KR102540944B1 - Digital content system supporting document management using meta data and integrated search based on artificial intelligent - Google Patents

Abstract

The present invention relates to an artificial intelligence-based document management and integrated search support digital content system using metadata, a document information analysis unit that analyzes documents collected from an external system to generate keyword information and a summary, and a document information analysis unit that analyzes documents collected from an external system. An index information generation unit that generates index information for search by combining metadata, abstracts, and keywords of the document, and provides search results by combining search information included in a user's search request and comparing index information and similarity The present invention relates to a digital content system having an intelligent search unit and an interface providing unit for providing an interface for providing at least one service among searching, job classification recommendation, and legal information provision to an external system.

Description

AI-based document management and integrated search support digital content system using metadata

The present invention relates to a digital content system that supports artificial intelligence-based document management and integrated search using metadata, and collects documents and related metadata generated and managed in several independently operated information systems and builds them through deep learning analysis. It relates to a digital content system that supports integrated search and information sharing at work through search information.

Electronic documents managed by government agencies exist independently of a system that produces documents and an information system that transfers produced documents and manages them at each record level. The method composed of these individual systems is inconvenient for the user to check the status of the document according to the transfer stage of the document and connect to the individual system to search, and it is difficult to apply the searched document to the business system and utilize it. Have.

In order to solve the above problems, the present invention manages status information about documents managed in an independent system for each stage as metadata, and indexes information, It supports integrated search by constructing keywords and summary information. In addition, the purpose is to provide an artificial intelligence-based document management and integrated search support digital content system using metadata that provides business classification information related to search keywords and related laws and regulations.

An artificial intelligence-based document management and integrated search support digital content system using metadata according to an embodiment of the present invention includes a document information analysis unit that analyzes documents collected from an external system to generate keyword information and a summary; an index information generation unit generating index information for search by combining the metadata of the document collected from the external system, the summary, and the keyword; an intelligent search unit that combines search information included in a user's search request, compares a similarity with the index information, and provides a search result; and an interface providing unit that provides an interface for providing at least one service of search, work classification recommendation, and provision of legal information to the external system.

Also, the external system may include at least one of a system for producing work-related documents, a system for transferring and recording/storing the documents, and a system for managing classification information for the work.

In addition, the metadata may include at least one or more of document information, linked work information, and document status information.

The interface providing unit may include: a search interface unit providing an interface capable of searching for at least one of a specific document and business information associated with the document to the external system; and an information change interface unit that receives at least one state change information among the creation, change, and transfer of documents made in the external system.

In addition, the search interface unit may support natural language search words and provide an interface including at least one of an interworking module for search, a search API, and a search screen.

In addition, the intelligent search unit may include a search word processing unit analyzing the search word according to the form of the search word included in the search request; a document search unit generating at least one search query and calculating a similarity of results extracted through the search query; and a result providing unit providing search results in a manner determined according to the degree of similarity.

In addition, the document information analysis unit may include a text extraction unit for extracting text to be analyzed according to the format of the document; a summary sentence extraction unit configured to divide the text into sentence units, determine sentence importance, and extract summary sentences; a summary sentence generating unit generating a summary sentence according to the context; and a keyword extraction unit extracting keywords from the summary.

In addition, the digital content system may include a business classification recommendation unit for recommending business classification information to be included in the document; It may further include a business information providing unit having a; and a legal information providing unit for providing legal information related to the document.

In addition, the job classification recommendation unit may be characterized in that it recommends the job classification information related to the keyword of the analyzed document or the search information input by the user by applying a deep learning algorithm.

In addition, the statutory information providing unit may be characterized in providing the statutory information associated with keywords of the analyzed document or search information input by the user by applying a deep learning algorithm.

By managing the document information managed by the system at each stage from document production to document record management, integrated search can be performed regardless of document status.

It is possible to search not only the contents of the document but also work-related information related to the document, thereby increasing work productivity.

By providing various interfaces for document sharing and integrated search, it is possible to facilitate interworking with existing information systems, and thus, AI-based document integrated search function can be applied to existing information systems.

