Document-based intelligent question and answer method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a document-based intelligent question and answer method and device, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the following steps: performing document analysis on a target electronic document to obtain document attributes and document contents; based on the document attribute and the document content, document segmentation is conducted on the target electronic document, document segmentation units and unit position information of the document segmentation units are obtained, and the document segmentation units comprise segmentation unit identifiers; based on the document attribute, performing network construction on the target electronic document to obtain a document structure tree network; on the basis of the unit position information, the document segmentation units are arranged in sequence, and a document linear list is obtained; performing document query on the document structure tree network or the document linear list based on the document problem information and the segmentation unit identifier to obtain document query information; and based on the document query information, performing question reply on the document question information. According to the embodiment of the invention, the intelligent question and answer accuracy based on the document can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a document-based intelligent question-answering method and device, an electronic device, and a storage medium. Background Art

[0002] At present, the document-based intelligent question answering method parses the target document input by the user to obtain the document information, and then answers the user's questions about the target document based on the document information. However, in the process of parsing the target document, the target document is often converted into a string, which makes it easy to lose the contextual semantic information of the target document when parsing the target document, thereby reducing the accuracy of intelligent question answering. Therefore, how to improve the accuracy of document-based intelligent question answering has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a document-based intelligent question and answer method and device, an electronic device and a storage medium, aiming to improve the accuracy of document-based intelligent question and answer.

[0004] To achieve the above objectives, a first aspect of an embodiment of the present application proposes a document-based intelligent question answering method, the method comprising:

[0005] Obtain the target electronic document and obtain document problem information;

[0006] Parsing the target electronic document to obtain document attributes and document content;

[0007] Based on the document attribute and the document content, the target electronic document is segmented to obtain a document segmentation unit and unit position information of the document segmentation unit, wherein the document segmentation unit includes a segmentation unit identifier;

[0008] Based on the document attributes, a network is constructed for the target electronic document to obtain a document structure tree network;

[0009] Based on the unit position information, the document segmentation units are arranged in sequence to obtain a document linear list;

[0010] Based on the document question information and the segmentation unit identifier, a document query is performed on the document structure tree network or the document linear list to obtain document query information;

[0011] Based on the document query information, answer questions about the document question information.

[0012] In some embodiments, the step of performing a document query on the document structure tree network or the document linear list based on the document problem information and the segmentation unit identifier to obtain document query information includes:

[0013] Performing semantic understanding on the document problem information to obtain text semantic information;

[0014] Calculating the richness of the text semantic information to obtain semantic richness;

[0015] If the semantic richness is less than a preset richness threshold, keyword extraction is performed on the document question information based on the text semantic information to obtain search keywords;

[0016] Based on the search keyword, perform document search on the document linear list to obtain the document query information;

[0017] If the semantic richness is greater than or equal to the richness threshold, a document query is performed on the document structure tree network based on the text semantic information and the segmentation unit identifier to obtain the document query information.

[0018] In some embodiments, if the semantic richness is greater than or equal to the richness threshold, a document query is performed on the document structure tree network based on the text semantic information and the segmentation unit identifier to obtain the document query information, including:

[0019] When the semantic richness is greater than or equal to the richness threshold, based on the text semantic information and the segmentation unit identifier, a retrieval language is constructed on a pre-constructed retrieval language database to obtain a semantic retrieval instruction;

[0020] Based on the semantic search instruction, a document search is performed on the document structure tree network to obtain the document query information.

[0021] In some embodiments, the step of segmenting the target electronic document based on the document attribute and the document content to obtain document segmentation units and unit position information of the document segmentation units includes:

[0022] Based on the document attributes and the document content, the target electronic document is structured to obtain a document unit structure;

[0023] Based on the document unit structure, performing element demarcation on the target electronic document to obtain a document segmentation boundary;

[0024] Based on the document segmentation boundary, the target electronic document is segmented to obtain initial segmentation units;

[0025] Based on the document unit structure, performing unit identification on the initial segmentation unit to obtain the document segmentation unit;

[0026] Based on the target electronic document, position identification is performed on the document segmentation unit to obtain the unit position information.

[0027] In some embodiments, the network construction of the target electronic document based on the document attribute and the document segmentation unit to obtain a document structure tree network includes:

[0028] Based on the document unit structure, determining a unit hierarchical relationship between each of the document segmentation units;

[0029] Based on the unit hierarchical relationship, the target electronic document is network constructed to obtain an initial structure tree network;

[0030] Based on the document segmentation unit, the initial structure tree network is filled with node information to obtain the document structure tree network.

[0031] In some embodiments, answering the document question information based on the document query information includes:

[0032] Performing text context understanding on the document query information to obtain query document semantic information;

[0033] Based on the query document semantic information, the document question information is answered and processed to obtain target answer information.

[0034] In some embodiments, the sequentially arranging the document segmentation units based on the unit position information to obtain a document linear list includes:

[0035] Based on the unit position information, determining an arrangement order between each of the document segmentation units;

[0036] Based on the arrangement order, the document segmentation units are stored to obtain the document linear list.

