Intelligent customer service system with self-adaptive semantic understanding

Abstract

The invention discloses an intelligent customer service system with self-adaptive semantic understanding. The intelligent customer service system comprises a first receiving module, a semantic analysis module, a second receiving module, a retrieval module and a question and answer standard library, the first receiving module is used for receiving user input information; the semantic analysis module is used for matching a question and answer standard library according to user input information and generating a standard question index sequence; the second receiving module is used for receiving one or more standard problem indexes; the retrieval module is used for performing retrieval in the question and answer standard library according to the standard question index and generating a standard answer; the question-answer standard library is used for constructing a knowledge graph; the storage module is used for storing standard questions and associated standard answers; according to the intelligent customer service system, the user can confirm own problems after semantic analysis, blind identification of the system is avoided, self-adaptive semantic understanding is achieved, the knowledge graph is used for managing the question and answer standard library, and the accuracy and the real-time performance of the intelligent customer service system are guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically to an intelligent customer service system with adaptive semantic understanding. Background Art

[0002] At present, with the development and popularization of Internet technology, intelligent customer service systems have become an important tool for many enterprises and institutions to improve service efficiency and reduce labor costs. However, in actual application, traditional intelligent customer service systems have some significant limitations, especially in terms of problem identification accuracy and knowledge update mechanism.

[0003] First, existing intelligent customer service systems mainly rely on keyword matching or simple natural language processing technology to identify user questions. Although this method can meet basic service needs to a certain extent, it often fails to accurately understand the user's intentions for complex or ambiguous queries, resulting in low relevance of responses and poor user experience. Users sometimes have to try different ways of expression multiple times to get a satisfactory answer, which not only increases the user's communication cost, but also reduces the practicality of the system.

[0004] Secondly, the knowledge base of traditional customer service systems is usually static, with a long update cycle and requiring manual intervention. This update method is difficult to adapt to the rapidly changing market demand and service environment. For example, when a company launches a new product or service, if the relevant information cannot be incorporated into the knowledge base of the customer service system in a timely manner, it may result in the customer service being unable to correctly answer users' inquiries about the new product, which in turn affects user satisfaction and the company's image.

[0005] In addition, due to the lack of an effective knowledge management mechanism, the existing intelligent customer service system is difficult to achieve personalized services. Personalized data such as different users' background information, preferences, and historical interaction records have not been effectively utilized, making the services provided by the customer service system relatively single and unable to meet the personalized needs of users.

[0006] Therefore, it is particularly important to develop an intelligent customer service system that can accurately identify user intentions and has dynamic update capabilities. Such a system can not only improve the accuracy of problem identification, but also ensure that the customer service system always has the latest and most complete information through an automated knowledge graph maintenance mechanism, thereby better serving users. Summary of the invention

[0007] In view of this, the present invention provides an intelligent customer service system with adaptive semantic understanding, which can allow users to confirm their own problems through semantic analysis, avoid blind identification of the system, realize adaptive semantic understanding, and use knowledge graph to manage the question and answer standard library, thereby ensuring the accuracy and real-time performance of the intelligent customer service system.

[0008] In order to achieve the above object, the present invention adopts the following technical solution:

[0009] An intelligent customer service system with adaptive semantic understanding, comprising a first receiving module, a semantic parsing module, a second receiving module, a retrieval module and a question-answering standard library;

[0010] The first receiving module is used to receive user input information;

[0011] The semantic parsing module is used to match the question and answer standard library according to the user input information to generate a standard question index sequence;

[0012] The second receiving module is used to receive one or more standard question indexes;

[0013] The retrieval module is used to search the question and answer standard library according to the standard question index and generate a standard answer;

[0014] The question-answer standard library is used to construct a knowledge graph; it is used to store standard questions and associated standard answers.

[0015] Preferably, it also includes a standard updating module, which is used to receive updated content to update the knowledge graph, and update standard questions and corresponding standard answers according to the updated knowledge graph.

[0016] Preferably, the standard update module includes a question-answer pairing unit and a node configuration unit;

[0017] The question-answer pairing unit is used to extract entities and relationships from the knowledge graph, generate corresponding questions and answers, and form a question-answer standard pair;

[0018] The node identification unit is used to receive a configuration file of an administrator, update the knowledge graph according to the configuration file, and synchronously update the generated question-answer standard pairs.

[0019] Preferably, the node configuration unit confirms the update operation type and the update object by identifying the file content of the configuration file; the update operation type is addition, replacement or abolition; and the update object is an entity or an edge.

