JP2021114167A - Document management/viewing system and annotation text display method thereof - Google Patents

Abstract

To provide a document management/viewing system that can display appropriate annotation text for a term when the term included in a document is selected and that can reduce labor required for the creation of the annotation text.SOLUTION: A document management/viewing system 10 includes: a first storage unit 11 that stores a document; and a control unit 13 that performs a process of displaying the document stored in the first storage unit 11 to a user, and for a term selected by the user in the document displayed to the user, performs a process of displaying annotation text corresponding to the term to the user. As the annotation text, text generated by artificial intelligence is used.SELECTED DRAWING: Figure 1

Description

The present invention relates to a document management / viewing system, and more particularly to the display of annotations that support the understanding of terms contained in a document managed by the document management / viewing system.

There is a document management / viewing system that accumulates and manages various documents, presents them by displaying or outputting documents searched or specified by the user, and makes them available for viewing by the user. The user searches for the required document and displays the searched document on, for example, a display screen of a computer. There are contracts and contracts as documents to be managed by the document management / viewing system, but it is difficult to understand these documents unless they are familiar with the terms and laws used in the contract. Similarly, there are manuals and regulation collections as documents to be managed by the document management / viewing system, but these documents are enormous in volume and it is difficult to immediately grasp the contents of the description. In order to properly understand a document managed by a document management / viewing system, it is necessary to know the meaning and definition of terms contained in the document, but that is actually difficult and the viewer of the document Whenever a term that you want to know in detail or a term that you do not understand is encountered, you need to search using that term as a key. For example, in the case of performing counter-service operations using a document management / viewing system, a document is searched based on an inquiry from the customer, and the content of the searched document is communicated to the customer. When a question arises about a term contained in a document, it is necessary to search within the document or another document containing the term using that term as a key, and only for the time required for another search. Business efficiency will decrease. For proper understanding of a document, it is required to be able to present commentary sentences explaining the terms contained in the document to the readers of the document in a timely manner.

As a way to timely present commentary on a term contained in a document to the viewer of the document, the document itself is constructed as a hypertext document, and a file containing the commentary for that term is linked to the document for each term. The method of embedding is known. According to this method, if a term is clicked with a mouse or the like while the document is being displayed on a display device of a computer, for example, a browser or the like is started and an annotation sentence for the term is displayed. Further, Patent Document 1 is a document management / browsing system, and when a term in a displayed document is selected by a mouse operation or the like, the selected term is used as a key in the document management / browsing system. We disclose a system that can display search results by searching documents and searching on the Internet.

Japanese Unexamined Patent Publication No. 2006-79366

The method described in Patent Document 1 performs a search using a term included in the document and selected by the user as a key, but guarantees that a search result useful for understanding the term is always obtained. is not it. On the other hand, when a document is a hypertext document and a link to a file of annotations about the terms is embedded in the terms contained in the document, it is necessary to create annotations for each term in advance and make it a file. It takes a lot of effort to create. In particular, when the amount of documents stored in the document management / viewing system is large, it takes a lot of work time to create the annotation text, which makes it unrealistic to execute it manually.

An object of the present invention is a document management / viewing system and an annotation text display thereof that can display an appropriate annotation text for a term when a term included in the document is selected and can reduce the labor required for creating the annotation text. To provide a method.

In order to achieve the above object, the document management / viewing system of the present invention
In a document management / viewing system that accumulates and manages documents and presents them to users
A first storage unit for storing the documents and
A control unit that executes a process of displaying documents stored in the first storage unit to the user, and a control unit.
With
With respect to the term selected by the user in the document displayed to the user, a commentary sentence corresponding to the term is displayed to the user.
The commentary is characterized by being generated by artificial intelligence.

In order to achieve the above object, the annotation text display method of the present invention is:
In the annotation text display method in the document management / viewing system that accumulates and manages documents and presents them to users
The steps to display the documents stored in the document management / viewing system to the user,
For the term selected by the user in the document displayed to the user, a step of displaying an annotation sentence corresponding to the term to the user, and
Have,
The commentary is characterized by being generated by artificial intelligence.

In the present invention configured as described above, when a user selects a term in the displayed document, an annotation sentence generated by artificial intelligence is displayed with respect to the selected term. By using artificial intelligence, it is possible to display accurate annotations to the terms in the document to the user without human intervention, and it is possible to reduce the labor required to create the annotations. ..

In the present invention, annotation sentences for each term are accumulated in the document management / viewing system in advance, and based on the terms selected by the user, annotations to be displayed to the user from among the annotation sentences accumulated in advance. You may choose a statement. By accumulating the annotation text in the document management / viewing system in advance, when the user selects a term in the document, the annotation text for that term can be immediately displayed to the user.