1 is a relationship diagram of an artificial intelligence-based document management and integrated search support digital content system using metadata according to an embodiment of the present invention.
2 is a block diagram of a digital content system in an artificial intelligence-based document management and integrated search support digital content system using metadata according to an embodiment of the present invention.
3 is a step-by-step performance process of a digital content system that supports document management and integrated search based on artificial intelligence using metadata according to an embodiment of the present invention.

Hereinafter, specific embodiments of the present invention will be described in detail with reference to the drawings. However, the spirit of the present invention is not limited to the presented embodiments, and those skilled in the art who understand the spirit of the present invention may add, change, delete, etc. other elements within the scope of the same spirit, through other degenerative inventions or the present invention. Other embodiments included within the scope of the inventive idea can be easily proposed, but it will also be said to be included within the scope of the inventive concept.

In addition, the terms to be described later are terms set in consideration of functions in the present invention, which may vary according to the intention or custom of the inventor, so the definitions should be made based on the contents throughout this specification, and in this specification related to the present invention If it is determined that a detailed description of a well-known configuration or function may obscure the gist of the present invention, a detailed description thereof will be omitted.

Hereinafter, an artificial intelligence-based document management and integrated search support digital content system 200 using metadata according to the present invention will be described with reference to the drawings.

1 is a relationship diagram of a digital content system (hereinafter referred to as a digital content system) supporting document management and integrated search based on artificial intelligence using metadata according to an embodiment of the present invention.

The digital content system 200 according to an embodiment of the present invention is connected to the business management system 100, the information system 300, and the user terminal 400 through a network to communicate with each other.

The online network referred to in the present invention may be a core network integrated with a wired public network, a wireless mobile communication network, or a mobile Internet, etc. , HTTPS (Hyper Text Transfer Protocol Secure), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System), SMTP (Simple Mail Transfer Protocol), etc. It comprehensively means a data communication network capable of transmitting and receiving data in various forms without being limited thereto.

The information system 100 according to an embodiment of the present invention is an external system that works with digital content documents through a network, and may be a system that collects documents and processes data to calculate processed information.

For example, the information system 100 may be a business management system for efficiently managing resources and processes necessary for the entire process of business processing within an institution in order to efficiently perform business of an administrative institution.

Also, the information system 100 may be an on-nara document system or an electronic payment system that produces electronic documents.

In addition, the information system 100 may be a record management system and a permanent record management system for recording and storing documents.

However, it is not limited to these examples and may mean various types of information systems comprehensively.

The system mentioned in the present invention may be constructed as a server serving at least one of a web server, a data analysis server, an application server, a database, and a mobile server.

Also, the system may be a cloud server configured in a cloud environment.

In addition, the processed result can be displayed on a web page through an online network or necessary input data can be transmitted through a web page, where the web page performs a specific task such as a web application in addition to simple text, image, sound, and video. It should be interpreted as including software for doing so, and may also be constructed to provide applications and interfaces installed on desktops, laptops, smart phones, tablet PCs, and the like. However, it is not limited to this, and the type of system can be variously changed at a level obvious to those skilled in the art.

The user terminal 300 according to an embodiment of the present invention should be interpreted as including various communication means such as a desktop, tablet, laptop, smartphone, wearable smart device, and the information system 100 and digital contents Various functions provided by at least one of the systems 200 may be executed through web-based or separate software/applications.

Meanwhile, the database (DB) referred to in the present invention may include a physical storage space in the digital content system 200 or may be configured as a separate and independent database server externally and connected to and communicated on a network.

A DB can be composed of one or a plurality of DBs within one server.

For example, an AI analysis DB that manages keywords extracted by artificial intelligence analysis through collected document information, summary information, attached document information, business classification information, and related statute information, and a metadata DB that manages metadata about documents. and a search DB including index information for document search.

However, it is not limited to this, and the configuration of the DB can be variously changed at a level obvious to those skilled in the art.

The document of the present invention may be an electronic document prepared or managed within an institution, and may include, for example, reports, papers, payment documents, memos, press releases, evaluation reports, and the like.

In addition, the electronic document may have a file format such as odt, pdf, ppt, pptx, xls, xlsx, doc, docx, hwp, and hwpx, and may be stored in a storage including a database.

However, it is not limited to this, and the type and format of the document can be variously modified at a level obvious to those skilled in the art.