[0037] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes a document-based intelligent question-answering device, the device comprising:

[0038] An information acquisition module is used to acquire the target electronic document and obtain document problem information;

[0039] A document parsing module, used to parse the target electronic document to obtain document attributes and document content;

[0040] A document segmentation module, configured to segment the target electronic document based on the document attribute and the document content, and obtain document segmentation units and unit position information of the document segmentation units, wherein the document segmentation unit includes a segmentation unit identifier;

[0041] A network construction module, used to perform network construction on the target electronic document based on the document attributes to obtain a document structure tree network;

[0042] A document arrangement module, used to arrange the document segmentation units in sequence based on the unit position information to obtain a document linear list;

[0043] A document query module, configured to perform a document query on the document structure tree network or the document linear list based on the document problem information and the segmentation unit identifier to obtain document query information;

[0044] The question answering module is used to answer the document question information based on the document query information.

[0045] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0046] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0047] The document-based intelligent question answering method and device, electronic device and storage medium proposed in the present application obtain document attributes and document content by performing document parsing on the acquired target electronic document, and then performing document segmentation on the target electronic document according to the document attributes and document content to obtain document segmentation units including segmentation unit identifiers and unit position information of the document segmentation units. At the same time, a document structure tree network of the target electronic document is constructed according to the document attributes, which can effectively retain the contextual semantic information of the target electronic document. Secondly, according to the unit position information of the document segmentation units, the document segmentation units are arranged in sequence to obtain a document linear list. Finally, according to the document question information and the segmentation unit identifier, a document query is performed on the document structure tree network or the document linear list to obtain document query information, and the document question information is answered according to the document query information, which can improve the efficiency of keyword retrieval or semantic retrieval of the target electronic document based on the document question information during the intelligent question answering process, thereby improving the overall efficiency and effect of document-based intelligent question answering. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a flowchart of a document-based intelligent question-answering method provided in an embodiment of the present application;

[0049] Figure 2 yes Figure 1 Flow chart of step S103 in FIG.

[0050] Figure 3 yes Figure 1 Flow chart of step S104 in FIG.

[0051] Figure 4 yes Figure 1 Flow chart of step S105 in FIG.

[0052] Figure 5 yes Figure 1 Flow chart of step S106 in FIG.

[0053] Figure 6 yes Figure 5 Flow chart of step S504 in FIG.

[0054] Figure 7 yes Figure 5 Flow chart of step S107 in FIG.

[0055] Figure 8 is a structural diagram of a document-based intelligent question-answering device provided in an embodiment of the present application;

[0056] Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0060] First, some nouns involved in this application are analyzed:

[0061] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0062] Intelligent question-answering system: An intelligent question-answering system is a software application based on artificial intelligence technology. It uses natural language processing, machine learning and other related algorithms to understand the questions asked by users, and retrieve and analyze information from documents input by users to generate accurate, relevant and useful answers. Intelligent question-answering systems can process and understand the complexity of natural language, provide an interactive dialogue experience, and are widely used in many fields such as education and research, aiming to improve the efficiency and convenience of information acquisition.

[0063] At present, the document-based intelligent question answering method parses the target document input by the user to obtain the document information, and then answers the user's questions about the target document based on the document information. However, in the process of parsing the target document, the target document is often converted into a string, which makes it easy to lose the contextual semantic information of the target document when parsing the target document, thereby reducing the accuracy of intelligent question answering. Therefore, how to improve the accuracy of document-based intelligent question answering has become a technical problem that needs to be solved urgently.

[0064] Based on this, the embodiments of the present application provide a document-based intelligent question and answer method and device, an electronic device and a storage medium, aiming to improve the accuracy of document-based intelligent question and answer.

[0065] The document-based intelligent question-answering method and device, electronic device and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the document-based intelligent question-answering method in the embodiments of the present application is described.

[0066] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0067] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0068] The document-based intelligent question-answering method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The document-based intelligent question-answering method provided in the embodiment of the present application can be applied to a terminal, can also be applied to a server side, and can also be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements a document-based intelligent question-answering method, etc., but is not limited to the above forms.

[0069] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0070] Figure 1is an optional flow chart of a document-based intelligent question-answering method provided in an embodiment of the present application, which is applied to an intelligent question-answering system. Figure 1 The method may include but is not limited to steps S101 to S107.

[0071] Step S101, obtaining a target electronic document and obtaining document problem information;

[0072] Step S102, parsing the target electronic document to obtain document attributes and document content;

[0073] Step S103, based on the document attributes and the document content, the target electronic document is segmented to obtain document segmentation units and unit position information of the document segmentation units, wherein the document segmentation unit includes a segmentation unit identifier;

[0074] Step S104, based on the document attributes, a network is constructed for the target electronic document to obtain a document structure tree network;

[0075] Step S105, based on the unit position information, the document segmentation units are arranged in sequence to obtain a document linear list;

[0076] Step S106, based on the document problem information and the segmentation unit identifier, a document query is performed on the document structure tree network or the document linear list to obtain document query information;

[0077] Step S107: answer the document question information based on the document query information.

