Method, apparatus and storage medium for generating auxiliary expansion ideas of a conference assistant

Abstract

The present invention discloses a method, apparatus, and storage medium for generating auxiliary expansion ideas of a meeting assistant. The method includes: obtaining a user's question; extracting entity relationships from the user's question to obtain an extraction result of the question; searching for the extraction result of the question in an information library; matching the extraction result of the question in a knowledge graph; comparing the search result of the question and the matching result of the question to determine whether there are entities in the matching result of the question that are not in the search result of the question. If so, perform sentence processing on the corresponding entities to generate an idea expansion result and feedback it to the user; if not, process the extraction result of the question based on the entity fusion degree to obtain new entities, perform sentence processing on the new entities to generate an idea expansion result and feedback it to the user, and at the same time add the new entities to the nodes of the entities corresponding to the extraction result of the question, thereby updating the knowledge graph. The present invention can collect information, integrate information, expand ideas, and innovate knowledge for the current discussion topic.

Description

Technical Field

[0001] The present invention relates to a method, device and storage medium for generating auxiliary expansion ideas of a conference assistant, belonging to the technical field. Background Art

[0002] The development of enterprises is inseparable from product R & D and business innovation. During the process of enterprise product R & D or business innovation, it is often necessary to hold meetings for discussion and analysis. How to innovate is one of the many important issues faced by enterprises in the development process.

[0003] To innovate based on existing business and products, it is necessary to first search and query existing business and technologies, integrate and analyze the collected information, and obtain inspiration from this, and then fuse to generate new ideas and new thoughts. Due to the limited energy, memory and experience of humans, when thinking and expanding around a certain topic to generate valuable and meaningful new ideas in a seminar, it is often limited by the energy, memory and experience of the participants in the seminar, resulting in the omission of some key and inspiring thinking directions, thus affecting the possibility of innovation. If a conference assistant is used to assist participants in collecting information, integrating information, and expanding ideas on the current discussion topic during the conference discussion process, it will greatly improve the effective utilization rate of information and help enterprises better carry out business and product innovation.

[0004] Currently popular and mature intelligent search and question - answering products, such as Xiaodu Voice Robot of Baidu, Xiaoai Voice Robot of Xiaomi, etc., can only support single tasks, such as turning on the TV, or simple question - answering searches, such as what is speech recognition technology. Such search - based question - answering intelligent assistants cannot perform the tasks of information integration and knowledge generation, and naturally cannot expand ideas and innovate knowledge based on existing information. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method, device and storage medium for generating auxiliary expansion ideas of a conference assistant, which can collect information, integrate information, expand ideas and innovate knowledge on the current discussion topic, improve the effective utilization rate of information, and help enterprises better carry out business and product innovation.

[0006] To achieve the above - mentioned purpose, the present invention is implemented by the following technical solutions:

[0007] In the first aspect, the present invention provides a method for generating auxiliary expansion ideas of a conference assistant, including:

[0008] Obtain a user's question sentence;

[0009] Extract entity relationships from the user's question sentence to obtain an extraction result of the question sentence;

[0010] Search the extraction results of the question sentence in the information library to obtain the question sentence search results;

[0011] Match the extraction results of the question sentence in the knowledge graph to obtain the question sentence matching results;

[0012] Compare the question sentence search results with the question sentence matching results to determine whether there are entities in the question sentence matching results that are not in the question sentence search results. If there are, perform sentence processing on the corresponding entities to generate idea expansion results and feedback them to the user;

[0013] If not, process the extraction results of the question sentence based on the entity fusion degree to obtain new entities, perform sentence processing on the new entities to generate idea expansion results and feedback them to the user, and at the same time add the new entities to the nodes of the entities corresponding to the extraction results of the question sentence, thereby updating the knowledge graph.