2 is a configuration diagram of a digital content system 200 according to an embodiment of the present invention.

Hereinafter, the digital content system 200 will be described in detail with reference to FIG. 2 .

The digital content system 200 according to an embodiment of the present invention calculates an analysis model through a document information collection unit 210 for collecting documents and metadata, the collected document information through deep learning, and keywords through the analysis model. and a document information analysis unit 220 that extracts a summary, an index information generation unit 230 that generates or updates index information for search by combining metadata, keywords, and summary information of documents, and natural language through the built index information. An intelligent search unit 240 that supports based document search, a business information providing unit 250 that provides job classification recommendation and statutory information related to documents, and an interface that provides an interface so that an external system can search integrated documents and business information. A provision unit 260 may be provided.

The document information collection unit 210 collects documents and metadata from external systems.

The document information collection unit 210 may collect documents and metadata from at least one external system, use an interworking API provided by an external system, FTP, HTTP protocol-based communication module, TCP/IP-based Socket communication module. Through this, information can be collected in a way agreed upon in advance with an external system.

However, it is not limited to this, and the collection method can be variously modified at a level obvious to those skilled in the art.

The document information collection unit 210 may determine a document newly created in each external system or a document whose metadata of an already existing document is changed as a collection target.

The metadata may include at least one or more of document information, linked business information, and status information.

For example, document information may include a document title, document size, file format, attached document information, format information, creator, creation time, storage location, file name, and the like.

Linked task information may include task classification, department in charge, schedule, linked memo information, sender, receiver, sharer, document viewing authority, and the like.

The document status information may include the processing date and time and details of the work progress stage and creation/change/transfer of the document.

The metadata may be in an xml or html format including a hierarchical structure, and after receiving the metadata, data items and corresponding data values may be extracted, converted into a table form of a database, stored, and managed.

The document information analysis unit 220 extracts text from the collected documents and extracts keywords and summaries through deep learning analysis. In this case, the text extraction target may include body contents and attached documents.

The document information analysis unit 220 includes a text extraction unit 221 that extracts text according to the format of the document, and a summary sentence extraction unit 222 that extracts a summary sentence by determining the importance of the sentence after separating the extracted text into sentence units. , It may include a summary sentence generation unit 223 that generates a summary sentence according to the context of the extracted summary sentence and a keyword extraction unit 224 that extracts keywords from the text.

The text extraction unit 221 analyzes the structure of the document and automatically recognizes the format of the document. In this case, the document format may mean a document structure such as a cover page, a title, a table of contents, a body, and a table.

In addition, in the case of a document that follows a predefined document format, it is possible to extract only the text of some areas containing the main contents such as the table of contents, summary, and results.

To classify document types, document types such as result reports, evaluation reports, utilization reports, papers, and press releases can be defined in advance, and document types can be classified using regular expressions, classification models through TextCNN and BERT.

The extracted text is tokenized in sentence units and stored as data in the form of sentence bundles for easy analysis.

At this time, in the sentence unit tokenization, sentences may be separated or connected by determining the last morpheme after analyzing the morpheme through the morpheme analyzer.

In addition, tokenized sentences can be managed in the form of sentences with attached morphemes calculated through morpheme analysis.

In addition, special characters, stop words, unused object names, etc. may be removed, and different removal rules may be applied for each document type for text analysis accuracy.

The summary sentence extractor 222 attaches importance tags for each sentence using a designated analysis model to the sentence unit bundle data extracted from the text extractor 221, and extracts a predefined number of sentences in order of highest importance. do.

In this case, the designated analysis model may include a plurality of models such as TextRank and SBERT, and a model with the highest accuracy among the models may be designated.

In addition, when the evaluation accuracy of the sentence importance analysis model is higher than or lower than a predetermined threshold, automatic model relearning and evaluation may be performed.

In addition, some data used when determining the importance of sentences may be stored as sentences for verification for re-learning.

The summary sentence generator 223 generates a summary sentence based on the summary sentence extracted by the summary sentence extractor 222 .

The purpose of the summary sentence generating unit 223 is to generate the extracted summary sentence into a concise and context-natural summary sentence.

