  The input detection unit 401 acquires input information using a touch panel, voice, buttons, and the like from the user.

  The control unit 402 performs transmission / reception of information with the input detection unit 401 and the display unit 403 and exchanges with the communication unit 404. For example, when information indicating a criterion such as “knowledge level: 1” or information indicating a question such as “ineffectiveness of air conditioner” is received from the user, the information is transmitted to the communication unit 404 through the control unit 402.

  The display unit 403 displays information obtained from the control unit 402. For example, an answer to a user's question.

  The communication unit 404 transmits the information of the input detection unit 401 to the tagging server 2 or the question answering server 3 via the control unit 402, and displays the screen from the tagging server 2 or the question answering server 3 through the control unit 402. Receive information about.

  FIG. 6 is a sequence diagram showing an example of a series of processes of the question answering system in the present embodiment. In the example of FIG. 6, the knowledge level of the user is used as a reference for selecting the tag adjustment document. In the example of FIG. 6, a series of communication processes until the answer database 5 is updated and the answer is output to the operation terminal 4 in response to the user's question is shown. This sequence is started when any input is detected on the operation terminal 4.

  In step S001, the operation terminal 4 acquires an input related to the user's knowledge level.

  In step S002, it is selected from the information obtained in step S001 whether the process is to apply a tag that matches the knowledge level of the user to the answer database 5 or the process of answering the user's question. Here, since information related to the knowledge level has been input, a tagging process is performed. Since these processes are shown in FIG. 7, they will be described later.

  In step S003, information about the user's knowledge level is acquired from S002. In step S 004, a tagging target document is acquired from the answer database 5. Further, referring to a table showing the relationship between the user knowledge level and the tag adjustment document as shown in FIG. 10, an appropriate tag adjustment document is selected from the tag adjustment document database 6 in step S005. The processing here is shown in FIG. 8, and will be described later.

  In step S006, a tag is extracted from the document selected from steps S003, S004, and S005, and the tag that most indicates the content of the tagging target document is associated with the document. The processing here is shown in FIG. 9 and will be described later.

  In step S007, the database is updated by reflecting the tag in the answer database 5.

  In step S008, the operation terminal 4 acquires an input of a user's question.

  In step S009, it is selected from the information obtained in step S008 whether the process is to apply a tag that matches the user's knowledge level to the answer database 5 or to answer the user's question. Here, since the information about the question is input, the question answer process is performed. Since these processes are shown in FIG. 7, they will be described later.

  In step S010, keywords are extracted from the user's question text.

  In step S011, a process for searching for an answer to the user's question is performed.

  In step S012, the answer database 5 is searched.

  In step S013, the answer to the question is displayed on the operation terminal 4.

  In the conventional method, the same tag is attached to any user. Because the words used for questions differ depending on the user's level of knowledge, it was not always possible to perform a correct search for everyone. According to this method, the user himself / herself declares the knowledge level, or a word suitable for the user's knowledge level can be automatically given as a tag based on the use history of the user's question answering system. For this reason, any user can obtain a desired answer. In addition, since the usage history of the question answering system can be automatically updated, the tag can be automatically updated even when the knowledge level is improved.

  FIG. 7 is a flowchart showing an example of the procedure of the selection process of the tag assignment server 2 or the question answering server 3 in the present embodiment. This process is the details of the server selection process (S002, S009) in FIG.

  In step S201, the user input obtained from the operation terminal 4 is acquired.

  In step S202, it is determined whether the data obtained from step S201 includes data relating to the knowledge level (that is, information relating to criteria for selecting the tag adjustment document). If it is determined that information on the knowledge level is included, the knowledge level information is transmitted to the tag assignment server 2 in step S203. In step S204, it is determined whether or not the user's question is included in the data obtained from step S201. If it is determined that the information related to the question is included, the information related to the question is transmitted to the question answering server 3 in step S205.

  For example, when the information “I want to know how to connect to the Internet” is obtained, the information is related to the question, so the information is transmitted to the question answering server 3 and the question answering process is performed. Further, when information such as “knowledge level: 1” is obtained, the information is related to the knowledge level, so the information is transmitted to the tag assignment server 2 and the tagging process is performed. FIG. 11 is a diagram showing an example of a display screen of the operation terminal 4 when a user's knowledge level is input according to the present embodiment. However, these display methods are examples and are not limited thereto.

  11A and 11B are diagrams each showing an example of a screen displayed on the operation terminal 4 when the user asks a question. For example, as shown in FIG. 11A, the user's own knowledge level is input for each product or the entire product. Further, for example, as shown in FIG. 11B, by inputting the user's own knowledge level to the question target product at the same time as the user's question, the search can be performed with the tag suitable for the user.