When accumulating the annotation sentences for each term in advance, it is preferable to execute the process of generating the annotation sentences for each term in the document management / viewing system. By generating and accumulating annotations in the document management / viewing system, it becomes possible to automatically generate annotations when a new document is added to the document management / viewing system. The commentary is, for example, a process of extracting a term from a document stored in the document management / viewing system and a search for related documents related to the extracted term from at least the documents stored in the document management / viewing system. It can be generated by the process of acquiring the document and the process of generating a summary sentence of the related document using artificial intelligence and accumulating it as an annotation sentence in association with the corresponding term.

In the acquisition of related documents, not only the related documents may be acquired from the documents stored in the document management / viewing system, but also the related documents may be searched and acquired from an external server connected via a network such as the Internet. .. By acquiring the related document from the external server, it is possible to generate an annotation even if an appropriate related document cannot be acquired from the document stored in the document management / viewing system. However, since there is a high possibility that the related documents acquired from the external server include those equivalent to search noise, the related documents acquired from the document management / viewing system should be used with priority when generating the summary sentence. Is preferable. In addition, when a certain number of related documents are acquired, the similarity with the term is calculated for each related document, and a summary sentence is generated from a predetermined number of related documents whose similarity exceeds the threshold value. It is preferable to do so. By selecting related documents based on the degree of similarity, it is possible to obtain annotations with more appropriate contents. In addition, by aggregating the information scattered in a plurality of related documents, it is possible to obtain an annotation sentence including more appropriate information at once.

According to the present invention, when a term included in a document is selected, an appropriate commentary sentence for the term can be displayed, and the labor required for creating the commentary sentence can be reduced.

It is a block diagram which shows the structure of the document management / viewing system of one Embodiment of this invention. It is a block diagram which shows the structure of the annotation sentence generation part. It is a flowchart which shows the process of generating an annotation sentence. It is a figure which shows the selection of the term in a document, and the display of an annotation sentence.

Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a document management / viewing system according to an embodiment of the present invention. The document management / viewing system 10 is composed of, for example, a computer for a server, and accumulates and manages documents generated in a business entity or the like, and displays or outputs a document searched or specified by a user. Presented by the user for viewing.

One or more terminals 50 used by the user are connected to the document management / viewing system 10 via a network 30 such as a local area network (LAN) or a virtual private network (VPN). As the terminal 50, a personal computer or the like provided with a graphical user interface (GUI) is used, and the document displayed to the user by the document management / viewing system 10 is displayed on the display device of the terminal 50. The document management / viewing system 10 corresponds to a first storage unit 11 that stores documents to be managed by the document management / viewing system 10 and terms included in the documents stored in the first storage unit 11. It includes a second storage unit 12 for accumulating annotation texts, a comment text generation unit 20 for generating annotation texts, and a control unit 13. The commentary sentence stored in the second storage unit 12 is a sentence that explains each term contained in the document stored in the first storage unit 11 and supports the understanding of the term. Yes, it was generated using artificial intelligence (AI). Regarding terms that are commonly included in a plurality of documents stored in the first storage unit 11, it is not necessary to store duplicate annotations in the second storage unit 12, and documents containing the same terms do not need to be stored. Regardless of the number, one commentary sentence may be stored in the second storage unit 12. In the second storage unit 12, commentary sentences for each term are stored in association with the term, and by searching the second storage unit 12 by the term, the commentary sentence corresponding to the term can be obtained.

The control unit 13 is a part of the document management / viewing system 10 that serves as an interface with the user's terminal 50 via the network 30, and searches for documents in the first storage unit 11 based on a request from the user. The process of displaying the searched document on the display screen of the user's terminal 50 is executed. Further, when the user selects a term included in the document on the screen of the terminal 50 in the document displayed on the screen of the user's terminal 50, the control unit 13 selects the term from the terminal 50. The process of displaying the commentary sentence corresponding to the notified term on the screen of the user's terminal 50 is also executed. In the process of displaying the annotation text on the screen of the terminal 50, for example, the control unit 13 selects the annotation text to be displayed on the terminal 50 by searching in the second storage unit 12 based on the notified term. This is done by sending the selected annotation text to the terminal 50. The function of executing these processes of the control unit 13 can be realized by a computer program executed in the document management / viewing system 10 configured by the server computer.

The terminal 50 displays a document or commentary sent from the document management / viewing system 10 on the screen, and when a term included in the document displayed on the screen is selected by the user, the term is displayed. It has a function to send to the document management / viewing system. The function of such a terminal 50 is executed by general browser software or a dedicated application. Of the document management / viewing system 10, the function of the control unit 13 can also be realized in the terminal 50.