[0078] In steps S101 to S107 shown in the embodiment of the present application, the intelligent question system can obtain the target electronic document and parse the target electronic document to obtain document attributes and document content. Then, the target electronic document can be segmented using the document attributes and document content to form document segmentation units including segmentation unit identifiers, and the location of each document segmentation unit is recorded to obtain unit location information. At the same time, based on the document attributes, a document structure tree network of the target electronic document is constructed. Thereafter, the segmentation units are arranged in sequence according to the unit location information to generate a document linear list. Further, document question information of the target electronic document is obtained, and based on the document question information and the segmentation unit identifier of the document segmentation unit, a query is performed in the document structure tree network or the document linear list to obtain document query information. Finally, the document question information is answered according to the document query information to complete the intelligent question-answering process based on the target electronic document. Therefore, the present application parses the acquired target electronic document to obtain document attributes and document content, and then performs document segmentation on the target electronic document according to the document attributes and document content to obtain document segmentation units including segmentation unit identifiers and unit position information of the document segmentation units. At the same time, according to the document attributes, a document structure tree network of the target electronic document is constructed, which can effectively retain the contextual semantic information of the target electronic document. Secondly, according to the unit position information of the document segmentation units, the document segmentation units are arranged in sequence to obtain a document linear list. Finally, according to the document question information and the segmentation unit identifier, a document query is performed on the document structure tree network or the document linear list to obtain document query information, and according to the document query information, questions are answered for the document question information, which can improve the efficiency of keyword retrieval or semantic retrieval of the target electronic document based on the document question information in the intelligent question and answer process, thereby improving the overall efficiency and effect of document-based intelligent question and answer.

[0079] In step S101 of some embodiments, the target electronic document refers to the document uploaded by the user to the intelligent question-answering system, wherein the format of the target electronic document is not limited, for example, the target electronic document may be a PDF document, a Word document, a TXT document, etc. Document question information refers to the question raised by the user to the intelligent question-answering system, and it should be noted that the document question information needs to be associated with the target electronic document. For example, when the content of the target electronic document is to introduce the basic situation of the company, the document question information may be questions such as "what is the main content of the document", "what risks does the company disclose", "what are the company's business areas", etc.

[0080] In an embodiment of the present application, the intelligent question and answer system can generate a target electronic document by reading a document uploaded by a user, and then obtain document problem information by performing text parsing on the question text input by the user. For example, when a user uploads a company's information introduction PDF document to the intelligent question and answer system through the document upload function, and enters the question "What is the main content of the document" in the question input of the intelligent question and answer system, the intelligent question and answer system can generate a target electronic document based on the information introduction PDF document, and obtain document problem information by parsing the text "What is the main content of the document".

[0081] In step S102 of some embodiments, the document attributes refer to the date, title, author, category and other attributes of the target electronic document. It should be noted that when the target electronic document is a chapter-structured document, the document attributes also include attributes such as directory. The document content refers to the content information of the target electronic document, wherein the content information can be data and vector representation in any format such as text, table, image, etc.

[0082] In an embodiment of the present application, the intelligent question-answering system can obtain the date, title, author, category, directory and other information of the document by reading the metadata embedded in the target electronic document. It should be noted that before obtaining the target electronic document, the metadata such as the date, author, category and the target electronic document are already associated with the target electronic document, and the metadata such as the title and directory can be obtained through document parsing. Furthermore, the date, title, author, category, directory and other information are integrated to obtain the document attributes. Secondly, by performing element recognition on the target electronic document, the document elements such as text elements, table elements, image elements in the target electronic document are determined. Finally, the text content of the text elements is extracted by a text information extraction method to obtain the document text content, the table content of the table elements is extracted by a table information extraction method to obtain the document table content, and the image content of the image elements is extracted by an image information extraction method to obtain the document image content. It should be noted that the above-mentioned text information, table information and image information extraction methods can be based on PDF processing, image recognition, natural language processing and other parsing technologies. It should be noted that the information includes text and corresponding vectors.

[0083] In step S103 of some embodiments, the document segmentation unit refers to the part of the document information obtained by segmentation based on the element layout of the target electronic document. For example, when the element layout of the target electronic document is three paragraphs of text, two tables and three images, by segmenting the target electronic document, text 1, text 2, text 3, table 1, table 2, image 1, image 2 and image 3 can be obtained, and the above eight parts of document information can be called document segmentation units. Unit position information refers to the position coordinate information of the document segmentation unit in the target electronic document.

[0084] In an embodiment of the present application, the intelligent question-answering system can identify the element layout of the target electronic document through document attributes and document content, thereby realizing structural recognition of the target electronic document and obtaining a document unit structure. Furthermore, according to the document unit structure, the target electronic document is structurally split to obtain an initial segmentation unit, and then the initial segmentation unit is identified to obtain a document segmentation unit. Finally, by recording the coordinate information of each document segmentation unit in the target electronic document, the unit position information of the document segmentation unit can be obtained.