  11A and 11B, the user directly inputs the knowledge level. However, information associated with the user's knowledge level may be input. For example, in the case of a question about how to use home appliances, a screen for selecting whether the user is “young” or “elderly”, or “general user”, “salesman”, “ A screen for selecting one of “engineer” may be presented, and the knowledge level may be determined based on the input information. Also, if the product is primarily used by users of a specific gender, users of a specific age, or users who live in a specific region, the gender, age, A screen to be input may be presented, and the knowledge level may be determined based on the input information.

  In addition, before accepting a question as shown in FIG. 11A, if there is a usage like an authoring tool for tagging a response document in advance by an answer provider, a tag is given when inquiring a question as shown in FIG. 11B It does not matter if you do this. By attaching a tag when inquiring a question, the authoring tool can cope with things that could not be handled in advance. Furthermore, since only a tag tailored to the user can be assigned, an unnecessary tag is not searched. For example, even if the tag “Internet, connection” is given in the authoring tool, if the user's knowledge level is improved at the time of the question, tagging is performed at the time of the question, and “Internet, LAN, IP address, browser” If it is changed, etc., a more appropriate answer can be presented.

  With the above processing, it is possible to separate the processing performed by each server. Accordingly, an appropriate answer can be made to the user's question while obtaining the user's knowledge level and sequentially updating the tags of the answer database 5.

  FIG. 8 is a flowchart showing an example of a processing procedure for selecting a tag adjustment document for a tagging target document in the present embodiment. This process is a detail of the document selection process (S003) in FIG.

  In step S301, information on the knowledge level of the user (that is, information on criteria for selecting a tag adjustment document) is acquired. Here, the information is converted into information that can be read by a table as shown in FIG. Examples of words representing the knowledge level include “low / medium / high”, “easy / ordinary / difficult”, and the like. These are only examples, and those other than those listed here may be used. It is assumed that a mechanism for converting these into information held in the table is provided. FIG. 13 is a diagram showing an example of the data structure of the tag adjustment document database 6 according to the embodiment of the present invention. As shown in FIG. 13, the data of the tag adjustment document database 6 may include a plurality of knowledge level expressions.

  In step S302, a tag adjustment document corresponding to the knowledge level acquired in step S301 is selected from the tag adjustment document database 6. At this time, for example, the correspondence relationship is acquired from a table of user proficiency levels and tag adjustment documents as shown in FIG. Data can be held in any format as long as the correspondence relationship is known, but the present invention is not limited to this. In addition, any number of documents may be selected as long as the text matches the knowledge level.

  In step S303, the tagging target document and the tag adjustment document selected in step S302 are combined. For example, suppose you want to add a tag to a page of an air conditioner instruction manual. At the same time, when the information that the user's knowledge level is “4 (high)” is obtained, the air conditioning construction installation special book is selected as the tag adjustment document by the correspondence table of the knowledge level and the tag adjustment document. Next, a document related to the air conditioner installation specialization book is acquired from the tag adjustment document database 6 and combined with the tagging target document. Here, when two air conditioner construction attachment technical books are selected, for example, one document to be tagged and three documents are sent to the next processing.

  Through the above processing, a document to be combined with a tagging target document is automatically selected. As a result, it is not necessary for the user to select a tag adjustment document suitable for the document to be tagged, and the burden on the user can be reduced.

  FIG. 9 is a flowchart illustrating an example of a processing procedure for extracting and linking a tag to a tagging target document according to the present embodiment. This process is a detail of the tag extraction / linking process (S006) in FIG.

  In step S601, morphological analysis of the acquired document is performed. A document subjected to morphological analysis is divided into meaningful words and becomes data having part-of-speech information.

  In step S602, the data obtained from step S601 is vectorized for each document. At this time, since information with parts of speech is obtained, only specific parts of speech may be employed. Further, words having a high appearance frequency may be weighted only for a specific document by TF-IDF (Term Frequency-Inverse Document Frequency). The weighting is an example, and methods other than those listed here may be used.

  In step S603, a document tag is extracted using the topic model. A topic model is a model that statistically estimates potential topics hidden in data. Here, topic refers to a group of topics and meanings. If a topic model is applied to a plurality of prepared documents and a topic model is applied, topics corresponding to the specified number of topics are obtained. Here, the topic name is not obtained, but the semantic relationship between documents and words is statistically found, and it is understood what topic each document is composed of. Further, in step S604, the topic that appears most in each document is regarded as the topic that best represents the document, and the words included in the topic are assigned as tags. Specific examples of the topic model include LDA (Lent Dirichlet Allocation), LSI (Lent Semantic Indexing), PLSA (Probabilistic Latin Semantic Analysis), and the like, as long as a topic representing the document can be extracted. Not limited to. In addition, the number of topics specifies the number of documents or more.