Next, the generation of the annotation text in this embodiment will be described. As described above, the commentary is generated by using artificial intelligence with respect to the terms contained in the document already stored in the first storage unit 11 or the document to be stored in the first storage unit 11. be. The procedure for generating the commentary may change depending on whether the commentary of the terms contained in the document already stored in the first storage unit 11 is generated or the commentary of the terms contained in the document to be stored in the future is generated. Since there is no such thing, here we will explain the case of generating annotations about terms contained in the documents to be accumulated from now on. FIG. 2 shows the configuration of the annotation sentence generation unit 20. In the following description, the document to be stored in the first storage unit 11 is referred to as document A. The process of generating an annotation sentence from the terms included in the document A and the process of accumulating the document A in the first storage unit 11 can be performed in parallel.

The commentary sentence generation unit 20 is at least first a term extraction unit 21 that extracts terms from the document A to be accumulated in the first storage unit 11 and a related document related to the terms extracted by the term extraction unit 21. It includes a related document acquisition unit 22 that searches and acquires from the documents stored in the storage unit 11, and a summary sentence generation unit 23 that generates a summary sentence of the related document by using artificial intelligence. The summary sentence generation unit 23 stores the generated summary sentence as an annotation sentence in the second storage unit 12 in association with the corresponding term. Here, the related document is a document that may include explanations and explanations about the target term. In many cases, the relevant document also includes the term itself in question.

Next, the process of generating an annotation sentence will be described with reference to FIG. First, in step 101, the term extraction unit 21 extracts a term for which an annotation sentence is to be generated from the document A to be stored in the first storage unit 11. When document A is a Japanese document, the terms for which commentary sentences should be generated are nouns including proper nouns, phrases with some modifiers added to the nouns, and some nouns connected by auxiliary words. It is thought that the words and phrases that were used are the main ones. Therefore, in order to extract a proper noun, a proper noun extraction software such as "Apache (registered trademark) OpenNLP" is applied to the document A, and the proper noun is extracted as a term from the document A. In parallel with this, for example, an algorithm based on TF-IDF is applied to the document A, and important words and phrases are extracted from the document A and used as terms. TF-IDF is a method of evaluating the importance of a word from the frequency of occurrence of the word and the frequency of reverse documents. By the way, the name of the law can be mentioned as a term that should be extracted from the document, but the name of the law includes "Act on the Use of Numbers for Identifying Specific Individuals in Administrative Procedures" and "Victims of Specific Emergency Disasters". Some particles cannot be extracted by ordinary morphological analysis because they contain a large number of particles, such as the "Act on Special Measures for the Protection of Rights and Interests". Therefore, in term extraction, in order to extract the legal name as a term when the document may contain the legal name, proper noun extraction by colocation extraction or search by the public legal database is used. It is also preferable. Since it is not necessary to generate the annotation sentences corresponding to the same term in duplicate, the term extraction unit 21 searches the second storage unit 12 as necessary, and the terms for which the annotation sentences have already been generated are searched. It is preferable not to extract terms again.

Subsequently, in step 102, the related document acquisition unit 22 searches for and acquires the related document from the documents stored in the first storage unit 11 using the term extracted in step 101 as a search key. A large number of documents containing the terms extracted in step 101 may be searched, and in that case, a document suitable for generating a summary sentence used as an annotation sentence is set as a related document. Since it is appropriate that the related document explains the target term, it is possible to narrow down the related document by examining the relationship between the reference and the reference to the target term. Alternatively, after obtaining a distributed representation of the term using "FastText" software or the like, the similarity of the document to the term is calculated by using a method such as cosine similarity to the document that is a candidate for the related document. Related documents may be selected based on similarity. You may use the "Doc2Vec" software to select relevant documents based on their similarity to the term. When narrowing down related documents based on the similarity to terms, basically, the documents with high similarity are regarded as related documents. By selecting related documents based on the degree of similarity, it is possible to obtain annotations with more appropriate contents. When more than a certain number of documents that are candidates for related documents are acquired, a summary sentence is generated from those related documents with a predetermined number of documents whose similarity exceeds the threshold value as related documents. Can be done. In this case, since the information scattered in a plurality of related documents is aggregated, it is possible to obtain an annotation sentence including more appropriate information at once. The threshold value and the predetermined number are predetermined according to the calculation method of the similarity and how many related documents the summary sentence is generated from.