[0085] For details, see Figure 2 In some embodiments, step S103 may include but is not limited to steps S201 to S205:

[0086] Step S201, based on the document attributes and document content, the target electronic document is structured to obtain a document unit structure;

[0087] Step S202, based on the document unit structure, the target electronic document is divided into elements to obtain a document segmentation boundary;

[0088] Step S203, splitting the target electronic document based on the document segmentation boundary to obtain initial segmentation units;

[0089] Step S204, based on the document unit structure, the initial segmentation unit is marked to obtain the document segmentation unit;

[0090] Step S205 : Based on the target electronic document, position identification is performed on the document segmentation unit to obtain unit position information.

[0091] In step S201 of some embodiments, the intelligent question-answering system may determine the document category of the target electronic document based on the document attributes. When the document category indicates that the target electronic document is a document with a chapter structure, the hierarchical structure of the target electronic document is first determined based on the directory attributes in the document attributes, and then the document scope of each title in the target electronic document is determined based on the titles in the directory attributes. Furthermore, based on the document content, the elements within the document scope of each title are identified, the element layout within the document scope of each title is determined, and then the element layout within the document scope of each title is integrated to obtain the document unit structure of the target electronic document. When the document category indicates that the target electronic document is a document without a chapter structure, the elements of the target electronic document are identified through the text content to obtain the element layout of the target electronic document, thereby determining the element structure and document unit structure of the target electronic document.

[0092] In step S202 and step S203 of some embodiments, according to the document unit structure, the boundary of each element in the target electronic document is identified to determine the segmentation boundary of the target electronic document and obtain the document segmentation boundary. Further, according to the document segmentation boundary, the target electronic document is split into independent initial segmentation units.

[0093] In step S204 of some embodiments, according to the above-mentioned document unit structure, each initial segmentation unit in the target electronic document is marked as a unit to obtain a segmentation unit mark, and then the segmentation unit mark is embedded in the initial segmentation unit to form a document segmentation unit. For example, when the document unit structure of the target electronic document is two first-level titles and three second-level titles, by marking the first-level titles as units, the segmentation unit marks of the first-level titles 1 and 2 can be obtained, and by marking the second-level titles as units, the segmentation unit marks of the second-level titles 1, 2 and 3 can be obtained. Furthermore, by respectively embedding the first-level title 1, the first-level title 2, the second-level title 1, the second-level title 2 and the second-level title 3 into the relevant initial segmentation units, the document segmentation unit containing the segmentation unit mark can be obtained.

[0094] In step S205 of some embodiments, for each document segmentation unit, the intelligent question and answer system can determine the pixel coordinates of the document segmentation unit by identifying the positions of the four corner pixel points in the boundary of the document segmentation unit in the target electronic document. Furthermore, the pixel coordinates of the four corner pixel points are stored as the position information of the document segmentation unit to obtain the unit position information.

[0095] In steps S201 to S205 shown in this embodiment, the structure of the target electronic document is identified according to the document attributes and the document content to obtain the document unit structure, and then the element boundaries of the target electronic document are divided according to the document unit structure to obtain the document segmentation boundary, and the target electronic document is split according to the document segmentation boundary to obtain the initial segmentation unit, thereby realizing the accurate segmentation of the target electronic document. Furthermore, according to the document unit structure, the initial segmentation unit is marked as a unit to obtain the document segmentation unit, and then according to the target electronic document, the position of the document segmentation unit is identified to obtain the unit position information, so that the document segmentation unit retains the context information of the target electronic document, thereby ensuring that the retrieval accuracy can be improved when the target electronic document is retrieved.

[0096] In step S104 of some embodiments, the document structure tree network is a data network model used to represent the organization and hierarchical relationship of the target electronic document. It should be noted that the titles, paragraphs, tables, images, coordinates, etc. in the target electronic document and their corresponding vectors can be organized in the form of a tree diagram according to their chapter structure logic and reading order in the target electronic document to form a document structure tree network.

[0097] In an embodiment of the present application, the hierarchical relationship between each document segmentation unit in the target electronic document, that is, the unit hierarchical relationship, can be extracted based on the above-mentioned document unit structure, and then a tree network can be constructed based on the unit hierarchical relationship and the unit information of the document segmentation unit to obtain the document structure tree network of the target electronic document.

[0098] For details, see Figure 3 In some embodiments, step S104 may include but is not limited to steps S301 to S303:

[0099] Step S301, based on the document unit structure, determining the unit hierarchical relationship between each document segmentation unit;

[0100] Step S302, constructing a network for the target electronic document based on the unit hierarchical relationship to obtain an initial structure tree network;

[0101] Step S303: Based on the document segmentation unit, the initial structure tree network is filled with node information to obtain a document structure tree network.