[0020] Preferably, the node configuration unit confirms the node object and the edge object by identifying the file content, and confirms the update operation type of the node object and the edge object by identifying keywords.

[0021] Preferably, the semantic parsing module includes a triple extraction unit, a question screening unit, a similarity analysis unit and a sorting output unit;

[0022] The triplet extraction unit is used to extract triples according to the user input information;

[0023] The question screening unit screens corresponding candidate standard questions from the question-answer standard library according to the triples to obtain a candidate question set;

[0024] The similarity analysis unit is used to calculate the similarity between the user input information and each candidate standard question in the candidate question set;

[0025] The sorting output unit is used to sort according to the similarity calculation results, and output the top N standard questions with the highest similarity to form a standard question index sequence.

[0026] Preferably, the similarity analysis unit calculates the similarity between the user input information and the question and answer standard library through a trained deep learning model, and outputs the top N matching results with the largest similarity ranking through similarity sorting to form a standard question index sequence.

[0027] Preferably, the deep learning model includes an input layer, a representation layer and a matching layer;

[0028] The input layer is used to convert the user input information text and the candidate standard question text into an embedded vector representation;

[0029] The representation layer is used to encode the embedded vector representation to obtain a latent vector;

[0030] The matching layer is used to calculate the distance between the user input information and the latent vector corresponding to the candidate standard question.

[0031] Preferably, the representation layer includes a position encoding layer, a multi-head attention layer, a gating mechanism layer, a feedforward neural network layer, and a residual connection and normalization layer connected in sequence; wherein the gating mechanism layer is also connected to the output of the position encoding layer.

[0032] A computer-readable storage medium stores a computer program, which implements the functions of the above-mentioned intelligent customer service system when executed by a processor.

[0033] It can be seen from the above technical solutions that compared with the prior art, the present invention discloses an intelligent customer service system with adaptive semantic understanding. By introducing a preliminary question identification module, it can perform multi-dimensional analysis on the questions input by users, and even in the face of complex or ambiguous statements, it can provide multiple possible interpretation options for users to choose from, effectively improving the accuracy of problem identification. Users can choose the option that best suits their intentions, thereby guiding the system to locate problems more accurately, reducing ineffective communication caused by misunderstandings, and improving user satisfaction. A knowledge graph system is constructed. Using this structured knowledge storage form, the system can more efficiently integrate and manage various types of information resources. Secondly, the present invention introduces a dynamic update mechanism for the knowledge graph, allowing the system to automatically update its knowledge base based on the latest market feedback, product changes, or service adjustments, ensuring that the customer service system can always provide the latest and most accurate information to users. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0035] Figure 1 A schematic diagram of the structure of an intelligent customer service system with adaptive semantic understanding provided by an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the structure of a semantic parsing module in an embodiment of the present invention;

[0037] Figure 3 Schematic diagram of the structure of the deep learning model used by the similarity analysis unit in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] Example 1

[0040] like Figure 1 The embodiment of the present invention discloses an intelligent customer service system with adaptive semantic understanding, including a first receiving module, a semantic parsing module, a second receiving module, a retrieval module and a question and answer standard library.

[0041] The first receiving module is used to receive user input information; the semantic parsing module is used to match the question and answer standard library according to the user input information and generate a standard question index sequence; the second receiving module is used to receive one or more standard question indexes; the retrieval module is used to search in the question and answer standard library according to the standard question index and generate a standard answer; the question and answer standard library is used to construct a knowledge graph; and is used to store standard questions and associated standard answers.

[0042] In this embodiment, the user can input information into the first receiving module or the second receiving module of the system through a visual interactive interface; the user information received by the first receiving module can be text, voice or other forms of inquiry or question, and the conversion of multimodal data to text can be achieved through technologies such as voice conversion or image recognition. The standard question index received by the second receiving module is actively confirmed by the user; when the semantic parsing module generates a standard question index sequence, the sequence is presented to the user. After the user selects the standard question index in the sequence, it is identified by the second receiving module and notified to the retrieval module. The retrieval module can output the standard answer based on the association or mapping relationship between the pre-stored standard questions and standard answers. Among them, the question and answer standard library is a database containing pre-defined standard questions and their corresponding standard answers. In addition, the question and answer standard library is also used to build and maintain the knowledge graph to ensure that the system can update itself according to the latest information and maintain the timeliness and accuracy of the answers.

[0043] In order to further implement the above technical solution, for the update of the knowledge graph, the system also includes a standard update module, which is used to receive updated content to update the knowledge graph, and update standard questions and corresponding standard answers according to the updated knowledge graph.

