CN119621955B

CN119621955B - Algorithm AI customer service dialogue management method and system based on big data

Info

Publication number: CN119621955B
Application number: CN202510157821.4A
Authority: CN
Inventors: 高婷
Original assignee: Shenzhen Rongcan Big Data Technology Co ltd
Current assignee: Shenzhen Rongcan Big Data Technology Co ltd
Priority date: 2025-02-13
Filing date: 2025-02-13
Publication date: 2025-05-23
Anticipated expiration: 2045-02-13
Also published as: CN119621955A

Abstract

The invention relates to the technical field of dialogue management, and discloses an algorithm AI customer service dialogue management method and system based on big data, wherein the method comprises the steps of acquiring user input content and a speech pattern; the method comprises the steps of carrying out keyword retrieval according to user input content and a voice map to obtain a voice tag and activate corresponding voice content, constructing a voice directed graph according to the voice tag and the voice map, carrying out greedy optimization to obtain a voice priority list, outputting all activated voice content according to the voice priority list, ending dialogue after voice output is completed, carrying out emotion state assessment according to the user input content, outputting a pacifying voice and ending dialogue when user emotion is detected to be negative, carrying out convergence degree calculation according to the user input content, and ending dialogue when topic convergence degree is larger than a preset convergence degree threshold value. The method has the following effect that the customer service dialogue efficiency can be optimized.

Description

Algorithm AI customer service dialogue management method and system based on big data

Technical Field

The invention relates to the technical field of dialogue management, in particular to an algorithm AI customer service dialogue management method and system based on big data.

Background

In the current digital age, with the vigorous development of industries such as electronic commerce, online service and the like, a customer service system is taken as an important bridge for communication between enterprises and users, and the efficiency and the quality of the customer service system are directly related to the quality of user experience. The traditional customer service mode depends on a manual seat, and has the disadvantages of low response speed and unstable service quality. To solve these problems, an algorithm AI customer service dialogue management method based on big data has been developed. The method can automatically understand the query intention of the user by utilizing advanced technologies such as Natural Language Processing (NLP), machine learning and the like, and provides accurate service or solution according to a preset speech pattern, thereby realizing high-efficiency customer service automation.

In the prior art, a traditional AI customer service system based on rules or template matching is adopted. Such systems respond to a user's query by presetting a series of questions or fixed-line telephone procedures. Specifically, when user input is received, the system first parses the text content, looks for a similar question pattern as in the built-in knowledge base, and then selects the best matching answer to reply. In addition, some systems began to introduce simple machine learning models for improving intent recognition and increasing the accuracy of answers.

Although the above-described conventional AI customer service systems improve service efficiency to some extent, they also have significant limitations. Because the system mainly depends on a fixed rule set or a fixed template, the adaptability and the flexibility of the system are poor for new problems or special situations beyond a preset range, and users can not obtain satisfactory solutions easily, so that the conversation efficiency of the users with customer service is low.

Disclosure of Invention

The invention provides an AI customer service dialogue management method and system based on a big data algorithm, which are used for improving AI customer service dialogue efficiency.

In order to solve the above technical problems, the present invention provides a big data-based algorithm AI customer service dialogue management method, including:

acquiring user input content and a speech surgery map;

keyword retrieval is carried out according to the user input content and the voice operation map, voice operation labels are obtained, and corresponding voice operation content is activated;

constructing a speaking directed graph according to the speaking tag and the speaking map, and performing greedy optimization to obtain a speaking priority list;

outputting all the activated speaking contents according to the speaking priority list, and ending the dialogue after completing speaking output;

carrying out emotion state evaluation according to the user input content, outputting a pacifying operation and ending the dialogue when the user emotion negative is detected;

And calculating the convergence degree according to the user input content, and ending the dialogue when the topic convergence degree is larger than a preset convergence degree threshold value.

In an optional implementation manner, the keyword searching according to the user input content and the voice graph, obtaining a voice tag and activating corresponding voice content includes:

Extracting keywords from the user input content to obtain user keywords;

Comparing and matching the user keywords in the conversation map, and searching conversation labels corresponding to the user keywords;

when a corresponding conversation label is found, recording the conversation label and activating corresponding conversation content;

And when the directly corresponding conversation label is not found, carrying out semantic recognition on the user input content to obtain the conversation label with the closest semantic meaning, and activating the corresponding conversation content.