[0102] In step S301 of some embodiments, the unit hierarchy relationship refers to the structural association relationship between document segmentation units. For example, if the second-level title 1 appears in the document scope of the first-level title 1 in the target electronic document, then the unit hierarchy relationship between the document segmentation units corresponding to the second-level title 1 and the first-level title 1 is a containment and being contained relationship.

[0103] In an embodiment of the present application, the association relationship between each document segmentation unit can be extracted from the document unit structure, and the hierarchical relationship between the document segmentation units can be analyzed based on the extracted association relationship to obtain the unit hierarchical relationship.

[0104] In step S302 and step S303 of some embodiments, the initial structure tree network refers to a data network model having a tree graph structure but with incomplete node information.

[0105] In an embodiment of the present application, after obtaining the unit hierarchical relationship, nodes can be set for the pre-constructed tree diagram according to the unit hierarchical relationship to determine the document segmentation unit corresponding to each node, thereby obtaining an initial structure tree network. Furthermore, the text content of the document segmentation unit is filled into the corresponding node in the initial structure tree network, so that the node information of the initial structure tree network is complete, thereby realizing the conversion of the initial structure tree network to the document structure tree network.

[0106] In steps S301 to S303 shown in the present embodiment, the unit hierarchical relationship between each document segmentation unit is determined according to the document unit structure, and then a network is constructed for the target electronic document according to the unit hierarchical relationship to obtain an initial structure tree network, which can enhance the organization and searchability of the document content. Furthermore, according to the document segmentation unit, the initial structure tree network is filled with node information to obtain a document structure tree network, which enriches the semantic information of the document structure tree network, so that the intelligent question and answer system can perform information retrieval and question answering more efficiently, thereby improving the efficiency and accuracy of intelligent question and answering.

[0107] In step S105 of some embodiments, the document linear list refers to a document storage model obtained by sorting and combining the target electronic documents by the document segmentation unit.

[0108] In the embodiment of the present application, the arrangement order of the document segmentation units in the target electronic document can be determined according to the unit position information, and the document segmentation units can be stored according to the arrangement order to obtain a document linear list.

[0109] For details, see Figure 4 In some embodiments, step S105 may include but is not limited to steps S401 to S402:

[0110] Step S401, determining the arrangement order between each document segmentation unit based on the unit position information;

[0111] Step S402: based on the arrangement order, the document segmentation units are stored to obtain a document linear list.

[0112] In step S401 of some embodiments, the arrangement order is an arrangement combination formed based on the spatial relationship of the coordinates of the document segmentation units in the target electronic document.

[0113] By extracting the coordinate data in the unit position information, the spatial ordering relationship between each document segmentation unit can be determined. Furthermore, by recording and saving the spatial ordering relationship, the arrangement order between each document segmentation unit can be obtained.

[0114] In step S402 of some embodiments, the document segmentation units are combined according to the above arrangement order to obtain a document linear list.

[0115] It should be noted that the document linear list is composed of document segmentation units, and the document segmentation unit contains the segmentation unit identifier, so the document linear list carries the segmentation unit identifier.

[0116] In steps S401 to S402 shown in this embodiment, the arrangement order between each document segmentation unit is determined according to the unit position information, and then the document segmentation units are stored according to the arrangement order to obtain a linear list of documents, which facilitates information retrieval by the intelligent question and answer system and improves the retrieval efficiency of document content.

[0117] It should be noted that in order to improve the efficiency and quality of document retrieval, the linear list of documents is obtained by parsing in advance. The present application can also obtain the linear list of documents by analyzing the information contained in the document structure tree network.

[0118] In step S106 of some embodiments, the document query information refers to the document content related to the document question information in the document structure tree network and the document linear list.

[0119] In an embodiment of the present application, the intelligent question-answering system can understand the text semantic information of the document question information to clarify the problem to be solved, and then retrieve and query the document linear list and document structure tree network based on the text semantic information, thereby obtaining document query information for solving the document question information.

[0120] For details, see Figure 5 In some embodiments, step S106 may include but is not limited to steps S501 to S505:

[0121] Step S501, semantic understanding of document question information to obtain text semantic information;

[0122] Step S502, calculating the richness of the text semantic information to obtain the semantic richness;

[0123] Step S503: if the semantic richness is less than a preset richness threshold, keyword extraction is performed on the document question information based on the text semantic information to obtain search keywords;

[0124] Step S504, based on the search keyword, perform document search on the document linear list to obtain document query information;

[0125] Step S505: if the semantic richness is greater than or equal to the richness threshold, a document query is performed on the document structure tree network based on the text semantic information and the segmentation unit identifier to obtain document query information.

[0126] In step S501 of some embodiments, text semantic information is used to represent the core meaning of the document question information. For example, when the document question information is “what does the document describe”, the text semantic information may be “what is the text content”.

[0127] In step S502 of some embodiments, semantic richness is used to characterize the diversity and depth of text semantic information.

[0128] In an embodiment of the present application, the richness of text semantic information can be calculated through methods such as interface selection, word frequency analysis, semantic similarity calculation, machine learning model or deep learning model, so as to obtain semantic richness.