[0044] Specifically, the standard update module includes a question-answer pairing unit and a node configuration unit; the question-answer pairing unit is used to extract entities and relationships from the knowledge graph, generate corresponding questions and answers, and constitute question-answer standard pairs; the node identification unit is used to receive the administrator's configuration file, update the knowledge graph according to the configuration file, and synchronously update the generated question-answer standard pairs.

[0045] The node configuration unit confirms the update operation type and update object by identifying the file content of the configuration file; the update operation type is add, replace or abolish; the update object is an entity or an edge. The node configuration unit extracts triples according to the file content, confirms the node object and the edge object, and confirms the operation type by comparing the current knowledge graph, and then performs the update.

[0046] Exemplarily, when a user modifies an order, the administrator notifies the node configuration unit of the user's behavior in the form of file content. The node configuration unit then identifies the user node, order number node, product node and other related nodes and edges to confirm the corresponding operation type. For order modifications, the corresponding nodes or edges are updated for the modified objects. For example, if the product is replaced, the operation type is confirmed to be node replacement by comparing with the current knowledge graph, and the product node is replaced; if a note is added, after comparing with the current knowledge graph, it can be found that the node is not included in the current knowledge graph, and therefore, the operation type can be confirmed to be addition.

[0047] like Figure 2 ,In order to further implement the above technical solution, the semantic parsing module includes a triple extraction unit, a question ,screening unit, a similarity analysis unit and a ,sorting output unit.

[0048] The triple extraction unit is used to extract triples based on user input information; the question screening unit screens out corresponding candidate standard questions in the question and answer standard library based on the triples to obtain a candidate question set; the similarity analysis unit is used to calculate the similarity between the user input information and each candidate standard question in the candidate question set; the sorting output unit is used to sort according to the similarity calculation results, and output the top N standard questions with the highest similarity to form a standard question index sequence.

[0049] In this embodiment, the standard questions in the standard question-answering database are configured with a label set, and each standard question is annotated by the nodes and edges involved. In the process of semantic analysis, in order to improve the analysis efficiency, after obtaining the question input by the user, the triples of the question input by the user are extracted, and the extracted structure is used as a label index, so that relevant content can be quickly located in the huge data in the standard question-answering database, and relevant questions can be screened.

[0050] Furthermore, after extracting triples, relevant entities and their relationships can be quickly queried in the knowledge graph through recall as the final result of triple extraction. In this way, even if the user uses synonymous or non-standard expressions when describing the problem, standard entities and edges in the knowledge graph can be selected under the recall of the knowledge graph, thereby improving the accuracy of semantic understanding.

[0051] Furthermore, the similarity analysis unit calculates the similarity between the user input information and the question-answer standard library through the trained deep learning model, and outputs the top N matching results with the largest similarity ranking through similarity sorting to form a standard question index sequence. The distance formula is:

[0052]

[0053] Among them, A and B represent the latent vectors corresponding to the input information and the candidate standard question respectively.

[0054] Specifically, the deep learning model includes an input layer, a representation layer, and a matching layer; the input layer is used to convert the user input information text and the candidate standard question text into an embedded vector representation; the representation layer is used to encode the embedded vector representation to obtain a latent vector; the matching layer is used to calculate the distance between the user input information and the corresponding latent vector of the candidate standard question.

[0055] The representation layer includes a position encoding layer, a multi-head attention layer, a gating mechanism layer, a feedforward neural network layer, and a residual connection and normalization layer connected in sequence; wherein the gating mechanism layer is also connected to the output of the position encoding layer, such as Figure 3 , Figure 3 The dotted part in the figure is the encoder part.

[0056] In this embodiment, multi-head attention allows the model to focus on different parts of the input when processing sequence data, and is able to process these parts in parallel, with each "head" focusing on different aspects of the input, and then merging their results so that the model can obtain different perspectives of the input. The gating mechanism is a method for controlling the flow of information, which determines how much information should be passed through the network to the next layer. The feedforward network is a sequence of two fully connected (FC) layers. These networks are applied in parallel at each position over the entire sequence, which means that they do not change the dependencies within the sequence. They provide additional nonlinear transformation capabilities, allowing the model to learn more complex feature mappings.

[0057] The input first passes through a multi-head self-attention layer, where it gains a new perspective on the input data. The original input is then added to the result of the self-attention via a residual connection, and this result is fed into a layer normalization layer. The normalized result is then fed into a gating mechanism and then passed through a feedforward neural network. After the feedforward network, residual connections and layer normalization are used again, and finally the output of the encoder module is obtained.