In an alternative embodiment, the constructing an outgoing call directed graph according to the call label and the call map, and performing greedy optimization to obtain a call priority list includes:

each call is taken as a node, directed edges among the nodes are established according to a preset initial call retrieval sequence, and a call outgoing directed graph is constructed;

Acquiring a tag priority list;

When traversing the speaking oriented graph, preferentially traversing the speaking corresponding to the speaking label with high priority in the label priority list;

After traversing the speaking directed graph, outputting all the activated speaking contents to obtain a speaking priority list according to the traversing sequence.

In an alternative embodiment, the performing the emotional state assessment according to the user input content, outputting pacifying and ending the dialogue when detecting the emotional negative of the user includes:

Inputting the user input content into a pre-trained emotion analysis model, and outputting to obtain a user emotion value;

When the emotion value of the user is lower than a preset negative emotion threshold value, indicating that the user has negative emotion, and recording the current turn;

Outputting a pacifying operation and ending the dialogue when the turn of the negative emotion is larger than the preset emotion upper limit;

The training process of the emotion analysis model comprises the following steps:

Constructing an emotion analysis model based on historical user texts, training the model, and judging that training is completed after detecting that a loss function of the model meets a condition to obtain a trained model;

and inputting the user input content into the trained emotion analysis model to obtain the emotion value of the user.

In an optional implementation manner, the calculating the convergence according to the user input content, and ending the dialogue when the topic convergence is greater than a preset convergence threshold value, includes:

Extracting key content according to the user input content to obtain user keywords;

vectorization is carried out according to the user keywords, and text vectors are obtained;

performing similarity calculation on the text vectors of two adjacent rounds to obtain text similarity;

When the text similarity is larger than a preset topic convergence threshold, topic convergence is indicated, and the current turn is recorded;

and outputting polite and ending the dialogue when the turns of the convergence of the topics continuously appear are larger than the preset convergence upper limit.

In an alternative embodiment, before said outputting all the active speaking contents according to the speaking priority list and ending the dialogue after completing the speaking output, the method further comprises:

the semantic confidence is calculated by the following formula:

Wherein, For the degree of confidence of the semantics,The number of matching nodes is represented and,The number of total nodes is represented and,The number of matching edges is represented and,The number of the total edges is represented,AndIs a weight coefficient;

when emotion tendencies of the user input content and the speech content are consistent, adding a preset first confidence coefficient difference value on the basis of the semantic confidence coefficient to obtain an optimized semantic confidence coefficient;

And when the optimized semantic confidence is smaller than a preset confidence threshold, converting the speech content into an inactive state.

In an optional implementation manner, when the emotion tendencies of the user input content and the speech content are consistent, adding a preset first confidence difference value to the semantic confidence coefficient to obtain an optimized semantic confidence coefficient, and before the semantic confidence coefficient is obtained, further including:

Inputting the speech content into a pre-trained emotion analysis model, and outputting to obtain a man-machine emotion value;

and when the emotion difference value between the user emotion value and the man-machine emotion value is smaller than a preset emotion threshold value, representing that emotion tendencies are consistent.

In a second aspect, the present invention provides an AI customer service dialogue management system based on big data, including:

The data acquisition module is used for acquiring user input content and a speech pattern;

the voice operation searching module is used for searching keywords according to the user input content and the voice operation map, obtaining a voice operation label and activating corresponding voice operation content;

The speaking list module is used for constructing a speaking directed graph according to the speaking tag and the speaking map, and performing greedy optimization to obtain a speaking priority list;

The conversation output module is used for outputting all the activated conversation contents according to the conversation priority list and ending the conversation after the conversation output is completed;

The emotion pacifying module is used for carrying out emotion state assessment according to the user input content, outputting pacifying voice operation and ending dialogue when detecting the emotion negative of the user;

and the topic convergence module is used for calculating the convergence degree according to the user input content, and ending the dialogue when the topic convergence degree is larger than a preset convergence degree threshold value.

In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, where the processor implements the big data based algorithm AI customer service dialogue management method according to any one of the above when executing the computer program.

In a fourth aspect, the present invention further provides a computer readable storage medium, where the computer readable storage medium includes a stored computer program, where when the computer program runs, the device where the computer readable storage medium is controlled to execute the big data based algorithm AI customer service dialogue management method according to any one of the above.