In the above description, the related document acquisition unit 22 acquires the related document from the documents stored in the first storage unit 11, but it is also possible to expand the search range of the related document. As shown by the broken line in FIG. 2, when the external server 61 is connected to the external network 60 such as the Internet, the related document acquisition unit 22 acquires the related document from the document stored in the external server 61. May be good. By acquiring the related document from the external server 61, it is possible to generate an annotation even if an appropriate related document cannot be acquired from the document stored in the first storage unit 11.

Subsequently, in step 103, the summary sentence generation unit 23 generates a summary sentence from the related document acquired by the related document acquisition unit 22 using artificial intelligence. The abstract is generated using, for example, "LexRank" software. When there is a related document acquired from the first storage unit 11 and a related document acquired from the external server 61, the related document acquired from the first storage unit 11 is prioritized to generate a summary sentence. This is because there is a high possibility that the related documents acquired from the external server 61 include those corresponding to the search noise, and there is a possibility that some of them lack accuracy. After that, in step 104, the summary sentence generation unit 23 stores the generated summary sentence as an annotation sentence in the second storage unit 12 in association with the corresponding term.

Although the generation of the annotation text in the present embodiment has been described above, the software of "Apache (registered trademark) OpenNLP", "FastText", "Doc2Vec", and "LexRank" is open source software and should be easily implemented. Can be done. In addition, software that realizes the TF-IDF method and software that calculates cosine similarity and the like are also available as open source software. In the present embodiment, the comment statement generation unit 20 can be realized by software executed on the document management / viewing system 10 configured by the server computer.

Next, the selection of terms in the document and the display of the annotation text on the user's terminal 50 in the present embodiment will be described with reference to FIG. Assuming that any of the documents stored in the first storage unit 11 of the document management / viewing system 10 is displayed on the display screen of the terminal 50, FIG. 4 shows a state in which the document is displayed on the display screen. Shown. The document is displayed in the document display window 70 in the display screen. Assuming that the document is, for example, a hypertext document or a PDF (portable document format) document in which a character code is embedded, the user can select a character string included in the displayed document by using a mouse or the like. Can be done. If the display screen is a touch panel, the character string can be selected by placing a finger on the first character of the target character string and dragging the range of the character string. In the present embodiment, the character string selected on the display screen is the term selected by the user in the displayed document. In FIG. 4, the character string 71 of the “personal information protection policy” surrounded by a square is the selected term. It is preferable that the selected character string 71 is highlighted by changing the display color or the background color of the portion of the character string 71. If you select a string in your browser or PDF document, that string is usually highlighted. Further, in the displayed document, if the same character string as the selected character string exists in a place other than the selected place, the same character string in those places may be highlighted.

Then, when the user calls the context menu by right-clicking the mouse, for example, with the term selected and selects "display commentary" in the context menu, the selected term is transmitted to the control unit 13 for control. An annotation sentence corresponding to the term is sent from the unit 13 to the terminal 50, and the terminal 50 displays a balloon 72 corresponding to the character string 71 selected on the display screen as shown in FIG. Inside the balloon 72, a commentary for the selected term (here, "personal information protection policy") is displayed. As a method of displaying the annotation text, in addition to the method of using a balloon, the method of using a pop-up window, the method of using a new tab or another page, the method of using another column, the method of using a new or another window, and the like. "Display annotation text" is added to the context menu, and when "Display annotation text" is selected, the data of the selected term is sent to the control unit 13, and the annotation text sent from the control unit 13 is sent. Can be displayed in some form by a dedicated application, and even when a general browser is used, it can be realized by the setting in the browser and the setting of the operating system (OS) of the terminal 50.

Further, in the document stored in the first storage unit 11, JavaScript (registered trademark) or the like is used to embed annotation information for each term in the document, and when the document is displayed, the user. When the mouse cursor or the like is placed on a term in the document, the annotation text stored in the second storage unit 12 and the annotation text corresponding to the annotation section information is displayed. Is also possible. In this case, by storing the commentary for each term in the second storage unit 12, the data size of the document can be reduced, and the commentary for the term needs to be updated flexibly. It will be possible to respond.

In order to display the comment text, it is also possible to embed a link to the comment text stored in the second storage unit 12 in the document. In this case, by clicking a term in the document displayed on the terminal 50 with a mouse, for example, an annotation sentence corresponding to the term is displayed on the display screen of the terminal 50. In this case, since the term can be selected by clicking the term with the mouse, it is not necessary to specify the range of the character string to be the term on the screen for selecting the term. However, when embedding a link to the commentary in the document, for example, a document containing the wording of the law name "Act on Special Measures for Preserving the Rights and Interests of Victims of Specified Emergency Disasters" is displayed. Whether the commentary that the user wants is for the "specific emergency disaster" itself, or the law "Act on Special Measures to Preserve the Rights and Interests of Victims of the Specific Emergency Disaster" There is a demerit that it is difficult to distinguish between the two.