[0058] This embodiment can effectively capture the long-term dependencies in the input data and maintain the stability of the model through residual connections and normalization layers.

[0059] Example 2

[0060] Based on the same inventive concept, an embodiment of the present invention discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the functions of the intelligent customer service system in Example 1 can be realized.

[0061] By introducing a preliminary question identification module, the present invention can perform a multi-dimensional analysis of the questions input by the user, and can provide multiple possible explanation options for the user to choose even in the face of complex or ambiguous statements, thereby effectively improving the accuracy of problem identification. Users can choose the option that best suits their intentions, thereby guiding the system to locate the problem more accurately, reducing ineffective communication caused by misunderstandings, and improving user satisfaction. A knowledge graph system has been constructed, and using this structured knowledge storage form, the system can more efficiently integrate and manage various types of information resources. Secondly, the present invention introduces a dynamic update mechanism for the knowledge graph, allowing the system to automatically update its knowledge base based on the latest market feedback, product changes, or service adjustments, ensuring that the customer service system can always provide the latest and most accurate information to users.

[0062] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0063] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent customer service system with adaptive semantic understanding, characterized in that: It includes a first receiving module, a semantic parsing module, a second receiving module, a retrieval module and a question-answering standard library; The first receiving module is used to receive user input information; The semantic parsing module is used to match the question and answer standard library according to the user input information to generate a standard question index sequence; The second receiving module is used to receive one or more standard question indexes; The retrieval module is used to search the question and answer standard library according to the standard question index and generate a standard answer; The question-answer standard library is used to construct a knowledge graph; it is used to store standard questions and associated standard answers.

2. The intelligent customer service system with adaptive semantic understanding according to claim 1, characterized in that: It also includes a standard updating module, which is used to receive updated content to update the knowledge graph, and update standard questions and corresponding standard answers according to the updated knowledge graph.

3. The intelligent customer service system with adaptive semantic understanding according to claim 2, characterized in that: The standard update module includes a question-answer pairing unit and a node configuration unit; The question-answer pairing unit is used to extract entities and relationships from the knowledge graph, generate corresponding questions and answers, and form a question-answer standard pair; The node identification unit is used to receive a configuration file of an administrator, update the knowledge graph according to the configuration file, and synchronously update the generated question-answer standard pairs.

4. The intelligent customer service system with adaptive semantic understanding according to claim 3, characterized in that: The node configuration unit confirms the update operation type and the update object by identifying the file content of the configuration file; the update operation type is addition, replacement or abolition; and the update object is an entity or an edge.

5. The intelligent customer service system with adaptive semantic understanding according to claim 4, characterized in that: The node configuration unit extracts triples according to the file content, confirms the node object and the edge object, and confirms the operation type by comparing the current knowledge graph.

6. The intelligent customer service system with adaptive semantic understanding according to claim 1, characterized in that: The semantic parsing module includes a triple extraction unit, a question screening unit, a similarity analysis unit and a sorting output unit; The triplet extraction unit is used to extract triples according to the user input information; The question screening unit screens corresponding candidate standard questions from the question-answer standard library according to the triples to obtain a candidate question set; The similarity analysis unit is used to calculate the similarity between the user input information and each candidate standard question in the candidate question set; The sorting output unit is used to sort according to the similarity calculation results, and output the top N standard questions with the highest similarity to form a standard question index sequence.

7. The intelligent customer service system with adaptive semantic understanding according to claim 6, characterized in that: The similarity analysis unit calculates the similarity between the user input information and the question and answer standard library through a trained deep learning model, and outputs the top N matching results with the largest similarity ranking through similarity sorting to form a standard question index sequence.

8. The intelligent customer service system with adaptive semantic understanding according to claim 7, characterized in that: The deep learning model includes an input layer, a representation layer and a matching layer; The input layer is used to convert the user input information text and the candidate standard question text into an embedded vector representation; The representation layer is used to encode the embedded vector representation to obtain a latent vector; The matching layer is used to calculate the distance between the user input information and the latent vector corresponding to the candidate standard question.

9. The intelligent customer service system with adaptive semantic understanding according to claim 8, characterized in that: The representation layer includes a position encoding layer, a multi-head attention layer, a gating mechanism layer, a feedforward neural network layer, and a residual connection and normalization layer connected in sequence; wherein the gating mechanism layer is also connected to the output of the position encoding layer.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the functions of the intelligent customer service system as described in any one of claims 1 to 9.

Images