[0014] Optionally, the entity relationship extraction of the user question sentence to obtain the question sentence extraction results includes:

[0015] Construct an RLC network model for entity relationship extraction, and perform entity relationship extraction on the user question sentence through the RLC network model; wherein, the RLC network model includes an input layer, an embedding layer, an encoding layer, a classification layer and an output layer;

[0016] The input layer: receives sentence information and processes it using a Chinese word segmentation tool to obtain word and maximum forward matching word vectors;

[0017] The embedding layer: receives the word and word vectors flowing out of the input layer and uses an improved BERT network model to generate sequence text;

[0018] The encoding layer: uses an LSTM network model to obtain the language features of the sequence text and outputs them;

[0019] The classification layer: performs entity recognition and relationship recognition based on language features. The entity recognition includes using a CRF model to perform entity recognition on the language features and using Viterbi decoding to output the name labels of the entities; the relationship recognition includes the entity position information determined by the entity labels, inputting the text encoding information between entities into the relationship classification module based on the entity position information, and using a CNN model to perform relationship recognition and using Softmax decoding to output the type labels of the relationships;

[0020] The output layer: in the entity recognition, outputs the name labels of the entities; in the relationship recognition, outputs the type labels of the relationships.

[0021] Optionally, the improved BERT network model includes:

[0022] Delete the next sentence prediction task of the BERT network model; improve the masking strategy to dynamic masking; use a vocabulary encoded with double-byte encoding to encode Chinese language representations.

[0023] Optionally, the searching for the question extraction result in the information library to obtain the question search result includes:

[0024] Access and search the question extraction result in the information library Info through a crawler tool;

[0025] Obtain the title information of the first several search results, and perform entity relationship extraction on the title information to obtain the title extraction result

[0026] Perform deduplication processing on the title extraction result to obtain the question search result.

[0027] Optionally, the processing of the question extraction result based on the entity fusion degree to obtain a new entity includes:

[0028] Calculate the entity fusion degree between each entity in the knowledge graph and the entity corresponding to the question extraction result in sequence;

[0029] Select multiple entities with the largest entity fusion degree from the knowledge graph as search targets;

[0030] Select entities within the single-hop path range of the search target in the knowledge graph and with the same relationship type as the question extraction result as candidate entities;

[0031] Calculate the entity fusion degree between each entity within the single-hop path range of the search target in the knowledge graph and the candidate entities in sequence;

[0032] Select the entity with the largest entity fusion degree from the candidate entities as the new entity.

[0033] Optionally, the entity fusion degree Fusion is:

[0034] Fusion(c i ,c j )=α*Sim sem (c i ,c j )+β*Sim str (c i ,c j )

[0035] In the formula, Fusion(c i ,c j ) is the entity fusion degree between entity c i and entity c j , α and β are preset weight coefficients, and Sim sem (c i ,cj )、Sim str (c i ,c j ) is entity c i and entity c j 's semantic similarity and structural similarity;

[0036] The obtaining of the semantic similarity includes: using the pre-trained BERT network model to encode entity c i and entity c j to generate word vectors, and calculating the cosine similarity between the word vectors to obtain the semantic similarity;

[0037] The structural similarity is:

[0038]

[0039] In the formula, c anc is the common entity connected to both entity c i and entity c j , dis(c i ), dis(c j ) respectively represent the distances between entity c i and entity c j and entity c anc .

[0040] Optionally, the updating of the knowledge graph by the new entity includes:

[0041] Adding the new entity to the node of the entity corresponding to the question extraction result to generate a new triple "new entity - question extraction result corresponding relationship type - question extraction result corresponding entity".

[0042] In a second aspect, the present invention provides an auxiliary expansion idea generation device for a conference assistant, and the device includes:

[0043] A question acquisition module for acquiring user questions;

[0044] A question extraction module for performing entity relationship extraction on the user question to obtain a question extraction result;

[0045] A result search module for searching the question extraction result in the information library to obtain a question search result;

[0046] A result matching module for matching the question extraction result in the knowledge graph to obtain a question matching result;

[0047] A thought expansion module is used to compare the question search results with the question matching results, determine whether there are entities in the question matching results that are not in the question search results. If there are, the corresponding entities are processed into sentences to generate thought expansion results and fed back to the user;

[0048] If not, the question extraction results are processed based on the entity fusion degree to obtain new entities. The new entities are processed into sentences to generate thought expansion results and fed back to the user. At the same time, the new entities are added to the nodes of the entities corresponding to the question extraction results, thereby updating the knowledge graph.