  In step S604, the extracted topic is linked to the tagging target document as a tag. At this time, the topic most representative of the document to be tagged is selected, and the answer database 5 is updated. For example, it is possible to select a topic that includes the word of the document to be tagged most, but this is an example, and methods other than those listed here may be used.

  For example, it is assumed that one page of an air conditioner instruction manual and two documents of tag adjustment documents (knowledge level: 1) are obtained. Assume that tags are extracted after morphological analysis and vectorization. As a result of the tag extraction, for example, topics such as Topic 1: “Daily, Sensor, Adjustment”, Topic 2: “TV Field, Smartphone, TV” Topic 3: “Sports, Competition, Player” are obtained. At this time, when one page of the document to be tagged is a topic of the daily sensor, the topic “daily, sensor, adjustment” is selected as the tag, and the tag of the answer database 5 is updated. Since the content of the tag adjustment document may be a sentence that is not directly related to the content of the air conditioner, words such as “TV column, smartphone, TV” in Topic 2 may be words extracted from the tag adjustment document High nature. When the knowledge level of the tag adjustment document is set high, words that are used by people with a high knowledge level such as “adjusted solar radiation sensor, outdoor temperature, outdoor unit” are assigned as tags.

  Through the above processing, different topics are extracted depending on the contents of the tag adjustment document, and an appropriate topic is assigned as a tag to the tagging target document. Thus, by automatically extracting the tag from the user's knowledge level and assigning it to the document, an answer tailored to the user becomes possible.

  FIG. 12 shows an example of XML held in the answer database 5 in the present embodiment. For example, in this way, the text, tags, etc. are held in a structured state. Further, every time the user's knowledge level is updated, the contents of this tag are also updated. The items listed here are merely examples, and there may be other items. Also, the XML format is given as an example of the data storage format, but it is not limited to this as long as the data can be structurally held.

  FIG. 13 shows an example of XML held in the tag adjustment document database 6 in the present embodiment. For example, in this way, the text, knowledge level, etc. are held in a structured state. The fluctuation of the knowledge level expression may be included in the knowledge level item as shown in FIG. 13, or a mechanism for converting the knowledge level expression may be provided separately. The items listed here are merely examples, and there may be other items. Also, the XML format is given as an example of the data storage format, but it is not limited to this as long as the data can be structurally held.

  FIG. 14 is a diagram illustrating an example of a result of topic extraction using LDA in the present embodiment. For example, suppose that tagging is performed using the topic model which is the present technique in order to add a tag to an item related to the Internet connection of the device in the instruction manual. When the document to be tagged is a page related to the Internet connection of the device and the tag adjustment document is an instruction manual having contents different from those of the Internet connection, words as shown in FIG. 14A are extracted. Further, when the tag target document is a page related to the Internet connection of the device and the tag adjustment document is a general content selected from Wikipedia, words as shown in FIG. 14B are extracted. Further, as words obtained by either combination, words as shown in FIG. 14C appear as common tags. In the original method, for example, a word as shown in FIG. 14A is uniquely assigned as a tag. However, when this method is used, a word as in FIG. 14B is also assigned as a tag. This indicates that tagging tailored to the user is possible. As described above, when the combination of documents is changed, since the words constituting the model in the topic model are different, the obtained topics and the words included therein are also different.

  As described above, the tag assignment device according to one aspect of the present invention has been described based on the embodiments. However, the present invention is not limited to these embodiments. Unless it deviates from the meaning of this invention, the form which carried out various deformation | transformation which those skilled in the art can think to this embodiment, or the structure constructed | assembled combining the component in different embodiment is also contained in the scope of the present invention. .

  For example, suppose you want to search only for functions newly installed in a certain home appliance. At this time, by extracting a tag by combining an instruction manual of a certain household electrical appliance and an instruction manual of a successor model (new model), for example, only features of the succeeding model (new model) can be assigned as a tag. . Similarly, it is needless to say that only the features of the old model can be added as a tag by extracting the tag by combining the instruction manual of a certain home appliance and the instruction manual of the old model. Further, by extracting a tag by a combination of a higher model and a lower model, for example, a function that is available only in a higher model can be given as a tag.