Compared with the prior art, the method and the system have the advantages that the method and the system for managing the AI customer service conversations based on the big data are disclosed, the method comprises the steps of obtaining user input content and a conversation map, carrying out keyword retrieval according to the user input content and the conversation map to obtain conversation labels and activate corresponding conversation content, constructing a conversation directed graph according to the conversation labels and the conversation map, carrying out greedy optimization to obtain a conversation priority list, outputting all the activated conversation content according to the conversation priority list, ending conversations after conversation output is completed, carrying out emotion state evaluation according to the user input content, outputting pacifying conversation and ending conversations when emotion negative of a user is detected, carrying out convergence calculation according to the user input content, and ending conversations when topic convergence is greater than a preset convergence threshold. The method has the following effect that the customer service dialogue efficiency can be optimized.

In particular, the method introduces a mechanism for intelligently determining when to end a session with a user. The mechanism first performs key content extraction on the user input content to identify keywords that can represent the user's intent and topic trend. By converting these keywords into text vectors, unstructured text information can be converted into a mathematically processable form. Then, by performing similarity calculation on text vectors in two adjacent rounds of conversations, the degree of variation of topics in the conversations can be quantitatively evaluated.

When the text similarity of the successive rounds of conversations exceeds a preset topic convergence threshold, meaning that the conversational content tends to repeat or no new information is generated, the system records the current round. If this high similarity condition continues to occur and exceeds a preset upper convergence limit, indicating that the dialog has reached a steady state, there is no further in-depth communication. At this point, the system will take steps to output polite to gracefully end the session, avoiding meaningless session extensions. The effectiveness of the AI customer service system is facilitated to be improved, and a better experience is ensured for the user because the conversation is neither terminated prematurely nor prolonged excessively.

Further, the method introduces an emotional state assessment mechanism to enhance the customer service experience. Specifically, the method quantifies the current emotional state of the user by transmitting the input content of the user to a pre-trained emotion analysis model, and outputs an emotion value capable of reflecting the positive and negative degrees of the emotion of the user. This process relies on advanced natural language processing techniques and deep learning models that are trained on large scale labeled datasets to accurately capture and understand emotion information in text.

When a user emotion value is detected below a set negative emotion threshold, indicating that the user is in an discontent, frustration or other negative emotional state, the system may record the turn in which this occurred. If the number of consecutive occurrences of negative emotions exceeds a preset upper emotion limit, the user's emotion is considered to have reached a level that requires special attention. At this point, the system will take steps to first output a pacifying speech, attempt to alleviate the user's emotion, and then end the conversation, avoiding further exacerbating the user's negative emotion or resulting in a more complex situation. The emotion recognition and response mechanism is not only helpful for timely finding and processing emotion demands of users, but also can effectively prevent problem upgrading caused by negative emotion accumulation. User satisfaction may be improved by terminating conversations that raise more problems in advance and giving the proper pacifying.

The invention also introduces a greedy optimization algorithm to construct a speech surgery directed graph and generate a speech surgery priority list. Specifically, each session is taken as a node, and directed edges among the nodes are established according to a preset initial retrieval sequence to form a structured dialogue flow network. Then, when traversing this graph, those call paths that are given high priority are prioritized so that important and urgent questions can be responded to in time. In addition, the system calculates semantic confidence, evaluates the degree of agreement between the user input and the phone call that the system is ready to respond, and when the semantic confidence is below a certain threshold, the related phone call will not be selected as the final answer, thereby avoiding low quality communication caused by misunderstanding the user's intent.

Drawings

Fig. 1 is a schematic flow chart of an algorithm AI customer service dialogue management method based on big data according to a first embodiment of the present invention;

fig. 2 is a schematic structural diagram of an algorithm AI customer service session management system based on big data according to a second embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

In order to solve the above problems, referring to fig. 1, a first embodiment of the present invention provides an algorithm AI customer service session management method based on big data, comprising the steps of:

S11, acquiring user input content and a speech surgery map;

S12, carrying out keyword retrieval according to the user input content and the voice operation map to obtain a voice operation label and activate corresponding voice operation content;

s13, constructing a speaking operation directed graph according to the speaking operation tag and the speaking operation map, and performing greedy optimization to obtain a speaking operation priority list;

s14, outputting all the activated speaking contents according to the speaking priority list, and ending the dialogue after completing speaking output;

S15, carrying out emotion state evaluation according to the user input content, outputting a pacifying operation and ending the dialogue when the emotion negative of the user is detected;

S16, performing convergence degree calculation according to the user input content, and ending the dialogue when the topic convergence degree is larger than a preset convergence degree threshold value.

In step S11, user input content and a speech pattern are acquired.