According to the document management / viewing system 10 of the present embodiment, since the commentary sentence is generated by using artificial intelligence, it is possible to create an appropriate commentary sentence about the term with little effort. In particular, the amount of work can be minimized by automatically generating the annotation text in parallel with the accumulation of the documents in the first storage unit 10. According to the present embodiment, a user who is browsing a document using the document management / browsing system 10 can immediately display an annotation sentence explaining terms in the document on the screen. Therefore, according to this document management / viewing system 10, it is possible to support the user to accurately understand the document. For example, when the document management / viewing system 10 of the present embodiment is used for window handling work, it is possible to reduce the time required for searching to find out the meaning of terms contained in the document, and to respond quickly to inquiries from customers. It becomes possible to do so, and business efficiency is greatly improved.

10 Document management / viewing system 11 1st storage unit 12 2nd storage unit 13 Control unit 20 Annotation sentence generation unit 21 Term extraction unit 22 Related document acquisition unit 23 Summary sentence generation unit 30 Network 50 Terminal 60 External network 61 External server

Claims (10)

In a document management / viewing system that accumulates and manages documents and presents them to users
A first storage unit for storing the documents and
A control unit that executes a process of displaying documents stored in the first storage unit to the user, and a control unit.
With
With respect to the term selected by the user in the document displayed to the user, a commentary sentence corresponding to the term is displayed to the user.
A document management / storage system characterized in that the commentary is generated by artificial intelligence. It is provided with a second storage unit that stores the commentary for each term in advance.
The control unit selects the commentary text from the second storage unit based on the term selected by the user, and executes a process of displaying the selected commentary text to the user. Document management / storage system according to 1. It is provided with a commentary sentence generation unit that generates the commentary sentence for each term.
The commentary sentence generation unit
A term extraction unit that extracts terms from documents stored in the first storage unit, and a term extraction unit.
A related document acquisition unit that searches and acquires related documents related to the term extracted by the term extraction unit from at least the documents stored in the first storage unit, and a related document acquisition unit.
An artificial intelligence is used to generate a summary sentence of the related document, and as the commentary sentence, a summary sentence generation unit that is associated with the corresponding term and is stored in the second storage unit, and a summary sentence generation unit.
The document management / storage system according to claim 2. The related document acquisition unit searches for and acquires the related document from an external server connected via a network, and obtains the related document.
The document management according to claim 3, wherein the summary sentence generation unit generates the summary sentence by giving priority to the related document acquired from the first storage unit over the related document acquired from the external server. Storage system. The related document acquisition unit calculates the degree of similarity between the acquired related document and the term.
The document management / viewing system according to claim 3, wherein the summary sentence generation unit generates the summary sentence from a plurality of related documents whose similarity exceeds a threshold value. In the annotation text display method in the document management / viewing system that accumulates and manages documents and presents them to users
The steps to display the documents stored in the document management / viewing system to the user,
For the term selected by the user in the document displayed to the user, a step of displaying an annotation sentence corresponding to the term to the user, and
Have,
A method of displaying an annotation text, characterized in that the annotation text is generated by artificial intelligence. The commentary sentence for each term is accumulated in the document management / viewing system in advance, and the commentary sentence to be displayed to the user from the accumulated commentary sentence based on the terminology selected by the user. The commentary text display method according to claim 6, wherein a text is selected. A comment statement generation step for generating the commentary sentence for each term is provided.
The comment statement generation step is
A term extraction step for extracting terms from documents stored in the document management / viewing system, and
A related document acquisition step of searching and acquiring at least related documents related to the term extracted in the term extraction step from the documents stored in the document management / viewing system, and a related document acquisition step.
A summary sentence generation step of generating a summary sentence of the related document using artificial intelligence, using the summary sentence as the commentary sentence, associating it with a corresponding term, and accumulating it in the document management / viewing system.
The commentary text display method according to claim 7. In the related document acquisition step, the related document is also searched and acquired from an external server connected via the network.
The annotation text display according to claim 8, wherein in the summary text generation step, the summary text is generated by giving priority to the related document acquired from the document management / viewing system over the related document acquired from the external server. Method. In the related document acquisition step, the degree of similarity between the acquired related document and the term is calculated.
The commentary text display method according to claim 8, wherein in the summary text generation step, the summary text is generated from a plurality of related documents whose similarity exceeds a threshold value.

Images