[0049] In a third aspect, the present invention provides an auxiliary expansion thought generation device for a meeting assistant, including a processor and a storage medium;

[0050] The storage medium is used to store instructions;

[0051] The processor is used to operate according to the instructions to execute the steps of the above method.

[0052] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the steps of the above method are implemented.

[0053] Compared with the prior art, the beneficial effects achieved by the present invention:

[0054] The present invention provides an auxiliary expansion thought generation method, device and storage medium for a meeting assistant, which extracts entity relationships from user questions in a meeting, searches an information library and matches a knowledge graph according to the extraction results, compares the search results with the matching results. If there are entities in the matching results that are not in the search results, thought expansion is performed based on these entities. If there are no entities in the matching results that are not in the search results, new entities are generated according to the entity fusion degree, and thought expansion is performed based on the new entities. Generally, it realizes assisting participants in a meeting to collect information, integrate information, expand thoughts and innovate knowledge on the current discussion topic during the meeting process, greatly improves the effective utilization rate of information, and helps enterprises better carry out business and product innovation. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flowchart of an auxiliary expansion thought generation method for a meeting assistant provided in Embodiment 1 of the present invention;

[0056] Figure 2 is a schematic diagram of the principle of the RLC network model provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0058] Embodiment 1:

[0059] As Figure 1 shown, the embodiment of the present invention provides an auxiliary expansion idea generation method for a meeting assistant, including the following steps:

[0060] 1. Obtain the user question UQ; Example: "What are the innovative application directions of speech recognition technology"

[0061] 2. Perform entity relationship extraction on the user question UQ to obtain the question extraction result UQS; Example: "Speech recognition technology - application direction", where speech recognition technology is the entity and application method is the relationship;

[0062] Specifically, it includes:

[0063] Construct an RLC network model for entity relationship extraction, and perform entity relationship extraction on the user question through the RLC network model;

[0064] As Figure 2 shown, the RLC network model includes an input layer, an embedding layer, an encoding layer, a classification layer, and an output layer;

[0065] Input layer: Receive statement information, and use a Chinese word segmentation tool to process to obtain word and maximum forward matching word vectors;

[0066] Embedding layer: Receive the word and word vectors flowing out of the input layer, and use an improved BERT network model (denoted as RoBERTa) to generate sequence text;

[0067] Encoding layer: Use an LSTM network model to obtain the language features of the sequence text and output;

[0068] Classification layer: Based on the language features, perform entity recognition and relationship recognition. Entity recognition includes using a CRF model to perform entity recognition on the language features, and using Viterbi decoding to output the name labels of the entities; Relationship recognition includes the entity position information determined by the entity labels, inputting the text encoding information between entities into the relationship classification module based on the entity position information, and using a CNN model to perform relationship recognition, and using Softmax decoding to output the type labels of the relationships;

[0069] Output layer: In entity recognition, output the name labels of the entities; in relationship recognition, output the type labels of the relationships.

[0070] Among them, the improved BERT network model includes:

[0071] Delete the next sentence prediction task of the BERT network model (simplify the model structure and improve the computational efficiency); improve the masking strategy to dynamic masking (enable the model to gradually adapt to different masking strategies and contain more semantic representations); use a vocabulary encoded with double-byte encoding to encode Chinese language representations (improve the accuracy of Chinese vocabulary representations and the model running efficiency).

[0072] 3. Search for the question extraction result UQS in the information repository Info to obtain the question search result UQSI;

[0073] Specifically include:

[0074] Access the information repository Info through a crawler tool and search for the question extraction result UQS;

[0075] Obtain the title information of the top multiple search results, perform entity relationship extraction on the title information, and obtain the title extraction result

[0076] Perform duplicate removal processing on the title extraction result to obtain the search result UQSI.