At this time, the designated generation model is GPT3 (Generative Pre-trained Transformer 3), BART (Bidirectional and Auto-Regressive Transformers), T5, PEGASUS (Pre-training with Extracted Gap-sentences for Abstractive Summarization), LED (Longformer Encoder-Decoder) ), etc., and a model with the highest accuracy among the models may be designated.

The keyword extractor 224 checks morpheme tags in the text extracted by the text extractor 221 or removes postpositions, endings, and the like.

In this case, morpheme tags to be removed, such as a particle to be removed and a suffix to be removed, may be defined in advance.

The keyword extraction unit 224 analyzes semantic similarity with the text for each word by applying a keyword extraction model to the words from which the search and suffixes have been removed, and attaches a similarity tag.

Thereafter, a number of words defined in the dictionary are extracted in order of high similarity and stored as core keywords.

In this case, the keyword extraction model may include a plurality of models such as TextRank and SBERT, and a model with the highest accuracy among the models may be designated.

In addition, when the evaluation accuracy of the keyword extraction model is higher than or lower than a predetermined threshold, automatic model relearning and evaluation may be performed.

In addition, some data used when extracting keywords can be stored as verification data for re-learning, and can be used for the purpose of learning similar words and new words.

The index information generation unit 230 generates, stores, and manages metadata of collected documents, summaries extracted by analyzing document contents, and keyword information as index information for search.

For example, the index information may be constructed in an inverted index structure in which keywords and document information including the keywords are connected and managed.

In addition, the number of keywords included in the document, the length of the document, and the weight for keywords can be managed.

In addition, metadata used as search conditions and document information can be linked and managed.

The index information generation unit 230 may update index information by receiving information collected by the document information collection unit 210 or document state change information collected by the information change interface unit 262 .

The intelligent search unit 240 combines the search information included in the user's search request, compares the index information of the document with a similarity, and provides a search result.

The intelligent search unit 240 analyzes the search word according to the form of the search word included in the search information and extracts the search word processing unit 241, generates a search query by combining the search word, and results extracted through the search query. It includes a document search unit 242 that calculates the similarity of and a result providing unit 243 that provides search results in a method determined according to the similarity.

The digital content system 200 according to an embodiment of the present invention may support natural language search and regular expression search words.

In this case, characters and patterns used in the regular expression may be defined in advance and guided to the user.

The search information may be formed in the form of a pair of at least one item name and a search word, and the search word may be composed of a sequence of at least one word or a sentence.

The search word processing unit 241 separates the search information received from the user from an item and a search word, and determines whether the search word is a regular expression.

If it is a regular expression, it is divided into token units according to the defined regular expression.

If it is not a regular expression, it is recognized as a natural language and separated into token units.

At this time, standard, nori, whitespace, pattern, and lowercase tokenizer can be used to separate token units.

In addition, it is possible to carry out the process of removing suffixes and suffixes through morpheme analysis.

The document search unit 242 generates a plurality of search queries by combining the separated tokens.

Here, a plurality of queries may be generated by applying synonyms of separated tokens.

The document search unit 242 may apply weights to a plurality of queries and calculate a similarity of documents derived as a search result of each query.

Indicators of weight evaluation can be the number of occurrences of tokens in a document according to the length of the document and the similarity of synonyms.

When a query is applied to the search term, the scope of a document search may be reduced by including a metadata item of a document included as a search condition.

For example, if items provided as metadata of documents, such as work stage, approval stage, person in charge, department in charge, creation date, document format, document transfer status, etc. are included in the search conditions, search terms for documents that satisfy the above conditions query can be applied.

The result providing unit 243 provides a search result according to the similarity of documents in the document search result extracted by the document search unit 242 in a predetermined manner.

Here, the predetermined method may arrange search results in descending order, and make and provide predefined search result items to be provided as search results as one data set.

Here, the data set may include a database table capable of containing search results, a resource object, an xml file, and the like.

The search result may include document information, keyword information, summary information, related work-related information, document status information, and the like.

The business information providing unit 250 provides business information related to documents, and includes a business classification recommendation unit 251 that recommends business classification information of documents and a legal information providing unit 252 that provides legal information related to documents. do.

The business classification recommendation unit 251 recommends a job classification associated with a keyword of a target document or search information input by a user.