In one embodiment, the user input refers to information that the user sends to the AI-service system via a chat interface, voice interaction, or other form. This information may be in a text format (e.g., questions typed in a chat box) or may be in a non-text format (e.g., voice input, which is then converted to text). The AI system needs to parse these inputs, understand the user's intent, and prepare for the next step of response.

It is noted that a speech atlas is a structured knowledge base that contains a large number of predefined dialogue patterns, answer pairs, and dialogue strategies for handling specific situations. This graph is organized in a graph structure in which nodes represent different dialog states or topics and edges represent transition paths from one state to another, i.e. candidate dialog flows. The session map is stored in a map database.

In step S12, keyword retrieval is performed according to the user input content and the voice pattern, so as to obtain a voice tag and activate the corresponding voice content.

In one embodiment, keyword extraction is performed on the user input content to obtain user keywords, the user keywords are compared and matched in the conversation map, conversation tags corresponding to the user keywords are searched, when the corresponding conversation tags are found, the conversation tags are recorded and the corresponding conversation content is activated, when no corresponding conversation tag is found, semantic recognition is performed on the user input content to obtain the conversation tag with the closest semantic meaning, and the corresponding conversation content is activated.

It is worth to say that the extracted user keywords are then compared and matched with a keyword library in a pre-stored speech pattern. For example, if the user asks "how to change my password? and find the tag associated with it in the session map, such as" account security-modifying password ". If no directly corresponding keywords are found, the system may further use Natural Language Processing (NLP) techniques for semantic analysis. By "activate corresponding session content" it is meant that once the appropriate session tag is determined, the system prepares a response in accordance with the content indicated by the tag. Specifically, the state of the corresponding voice tag is changed to the activated state. And outputting the conversation content with all the states being active states when the follow-up output is performed.

In step S13, a speaking directed graph is constructed according to the speaking tag and the speaking map, and greedy optimization is performed to obtain a speaking priority list.

In one embodiment, each call is used as a node, directed edges among the nodes are established according to a preset initial call retrieval sequence, a call directed graph is built, a tag priority list is obtained, when the call directed graph is traversed, calls corresponding to high-priority call tags in the tag priority list are traversed preferentially, and after the call directed graph is traversed, all activated call contents are output to the call priority list according to the traversal sequence.

It is worth noting that each node represents a specific session or dialogue state. For example, "account security-modifying password", "order query-logistics information", etc. Directed edges represent the transition relationship, i.e., the estimated path of progress of a conversation, from one conversation to another. For example, the "order inquiry-predicted delivery time" may be diverted from "order inquiry-logistics information". And establishing directed edges between different dialects according to a preset initial retrieval sequence. This sequence is based on historical data, and most users will ask the predicted delivery time immediately after querying the logistics information, then there will be a directional edge between these two words.

It is worth noting that the system maintains a tag priority list that defines the importance and urgency of different types of speech. For example, sensitive topics related to payment issues, account security, etc. may be given higher priority, while non-critical issues such as product recommendations are ranked later.

It is worth noting that when traversing the phone directed graph, the system first selects those phones marked as high priority to process. This means that when the user enters certain keywords, the system will prioritize topics that are considered most important or that require immediate response. For example, if the user mentioned "unable to login", the system would quickly locate a high priority session associated therewith, such as "account security-forget password".

It should be noted that, using a greedy algorithm, a best path from the start node to the target node is gradually constructed each time the currently best selection point is selected. For example, when solving the problem that the user cannot place an order, the system will first check if there is a shortage of stock, then check if the payment gateway is working properly, and finally confirm the network connection status of the user—such order is arranged according to the common reasons appearing in big data, so as to solve the problem as soon as possible. For example, suppose a user says "I want to know when My packages can be sent to. The system activates the sequence of "confirm order number- > check logistics status- > provide predicted arrival time". This is the final list of session priorities that is formed.

In step S14, all the active speaking contents are output according to the speaking priority list, and the dialogue is ended after the speaking output is completed.

In one embodiment, the semantic confidence is calculated by the following formula:

when the emotion difference value between the user emotion value and the man-machine emotion value is smaller than a preset emotion threshold value, representing that emotion tendencies are consistent;

It should be noted that the semantic confidence is an index that measures the degree of semantic similarity between the user input and the content of the utterance to which the system is ready to respond. The number of the matched nodes and the total number of the nodes respectively represent the number of the nodes matched by the user input keywords and the total number of the nodes in the whole graph in the constructed speaking directed graph. The number of matching edges and the total number of edges represent the number of edges to which the user inputs a keyword, and the total number of edges in the entire graph, respectively. A value between 0 and 1 is calculated by the above formula, reflecting the semantic fit between the user input and the system-prepared speech content. The higher the value, the closer the two are.