[0077] Among them, the information repository Info refers to the existing knowledge information repository including major resource search websites, public knowledge bases, enterprise domain databases, etc. Example: Search for "Speech Recognition Technology - Application Directions" in the Baidu search engine, take the top 10 search results, perform entity relationship extraction and entity duplicate removal, and obtain the search results: "Speech Recognition Technology - Application Directions - Customer Service", "Speech Recognition Technology - Application Directions - Translation", "Speech Recognition Technology - Application Directions - Smart Home Appliances".

[0078] 4. Match the question extraction result UQS in the knowledge graph KG to obtain the question matching result UQSK. Examples: "Speech Recognition Technology - Application Directions - Customer Service", "Speech Recognition Technology - Application Directions - Translation".

[0079] 5. Compare the question search result UQSI and the question matching result UQSK, and determine whether there is an entity AS in the question matching result UQSK that does not exist in the question search result UQSI. If so, perform statement processing on the corresponding entity AS to generate an idea expansion result and feedback it to the user;

[0080] If not, process the question extraction result UQS based on the entity fusion degree to obtain a new entity UQSN, perform statement processing on the new entity UQSN to generate an idea expansion result and feedback it to the user, and at the same time add the new entity UQSN to the node of the entity corresponding to the question extraction result UQS, thereby updating the knowledge graph KG.

[0081] 5.1. Processing the question extraction result UQS based on the entity fusion degree to obtain a new entity UQSN includes:

[0082] (1) Calculate the entity fusion degree between each entity in the knowledge graph KG and the corresponding entity in the question extraction result UQS in sequence;

[0083] (2) Select multiple entities with the largest entity fusion degree from the knowledge graph KG as the search targets;

[0084] (3) Select entities within the single-hop path range of the search targets in the knowledge graph KG and with the same corresponding relationship type as the question extraction result UQS as candidate entities;

[0085] (4) Calculate the entity fusion degree between each entity within the single-hop path range of the search targets in the knowledge graph KG and the candidate entities in sequence;

[0086] (5) Select the entity with the largest entity fusion degree from the candidate entities as the new entity UQSN.

[0087] Among them, the entity fusion degree Fusion is:

[0088] Fusion(c i ,c j ) = α * Sim sem (c i ,c j ) + β * Sim str (c i ,c j )

[0089] In the formula, Fusion(c i ,c j ) is the entity fusion degree between entity c i and entity c j . α and β are preset weight coefficients (usually set as α = 0.7, β = 0.3). Sim sem (c i ,c j ) and Sim str (c i ,c j ) are the semantic similarity and structural similarity between entity c i and entity c j ;

[0090] The acquisition of semantic similarity includes: using the pre-trained BERT network model to encode entity c i and entity c j to generate word vectors, and calculating the cosine similarity between the word vectors to obtain the semantic similarity;

[0091] The structural similarity is:

[0092]

[0093] In the formula, c anc is the common entity connected to both entity c i and entity c j . dis(c i ), dis(c j ) respectively represent the distances between entity c i and entity c j and entity c anc .

[0094] Examples are as follows: In the current knowledge graph, there are entity-relationship-entity pairs "speech recognition technology - application direction - customer service"; "speech recognition technology - application direction - translation"; "speech recognition technology - application direction - smart home appliances"; "communication technology - application direction - remote meeting"; "communication technology - application direction - chatbot";

[0095] For the question entity relationship UQS "speech recognition technology - application direction", within the current knowledge graph, first select an entity with a relatively high degree of fusion with the entity "speech recognition technology", such as "communication technology". Then obtain the entities within the single-hop path range of the UQS entity "speech recognition technology": "customer service", "translation", "smart home appliances"; and the entities within the single-hop range of "communication technology" and with the same UQS relationship type "application direction": "remote meeting", "chatbot". Next, calculate the entity fusion degrees between the two groups of entities {"customer service", "translation", "smart home appliances"} and {"remote meeting", "chatbot"} respectively. Finally, it is obtained that the entity fusion degree between "customer service" and "chatbot" is the largest. Therefore, "chatbot" is the new entity UQSN, that is, the new application direction of "speech recognition technology" is "chatbot", which is also the expansion idea after knowledge fusion.