The task classification recommendation unit 251 may apply a task classification recommendation model to which a deep learning algorithm is applied for task classification recommendation.

As an embodiment of the present invention, the task classification predicted by the task classification model may be a unit task included in the government functional classification system.

The governmental function classification system classifies tasks that the government regularly performs into policy areas, policy areas, large functions, medium functions, small functions, and unit tasks according to the functional level.

The task classification recommendation unit 251 may interwork with the government function classification system to collect government function classification system information.

However, it is not limited to this, and the business classification system and the collection target can be variously changed according to the general application field of the present invention.

The task classification recommendation model is a model built by performing deep learning learning and evaluation by applying the keywords of the document and the task classification information included in the document.

Here, TextCNN, GRU (Gated Recurrent Unit), BERT, and BiLSTM (Bidirectional Long Short-Term Memory) can be used as deep learning algorithms applied to the task classification recommendation model. can apply the best model.

The legal information provider 252 provides legal information related to keywords of target documents or search information input by a user.

Here, a statutory information provision model applied with a deep learning algorithm can be applied to provide related statutory information.

As an embodiment of the present invention, the statutory information predicted by the statutory information provision model may be laws, presidential decrees, ministerial ordinances, administrative rules, and self-governing laws included in the national statutory system.

The legal information providing unit 252 may be interlocked with the national legal information system to collect legal system information.

The statutory information provision model is a model built by performing deep learning learning and evaluation by applying the keywords of documents, business classification information that includes relevant documents, relevant agencies or departments, and related laws and regulations.

Here, TextCNN, GRU, BERT, and BiLSTM can be used as the deep learning algorithm applied to the statutory information provision model, and the model with the best prediction result can be applied through performance evaluation of the model to which each algorithm is applied.

The interface providing unit 260 provides an interface for providing at least one service of search, work classification recommendation, and provision of legal information to an external system.

The interface providing unit 260 includes a search interface unit 261 that provides an external system with an interface capable of searching for at least one of a specific document and business information associated with the document, and creates, changes, and transfers documents made in the external system. and a document state interface unit 262 for receiving at least one state change information among

The search interface unit 261 provides an interface that supports search in various ways by providing an interworking module for search, a search API, and a search screen according to the purpose of using the external system, thereby improving the usability and performance of the digital content system 200. accessibility can be increased.

For example, when an external system has a screen interface through which a user can access and search, a search function can be implemented by applying a search linkage module to the screen of the corresponding external system.

At this time, the search interworking module may be provided in the form of a library.

As another example, an external system may request a search using a search API provided by the search interface unit 261 and receive the search result in order to utilize the search function of the digital content system 200 according to the present invention.

At this time, the API may be based on a communication protocol such as HTTP-based SOAP or RESTful method.

As another example, the search interface unit 261 may provide a search screen according to a request from an external system. If necessary, when an external system requests a search screen in a defined manner, the search interface unit 261 may provide a link to the search screen by checking authentication and session information for the corresponding system.

The information change interface unit 262 receives event information about a document to be managed, such as creation, change, transfer, etc. of a document made in an external system, includes it as a collection target, or updates metadata of the document.

The information change interface unit 262 may be provided in the form of an API, and a request/response may be made with an external system in a predefined manner.

The information change interface unit 262 may update the metadata for the event request in real time and/or periodically.

3 is a process performed by each task of the digital content system 200 supporting artificial intelligence-based document management and integrated search using metadata according to an embodiment of the present invention.

Hereinafter, the process of performing the digital content system 200 for each business step will be described in detail with reference to FIG. 3 .

The user can access the task management system, one of the external systems, and create documents for the institution's task performance. At this time, necessary information and related documents may be searched using the search function of the digital content system 200 .

When a user inputs a search word through the search interface of the digital content system 200 provided as a business management system, the digital content system 200 searches the built search database and provides search results.

At this time, the search result selected by the user can be separately stored and applied as deep learning training data to increase the accuracy of the model.

Users can perform schedule registration, meetings, data sharing, etc. to perform tasks, and can utilize the search function at each work stage.

In addition, by inputting the subject or core keyword of the document, job classification information of the document can be automatically recommended and applied to the document.