[0129] In step S503 of some embodiments, the richness threshold is a pre-set condition for determining whether the semantic richness satisfies the conditions for semantic retrieval. The search keyword refers to the keyword used to describe the core problem in the document question information. For example, when the document question information is "what risks does the document disclose", the search keyword may be "risk". The search keyword may also be a keyword directly input by the user.

[0130] In an embodiment of the present application, when the semantic richness is less than a pre-set richness threshold, the intelligent question and answer system can determine the core issue mentioned in the document question information based on the text semantic information, and then search for relevant information from the document question information based on the core issue, so as to quickly locate the text content needed to solve the core issue.

[0131] In step S504 of some embodiments, keyword matching is performed between the search keyword and the phrase of each document segmentation unit in the document linear list, and the document segmentation unit with successful keyword matching is used as the document query information.

[0132] In step S505 of some embodiments, when the semantic richness is greater than or less than the above-mentioned richness threshold, a text retrieval function can be called according to the text semantic information and the above-mentioned segmentation unit identifier to perform a document query on the document structure tree network, thereby obtaining document query information.

[0133] For details, see Figure 6 In some embodiments, step S505 may include but is not limited to steps S601 to S602:

[0134] Step S601, when the semantic richness is greater than or equal to the richness threshold, based on the text semantic information and the segmentation unit identifier, a retrieval language is constructed for a pre-constructed retrieval language database to obtain a semantic retrieval instruction;

[0135] Step S602: Based on the semantic search instruction, a document search is performed on the document structure tree network to obtain document query information.

[0136] In step S601 of some embodiments, the retrieval language database is a database for storing pre-constructed retrieval instructions, wherein the retrieval instructions can be stored in the retrieval language database in the form of objects, attributes or functions. For example, the retrieval instructions in the form of objects can be title([N],[id]), paragraphs(id), etc., the retrieval instructions in the form of attributes can be id(), text(), etc., and the retrieval instructions in the form of functions can be abovetitles(N), etc.

[0137] It is important to note that the above title([N],[id]) indicates that the document retrieval object is a title, N indicates the level of the title, id in the retrieval instruction indicates the segmentation unit identifier of the title, paragraphs(id) indicates that the document retrieval object is a paragraph, id in the retrieval instruction indicates the segmentation unit identifier of the paragraph, id() indicates that the document retrieval return object is a segmentation unit identifier, text() indicates that the document retrieval return object is text information, abovetitles(N) indicates that the document retrieval return object is an array of all title objects within the specified level, and N indicates the level of the title.

[0138] When the semantic richness is greater than or equal to the pre-set richness threshold, it indicates that the semantic richness meets the conditions for semantic retrieval. Therefore, the document structure tree information to be retrieved can be clarified based on the text semantic information, and then the retrieval language query can be performed on the retrieval language database based on the document structure tree information to obtain the semantic retrieval instruction.

[0139] In step S602 of some embodiments, after obtaining the semantic search instruction, the semantic search instruction is used to perform a network query on the document structure tree network to obtain document content related to the text semantic information, that is, document query information.

[0140] In steps S601 to S602 shown in the present embodiment, when the semantic richness is greater than or equal to the richness threshold, a retrieval language is constructed for a pre-constructed retrieval language database based on the text semantic information and the segmentation unit identifier to obtain a semantic retrieval instruction. Furthermore, based on the semantic retrieval instruction, a document search is performed on the document structure tree network to obtain document query information, which can reduce the complexity of the semantic retrieval and thus improve the efficiency of the semantic retrieval.

[0141] In steps S501 to S505 illustrated in the present embodiment, semantic understanding is performed on the document question information to obtain text semantic information, and then the richness of the text semantic information is calculated to obtain semantic richness. When the semantic richness is less than a preset richness threshold, keyword extraction is performed on the document question information based on the text semantic information to obtain retrieval keywords. Based on the retrieval keywords, document retrieval is performed on a linear list of documents to obtain document query information. When the semantic richness is greater than or equal to the richness threshold, a document query is performed on a document structure tree network based on the text semantic information and segmentation unit identifiers to obtain document query information, so that keyword retrieval or semantic retrieval can be performed according to the semantic richness of the document question information during intelligent question answering, thereby improving the retrieval efficiency during intelligent question answering.

[0142] It should be noted that in the embodiments of the present application, keyword retrieval or semantic retrieval methods can be used alone to search for target electronic documents, or a combination of keyword retrieval and semantic retrieval can be used to search for target electronic documents. For example, the intelligent question-answering system can understand the text semantic information of the document problem information to clarify the problem to be solved, and then search and query the document linear list and document structure tree network based on the text semantic information, thereby obtaining document query information for solving the document problem information.

[0143] In detail, the embodiment of the present application can realize the combination of keyword search and semantic search through the following steps:

[0144] Perform semantic understanding on document question information to obtain text semantic information;

[0145] Based on the text semantic information, keyword extraction is performed on the document question information to obtain the search keywords;

[0146] Based on the search keywords, the document linear list is searched to obtain the key content of the document;

[0147] Based on the key content of the document and the identification of the segmentation unit, a document query is performed on the document structure tree network to obtain document query information.