[0096] 5.2. Updating the knowledge graph through the new entity includes:

[0097] Adding the new entity UQSN to the node of the entity corresponding to the question extraction result UQS, generating a new triple "new entity - question extraction result corresponding relationship type - question extraction result corresponding entity". Example: Adding "chatbot" to the entity node "speech recognition technology" to generate a new triple "speech recognition technology - application direction - chatbot".

[0098] 5.3. Performing statement processing on the corresponding entity AS to generate an idea expansion result and feedback it to the user; performing statement processing on the new entity UQSN to generate an idea expansion result and feedback it to the user; the statement processing here is:

[0099] Use custom rules to splice triple information. The custom rules are like "[entity]'s [relationship] has [entity]". For example: "The application directions of speech recognition technology include chatbots."

[0100] Embodiment 2:

[0101] The embodiment of the present invention provides an auxiliary expansion idea generation device for a conference assistant. The device includes:

[0102] A question acquisition module, used to acquire user questions;

[0103] A question extraction module, used to perform entity relationship extraction on user questions to obtain a question extraction result;

[0104] A result search module, used to search for the question extraction result in the information library to obtain a question search result;

[0105] A result matching module, used to match the question extraction result in the knowledge graph to obtain a question matching result;

[0106] An idea expansion module, used to compare the question search result and the question matching result, and determine whether there are entities in the question matching result that are not in the question search result. If so, perform statement processing on the corresponding entities to generate an idea expansion result and feedback it to the user;

[0107] If not, process the question extraction result based on the entity fusion degree to obtain new entities, perform statement processing on the new entities to generate an idea expansion result and feedback it to the user, and at the same time add the new entities to the nodes of the entities corresponding to the question extraction result, thereby updating the knowledge graph.

[0108] Embodiment 3:

[0109] Based on Embodiment 1, the present invention provides an auxiliary expansion idea generation device for a conference assistant, including a processor and a storage medium;

[0110] The storage medium is used to store instructions;

[0111] The processor is used to operate according to the instructions to execute the steps of the above method.

[0112] Embodiment 4:

[0113] Based on Embodiment 1, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0114] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0115] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0116] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, so that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and the instructions executed on the computer or other programmable devices provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks.

[0118] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A method for generating an auxiliary expansion idea of a meeting assistant, characterized in that, Including: Obtain the user's question sentence; Perform entity relationship extraction on the user's question sentence to obtain the question extraction result; Search for the question extraction result in the information repository to obtain the question search result; Match the question extraction result in the knowledge graph to obtain the question matching result; Compare the question search result and the question matching result, and determine whether there are entities in the question matching result that are not in the question search result. If so, perform sentence processing on the corresponding entities to generate a thought expansion result and feedback it to the user; If not, process the question extraction result based on the entity fusion degree to obtain new entities, perform sentence processing on the new entities to generate a thought expansion result and feedback it to the user, and at the same time add the new entities to the nodes of the entities corresponding to the question extraction result, thereby updating the knowledge graph.

2. The method for generating an auxiliary expansion idea of a meeting assistant according to claim 1, characterized in that, The performing entity relationship extraction on the user's question sentence to obtain the question extraction result includes: Construct an RLC network model for entity relationship extraction, and perform entity relationship extraction on the user's question sentence through the RLC network model; wherein, the RLC network model includes an input layer, an embedding layer, an encoding layer, a classification layer, and an output layer; The input layer: receives sentence information and processes it using a Chinese word segmentation tool to obtain word and maximum forward matching word vectors; The embedding layer: receives the words and word vectors flowing out of the input layer and uses an improved BERT network model to generate sequence text; The encoding layer: uses an LSTM network model to obtain the language features of the sequence text and outputs them; The classification layer: performs entity recognition and relationship recognition based on the language features. The entity recognition includes using a CRF model to perform entity recognition on the language features and using Viterbi decoding to output the name labels of the entities; the relationship recognition includes determining the entity position information through the entity labels, inputting the text encoding information between the entities into the relationship classification module based on the entity position information, and using a CNN model to perform relationship recognition and using Softmax decoding to output the type labels of the relationships; The output layer: in the entity recognition, outputs the name labels of the entities; in the relationship recognition, outputs the type labels of the relationships.