  If a tag is to be assigned to a certain instruction manual, the tagging target document may be one page of the instruction manual and the other pages may be tag adjustment documents. For example, in the case of an instruction manual related to an air conditioner, a tag is assigned to a tagging target document by setting the tagging target document as a page related to air volume adjustment and setting the tag adjustment document as a page other than the air volume adjustment of the air conditioner. Further, by performing these operations for all pages of the instruction manual, tags are assigned to all pages of the instruction manual. The page is an example, and an item or a table of contents may be used.

  Further, the tag adjustment document database 6 may be connected to a network at all times to have a mechanism for adding / updating the contents of the database. The contents for updating the tag adjustment document database 6 include automatically contents such as a dictionary on the Internet such as Wikipedia, news, blog, or a short text posting service such as Twitter (registered trademark). Can be updated. Information such as new product press releases may also be incorporated. These are only examples, and the document data is not limited thereto.

  Further, in FIG. 1 showing the entire system configuration, the control server 1, the tag assignment server 2, and the question answering server 3 may be configured by one server or may be different from the configuration shown in FIG. . For example, when used as an authoring tool, the question answering server 3 may not be provided. The answer database 5 and the tag adjustment document database 6 may be configured as one database.

  In the example of this embodiment, when a user's knowledge level is input from the operation terminal 4, a tag is assigned to the answer document in the answer database 5. In the question answering system, the knowledge level is acquired from the user's action history. If possible, the tag of the document in the answer database 5 may be updated automatically or manually. For example, if a user's web browsing history or words used in email exchange are acquired, the knowledge level may be statistically calculated using, for example, machine learning. Further, the tag of the answer database 5 may be updated based on information such as what kind of word is used from the history of past questions and whether the answer to the question is appropriate. These are merely examples, and the present invention is not limited to this as long as the information can obtain the user's knowledge level.

  Note that any standard may be used for the level of difficulty in the knowledge level in the example of the present embodiment. For example, the terms to be tagged may be changed in terms of skill levels for some technologies, differences in terms used by salespeople or engineers, and differences in terms used by young people and elderly people. Moreover, these are examples, and it is not restricted to this as long as a word changes with criteria.

  Further, in the present embodiment, the question answering system relating to the usage method of home appliances has been described as an example, but application to other fields is also possible. For example, in a medical diagnosis system, first, a user who asks a question is inquired about a knowledge level regarding the medical field or a symptom to be asked, and according to the obtained knowledge level, an appropriate answer document regarding a case or a treatment method is used. It is possible to add a tag. When acquiring the knowledge level from the user, an input of a numerical value or the like corresponding to the knowledge level may be received on the same screen as in FIG. 11A or FIG. 11B, or it may be a general patient or a medical staff. Such information as the knowledge level may be received. For the tag adjustment document, for example, when the knowledge level is high, a medical book or a paper in the medical field may be selected, and when the knowledge level is low, a document related to general knowledge may be selected.

  In addition, some or all of the components included in the tagging device according to the embodiment may be configured by one system LSI (Large Scale Integration).

  The system LSI is an ultra-multifunctional LSI manufactured by integrating a plurality of components on one chip. Specifically, a microprocessor, a ROM (Read Only Memory), a RAM (Random Access Memory), etc. It is a computer system comprised including. A computer program is stored in the ROM. The system LSI achieves its functions by the microprocessor operating according to the computer program.

  Although the system LSI is used here, it may be called IC, LSI, super LSI, or ultra LSI depending on the degree of integration. Further, the method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible. An FPGA (Field Programmable Gate Array) that can be programmed after manufacturing the LSI, or a reconfigurable processor that can reconfigure the connection and setting of circuit cells inside the LSI may be used.

  Further, if integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology or a derivative other technology, it is naturally also possible to carry out function block integration using this technology. Biotechnology can be applied.

  In addition, the present invention can be realized not only as a tag providing device including such a characteristic processing unit, but also as a tag providing method using the characteristic processing unit included in the tag adding device as a step. You can also It can also be realized as a computer program that causes a computer to execute the characteristic steps included in such a method. Needless to say, such a computer program can be distributed via a computer-readable non-transitory recording medium such as a CD-ROM or a communication network such as the Internet.

  A tagging method according to the present invention includes a tagging device for automatically tagging a document, a question answering system for searching for an answer to a question using a tag attached to the document, and managing the document by a tag attached to the document It can be applied to a document management system.