[0148] In some embodiments, after clarifying the text semantic information of the document question information, the intelligent question answering system can determine the core issue mentioned in the document question information based on the text semantic information, and then search for keywords from the document question information based on the core issue to quickly locate the text content required to solve the core issue. Furthermore, by keyword matching the search keywords with the phrases of each document segmentation unit in the document linear list, and taking the document segmentation unit with successful keyword matching as the document key content, after obtaining the document key content, according to the segmentation unit identifier embedded in the document key content, the corresponding key content node can be found in the document structure tree network, and then the upper and lower nodes of the key content node can be determined through the document structure tree network to obtain the target content context. Finally, content extraction is performed on the target content context and the document key content to obtain the document query information.

[0149] In step S107 of some embodiments, the target reply information can be obtained by understanding the retrieved document query information, clarifying the query document semantic information of the document query information, and then replying to the document question information according to the query document semantic information.

[0150] For details, see Figure 7 In some embodiments, step S107 may include but is not limited to steps S701 to S702:

[0151] Step S701, understanding the text context of the document query information to obtain the query document semantic information;

[0152] Step S702, based on the query document semantic information, answer the document question information to obtain target answer information.

[0153] In step S701 of some embodiments, querying document semantic information refers to the core content of document query information.

[0154] In the embodiment of the present application, the method of performing text context understanding on the document query information to obtain the query document semantic information is similar to the above-mentioned method of performing semantic understanding on the document question information to obtain text semantic information, so it is not repeated here.

[0155] In step S702 of some embodiments, a document summary of the document query information is extracted based on the query document semantic information, and then the wording of the document summary is modified and adjusted based on the document question information, thereby generating target response information.

[0156] In steps S701 to S702 shown in this embodiment, the text context of the document query information is understood to obtain the query document semantic information, and then the document question information is replied to based on the query document semantic information. When processing the document question information, the intelligent question and answer system can not only identify the key points of the document question information, but also combine the deep meaning and context information of the target electronic document to generate accurate target reply information, thereby improving the accuracy of document-based intelligent question and answer.

[0157] The present application parses the acquired target electronic document to obtain document attributes and document content, and then performs document segmentation on the target electronic document according to the document attributes and document content to obtain document segmentation units including segmentation unit identifiers and unit position information of the document segmentation units. At the same time, a document structure tree network of the target electronic document is constructed according to the document attributes, which can effectively retain the contextual semantics and chapter structure information of the target electronic document. Secondly, according to the unit position information of the document segmentation units, the document segmentation units are arranged in sequence to obtain a document linear list, which can improve the intelligent question-answering process. Finally, according to the document question information and the segmentation unit identifier, a document query is performed on the document structure tree network or the document linear list to obtain document query information, and based on the document query information, a question is answered for the document question information, which can improve the efficiency of keyword retrieval or semantic retrieval of the target electronic document based on the document question information during the intelligent question-answering process, thereby improving the overall efficiency and effect of document-based intelligent question-answering.

[0158] See also Figure 8 The embodiment of the present application further provides a document-based intelligent question-answering device, which can implement the above-mentioned document-based intelligent question-answering method, and the device includes:

[0159] The information acquisition module 801 is used to acquire the target electronic document and obtain document problem information;

[0160] The document parsing module 802 is used to parse the target electronic document to obtain document attributes and document content;

[0161] The document segmentation module 803 is used to segment the target electronic document based on the document attributes and the document contents, and obtain the document segmentation units and the unit position information of the document segmentation units, wherein the document segmentation units include segmentation unit identifiers;

[0162] A network construction module 804 is used to construct a network for the target electronic document based on the document attributes to obtain a document structure tree network;

[0163] The document arrangement module 805 is used to arrange the document segmentation units in sequence based on the unit position information to obtain a document linear list;

[0164] The document query module 806 is used to perform a document query on the document structure tree network or the document linear list based on the document question information and the segmentation unit identifier to obtain document query information;

[0165] The question answering module 807 is used to answer the document question information based on the document query information.

[0166] The specific implementation of the document-based intelligent question-answering device is basically the same as the specific implementation of the above-mentioned document-based intelligent question-answering method, and will not be repeated here.

[0167] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above document-based intelligent question-answering method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0168] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0169] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0170] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the document-based intelligent question-answering method of the embodiment of this application;

[0171] Input / output interface 903, used to implement information input and output;

[0172] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0173] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0174] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0175] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the above-mentioned document-based intelligent question-answering method is implemented.