3. The method for generating an auxiliary expansion idea of a meeting assistant according to claim 2, wherein The improved BERT network model includes: Delete the next sentence prediction task of the BERT network model; improve the masking strategy to dynamic masking; use a vocabulary encoded with double-byte encoding to encode Chinese language representations.

4. The method for generating an auxiliary expansion idea of a meeting assistant according to claim 1, wherein, The searching for the question extraction result in the information repository to obtain the question search result includes: Access and search the information repository Info through a crawler tool for the question extraction result; Obtain the title information of the top multiple search results, and perform entity relationship extraction on the title information to obtain the title extraction result Perform deduplication processing on the title extraction result to obtain the question search result.

5. The method for generating an auxiliary expansion idea of a meeting assistant according to claim 1, characterized in that, The processing the question extraction result based on the entity fusion degree to obtain new entities includes: Calculate the entity fusion degree of each entity in the knowledge graph and the entity corresponding to the question extraction result in turn; Select multiple entities with the largest entity fusion degree from the knowledge graph as search targets; Select entities in the knowledge graph within the single-hop path range of the search targets and with the same relationship type as the question extraction result as candidate entities; Calculate the entity fusion degree of each entity and candidate entity in the knowledge graph within the single-hop path range of the search target in sequence; Select the entity with the maximum entity fusion degree from the candidate entities as the new entity.

6. The method for generating an auxiliary expansion idea of a meeting assistant according to claim 5, characterized in that, The entity fusion degree Fusion is: Fusion(c i ,c j ) = α * Sim sem (c i ,c j ) + β * Sim str (c i ,c j ) In the formula, Fusion(c i ,c j ) is the entity fusion degree of entity c i and entity c j . α and β are preset weight coefficients. Sim sem (c i ,c j ), Sim str (c i ,c j ) are the semantic similarity and structural similarity of entity c i and entity c j ; The obtaining of the semantic similarity includes: using a pre-trained BERT network model to encode entity c i and entity c j to generate word vectors, and calculating the cosine similarity between the word vectors to obtain the semantic similarity; The structural similarity is: where c anc is the common entity connected to both entity c i and entity c j , dis(c i ), dis(c j ) respectively represent the distances between entity c i and entity c j and entity c anc .

7. A method for generating an auxiliary expansion idea of a meeting assistant according to claim 1, characterized in that, Updating the knowledge graph through the new entity includes: Add the new entity to the node of the entity corresponding to the question extraction result, and generate a new triple "new entity - question extraction result corresponding relationship type - question extraction result corresponding entity".

8. An auxiliary extended idea generation device for a meeting assistant, characterized in that, The device includes: A question acquisition module for acquiring user questions; A question extraction module for performing entity relationship extraction on user questions to obtain question extraction results; A result search module for searching the question extraction results in the information library to obtain question search results; A result matching module for matching the question extraction results in the knowledge graph to obtain question matching results; An idea expansion module for comparing the question search results and the question matching results, determining whether there are entities in the question matching results that are not in the question search results. If so, perform statement processing on the corresponding entities and feedback the idea expansion results to the user; If not, process the question extraction results based on the entity fusion degree to obtain a new entity, perform statement processing on the new entity to generate idea expansion results and feedback them to the user, and at the same time add the new entity to the node of the entity corresponding to the question extraction result, thereby updating the knowledge graph.

9. An auxiliary expansion idea generation device for a meeting assistant, characterized in that Includes a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method according to any one of claims 1-7.

Images