DESCRIPTION OF SYMBOLS 1 Control server 2 Tag grant server 3 Question answer server 4 Operation terminal 5 Answer database 6 Tag adjustment document database 7 Network 101 Input detection part 102 Process selection part 103 Control part 104,207,305,404 Communication part 201 Reference | standard acquisition part 202 Document Selection unit 203, 302 Morphological analysis unit 204 Vector generation unit 205 Tag automatic generation unit 206 Tag linking unit 208, 306 Morphological analyzer 301 Question input unit 303 Keyword extraction unit 304 Question search unit 401 Input detection unit 402 Control unit 403 Display unit

Claims (18)

A tagging method performed by a tagging device that grants a tag for searching for answers to an answer document that constitutes an answer to a question from a user,
Select a tag adjustment document for adjusting a topic to be given as a tag to the answer document according to a predetermined criterion related to a user who performs the search or the search target,
Extracting a topic representing the answer document and the selected tag adjustment document from the answer document and the selected tag adjustment document;
The extracted topic is given to the answer document as a predetermined reference tag.
Tagging method. The predetermined criterion is whether the user's proficiency level or knowledge level for the search target is high or low,
The tag providing method according to claim 1. The search target is a product, and the response document is at least a part of a document explaining the handling of the product.
The tag providing method according to claim 2. Selecting a document relating to general knowledge as the tag adjustment document for assigning a tag used for a search by a user having a low proficiency level or knowledge level with respect to the search target;
The tag providing method according to claim 3. The general knowledge document is at least one of a dictionary, textbook, news, blog, or short text posting service.
The tag providing method according to claim 4. Selecting a document relating to expertise as the tag adjustment document for assigning a tag used for a search by a user having a high proficiency level or knowledge level with respect to the search target;
The tag providing method according to claim 3. The document relating to the specialized knowledge is at least one of construction installation manual, instruction manual, academic technical book, in-house document, or sales information.
The tag providing method according to claim 6. The target of the search is a product, and the predetermined criterion is either information on a new function of the product or information on an old function.
The tag providing method according to claim 1. The answer document is at least a part of a document explaining the function of the product, and the tag adjustment document is at least a part of a document explaining a function of a new model or an old model from the product.
The tag providing method according to claim 8. The target of the search is a product, and the predetermined criterion is either information on a higher model of the product or information on a lower model.
The tag providing method according to claim 1. The answer document is at least a part of a document explaining the function of the product, and the tag adjustment document is at least a part of a document explaining a function of a higher model or a lower model than the product.
The tag providing method according to claim 10. The extraction of keywords to be given as the tag is performed by extracting features of each document using a topic model for automatically extracting the topics constituting the document.
The tag providing method according to claim 1. The topic model is at least one of LDA (Lent Dielectric Allocation), LSI (Lent Semantic Indexing), or PLSA (Probabilistic Lent Semantic Analysis).
The tag providing method according to claim 12. The assignment of the tag to the answer document is performed by selecting a tag group including the most answer documents among a plurality of tag groups obtained from the topic model.
The tag providing method according to claim 12. The tag adjustment document is selected based on the user's action history.
The tag providing method according to claim 1. A question search method performed by a question answering system that searches an answer document that matches a user's question from a document group to be searched,
Determining a predetermined criterion relating to the user or the search target of the search;
Based on the predetermined criteria, select a tag adjustment document for adjusting the topic to be given as a tag to the answer document,
Extracting a topic representing the answer document and the selected tag adjustment document from the answer document and the selected tag adjustment document;
The extracted topic is given to the answer document as a predetermined reference tag,
A keyword is extracted from the user's question, and the answer document to which a tag matching the keyword is assigned is searched;
Question answer search method. A tag assignment device for attaching a tag for searching for an answer to an answer document constituting an answer to a question from a user,
A document selection means for selecting a tag adjustment document for adjusting a topic to be given as a tag to the answer document according to a predetermined criterion relating to the user who performs the search or the search target;
Topic extracting means for extracting a topic representing the answer document and the selected tag adjustment document from the answer document and the selected tag adjustment document;
Tag giving means for giving the extracted topic to the answer document as a predetermined reference tag;
A tagging device comprising: A program for causing a computer to function as a tag assigning device for assigning a tag for searching for an answer to an answer document constituting an answer to a question from a user,
In the computer,
Select a tag adjustment document for adjusting a topic to be given as a tag to the answer document according to a predetermined criterion related to a user who performs the search or the search target,
Extracting a topic representing the answer document and the selected tag adjustment document from the answer document and the selected tag adjustment document;
The extracted topic is given to the answer document as a predetermined reference tag.
A program that lets you do that.

Images