[0176] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0177] The document-based intelligent question-answering method, document-based intelligent question-answering device, electronic device and storage medium provided in the embodiments of the present application are applied to an intelligent question-answering system, which obtains a target electronic document and performs document parsing on the target electronic document to obtain document attributes and document content. Then, the target electronic document is segmented using the document attributes and document content to form a document segmentation unit including a segmentation unit identifier, and the position of each document segmentation unit is recorded to obtain unit position information. At the same time, a document structure tree network of the target electronic document is constructed based on the document attributes. Thereafter, the segmentation units are arranged in sequence according to the unit position information to generate a document linear list. Furthermore, document question information of the target electronic document is obtained, and a query is performed in the document structure tree network and the document linear list according to the document question information and the segmentation unit identifier of the document segmentation unit to obtain document query information. Finally, the document question information is replied according to the document query information to complete the intelligent question-answering process based on the target electronic document.

[0178] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0179] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0180] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0181] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0182] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0183] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0184] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0185] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0186] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0187] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0188] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A document-based intelligent question-answering method, characterized in that: The method comprises: Obtain the target electronic document and obtain document problem information; Parsing the target electronic document to obtain document attributes and document content; Based on the document attribute and the document content, the target electronic document is segmented to obtain a document segmentation unit and unit position information of the document segmentation unit, wherein the document segmentation unit includes a segmentation unit identifier; Based on the document attributes, a network is constructed for the target electronic document to obtain a document structure tree network; Based on the unit position information, the document segmentation units are arranged in sequence to obtain a document linear list; Based on the document question information and the segmentation unit identifier, a document query is performed on the document structure tree network or the document linear list to obtain document query information; Based on the document query information, answer questions about the document question information.

2. The method according to claim 1, characterized in that The step of performing a document query on the document structure tree network or the document linear list based on the document problem information and the segmentation unit identifier to obtain document query information includes: Performing semantic understanding on the document problem information to obtain text semantic information; Calculating the richness of the text semantic information to obtain semantic richness; If the semantic richness is less than a preset richness threshold, keyword extraction is performed on the document question information based on the text semantic information to obtain search keywords; Based on the search keyword, perform document search on the document linear list to obtain the document query information; If the semantic richness is greater than or equal to the richness threshold, a document query is performed on the document structure tree network based on the text semantic information and the segmentation unit identifier to obtain the document query information.

3. The method according to claim 2, characterized in that If the semantic richness is greater than or equal to the richness threshold, a document query is performed on the document structure tree network based on the text semantic information and the segmentation unit identifier to obtain the document query information, including: When the semantic richness is greater than or equal to the richness threshold, based on the text semantic information and the segmentation unit identifier, a retrieval language is constructed on a pre-constructed retrieval language database to obtain a semantic retrieval instruction; Based on the semantic search instruction, a document search is performed on the document structure tree network to obtain the document query information.

4. The method according to claim 1, characterized in that The step of segmenting the target electronic document based on the document attribute and the document content to obtain document segmentation units and unit position information of the document segmentation units includes: Based on the document attributes and the document content, the target electronic document is structured to obtain a document unit structure; Based on the document unit structure, performing element demarcation on the target electronic document to obtain a document segmentation boundary; Based on the document segmentation boundary, the target electronic document is segmented to obtain initial segmentation units; Based on the document unit structure, performing unit identification on the initial segmentation unit to obtain the document segmentation unit; Based on the target electronic document, position identification is performed on the document segmentation unit to obtain the unit position information.

5. The method according to claim 4, characterized in that The network construction of the target electronic document based on the document attribute and the document segmentation unit to obtain a document structure tree network includes: Based on the document unit structure, determining a unit hierarchical relationship between each of the document segmentation units; Based on the unit hierarchical relationship, the target electronic document is network constructed to obtain an initial structure tree network; Based on the document segmentation unit, the initial structure tree network is filled with node information to obtain the document structure tree network.

6. The method according to any one of claims 1 to 5, characterized in that: The step of answering the document question information based on the document query information includes: Performing text context understanding on the document query information to obtain query document semantic information; Based on the query document semantic information, the document question information is answered and processed to obtain target answer information.

7. The method according to any one of claims 1 to 5, characterized in that: The step of sequentially arranging the document segmentation units based on the unit position information to obtain a document linear list includes: Based on the unit position information, determining an arrangement order between each of the document segmentation units; Based on the arrangement order, the document segmentation units are stored to obtain the document linear list.

8. A document-based intelligent question-answering device, characterized in that: The device comprises: An information acquisition module is used to acquire the target electronic document and obtain document problem information; A document parsing module, used to parse the target electronic document to obtain document attributes and document content; A document segmentation module, configured to segment the target electronic document based on the document attribute and the document content, and obtain document segmentation units and unit position information of the document segmentation units, wherein the document segmentation unit includes a segmentation unit identifier; A network construction module, used to perform network construction on the target electronic document based on the document attributes to obtain a document structure tree network; A document arrangement module, used to arrange the document segmentation units in sequence based on the unit position information to obtain a document linear list; A document query module, configured to perform a document query on the document structure tree network or the document linear list based on the document problem information and the segmentation unit identifier to obtain document query information; The question answering module is used to answer the document question information based on the document query information.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the document-based intelligent question-answering method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the document-based intelligent question answering method according to any one of claims 1 to 7 is implemented.

Images