The completed document may be registered in a specific information system 100 such as a record management system so that it can be shared with other people in charge. At this time, the digital content system 200 can collect documents and metadata related to the documents. there is. Document collection and metadata collection can be performed periodically after receiving an event for document generation through the information change interface unit 262 and storing it therein, or in real time when an event is received.

A summary sentence and keywords are extracted from the collected documents and metadata through the document information analysis unit 220, and index information for search is created based on these.

If document-related information such as transfer of person in charge, document transfer, business status change, document disposal, etc. is changed after document registration, the external system transmits an event for the change information through the information change interface unit 262, and the digital content system ( 200) updates metadata and index information by reflecting the changed information.

The digital content system 200 shares documents generated by one or more external systems, enables integrated search, monitors the status of documents transferred to other systems after creation, and manages them in an integrated manner, thereby providing a work process process. can provide the effect of improving work efficiency by seamlessly linking information and maximizing information sharing and utilization.

In the above, the configuration and characteristics of the present invention have been described based on the embodiments according to the present invention, but the present invention is not limited thereto, and various changes or modifications can be made within the spirit and scope of the present invention. It is ordered by those skilled in the art, and therefore such changes or modifications are included within the scope of the appended claims.

100: information system
200: Digital content system supporting document management and integrated search
210: document information collection unit
220: document information analysis unit
221: text extraction unit
222: summary sentence extraction unit
223: summary statement generation unit
224: keyword extraction unit
230: index information generation unit
240: intelligent search unit
241: search term processing unit
242: document search unit
243: result providing unit
250: business information provision unit
251: business classification recommendation unit
252: Legal information provision department
260: interface providing unit
261: search interface unit
262: information change interface unit
300: user terminal

Claims (10)

a document information analysis unit that analyzes documents collected from an external system and generates keyword information and a summary;
an index information generation unit generating index information for search by combining the metadata of the document collected from the external system, the summary, and the keyword;
an intelligent search unit that combines search information included in a user's search request, compares a similarity with the index information, and provides a search result; and
An interface providing unit providing an interface for providing at least one service of search, work classification recommendation, and provision of legal information to the external system;
The external system,
At least one of a system for producing work-related documents, including at least one of an electronic payment system and a business management system, and a system for transferring and recording/storing the documents;
The metadata,
Includes document information, linked business information and document status information;
The linked task information includes at least one of task classification, department in charge, schedule, linked memo information, sender, receiver, sharer, and document viewing authority,
The document status information includes at least one of a task progress stage, document creation information, document change information, and document transfer information,
The interface providing unit,
a search interface unit providing an interface for searching a specific document and business information associated with the document to the external system; and
An information change interface unit for receiving at least one document status change information among the creation, change, and transfer of documents in the external system and updating metadata of the document;
The search interface unit,
Supporting natural language search terms and providing an interface including at least one of an interlocking module for search, a search API, and a search screen,
digital content system.
delete delete delete delete The method of claim 1,
The intelligent search unit,
a search word processing unit that analyzes the search word according to the form of the search word included in the search request;
a document search unit generating at least one search query and calculating a similarity of results extracted through the search query; and
Including, a result providing unit providing search results in a predetermined manner according to the degree of similarity;
digital content system.
The method of claim 1,
The document information analysis unit,
a text extraction unit extracting text to be analyzed according to the format of the document;
a summary sentence extraction unit configured to divide the text into sentence units, determine sentence importance, and extract summary sentences;
a summary sentence generation unit that generates the summary sentence according to the context; and
A keyword extraction unit for extracting keywords from the summary; including,
digital content system.
The method of claim 1,
The digital content system,
a business classification recommendation unit for recommending business classification information to include the document; and
Having a legal information providing unit for providing legal information related to the document;
Further comprising a business information provider,
digital content system.
The method of claim 8,
The business classification recommendation unit,
Characterized in that by applying a deep learning algorithm, the job classification information related to the keyword of the analyzed document or the search information input by the user is recommended.
digital content system.
The method of claim 8,
The statutory information provider,
Characterized in that by applying a deep learning algorithm, the keyword of the analyzed document or the legal information related to the search information entered by the user is provided.
digital content system.

Images