It is worth noting that the system sends the content input by the user into a pre-trained emotion analysis model to obtain an emotion value (user emotion value) reflecting the current emotion state of the user. Similarly, the system also sends the speech content to be output into the same emotion analysis model to obtain a man-machine emotion value reflecting the emotion color of the speech. And if the difference value between the emotion value of the user and the emotion value of the man-machine is smaller than the preset emotion difference value threshold value, the emotion tendencies of the user emotion value and the emotion value of the man-machine are considered to be consistent. For example, if the user expresses a positive or neutral emotion and the answer provided by the system is also positive or neutral, the emotion is considered consistent, confidence is increased, and an optimized semantic confidence is obtained. If, after all of the above calculations, the optimized semantic confidence remains below the preset confidence threshold, the system will flag the speech content as inactive, meaning that it will not be selected as the final answer. This is because low confidence means that the system is not sufficiently well understood of the user's intent, and the direct use of such answers can result in poor user experience.

In step S15, an emotional state evaluation is performed according to the user input content, and when a user emotional negative is detected, a pacifying speech operation is output and the dialogue is ended.

In one embodiment, the user input content is input into a pre-trained emotion analysis model, a user emotion value is obtained through output, when the user emotion value is lower than a preset negative emotion threshold value, negative emotion appears in a user, the current turn is recorded, when the turn of the negative emotion appears is greater than a preset emotion upper limit, a pacifying operation is output and conversation is finished, the training process of the emotion analysis model comprises the steps of constructing the emotion analysis model based on historical user texts, training the model, judging that training is completed after a loss function of the model is detected to meet the condition, and obtaining a trained model, and the user input content is input into the trained emotion analysis model to obtain the user emotion value.

In one embodiment, the emotion analysis model employs an LSTM (long short term memory network) model that draws attention mechanisms, and the emotion analysis model employs an encoder-decoder architecture. The encoder is responsible for converting the input sentence into a context vector of fixed length, and the decoder generates a final emotion classification result based on this vector. During the encoding process, the attention layer calculates the importance score of each word for the entire sentence. These scores determine which words should be given more weight. For example, in a sentence, words such as "very angry" may get a higher degree of attention because they directly express strong emotions. Then, the state of each time step is multiplied by the corresponding attention weight by a weighted summation mode, and then summed to obtain a new context representation, and finally an emotion value is obtained. The emotion analysis model is trained by using historical dialogue content and historical emotion values as training sets.

In step S16, a convergence degree calculation is performed according to the user input content, and when the topic convergence degree is greater than a preset convergence degree threshold, the dialogue is ended.

In one embodiment, key content extraction is performed according to user input content to obtain a user keyword, vectorization is performed according to the user keyword to obtain a text vector, similarity calculation is performed on two adjacent rounds of the text vectors to obtain text similarity, topic convergence is represented and a current round is recorded when the text similarity is larger than a preset topic convergence threshold, and polite is output and dialogue is ended when the round of topic convergence continuously occurs is larger than a preset convergence upper limit.

In one embodiment, the text vector is obtained by vectorizing the user keywords, and the word embedding technique is used in the process of obtaining the text vector. Words may be mapped to points in a high-dimensional space such that semantically similar words are closer together in the space.

In one embodiment, cosine similarity is used for similarity calculation, and the formula is as follows:

Wherein, Is vector quantitySum vectorIs used for the cosine similarity of the (c),Is vector quantitySum vectorIs used for the dot product of (a),Is vector quantityIs used for the mold length of the mold,Is vector quantityIs a die length of the die.

It is assumed that in a customer service scenario, the user is asking different aspects of the same question for several consecutive rounds, but essentially without introducing new information. For example, "where is my order" and then "how is this order now" are the user asked "first. The system calculates the text vectors of the two sentences through the above process and finds that the cosine similarity between them is very high. When this happens continuously several times, the system realizes that the dialogue has begun to repeat, and then chooses to output a similar "we have struggled to provide your with the latest information, please contact us at any time if you have other questions, and end the dialogue.

In summary, the invention discloses an algorithm AI customer service dialogue management method based on big data, aiming at optimizing customer service dialogue efficiency in the human-computer interaction process. The method first involves the acquisition of user input content, which refers to various forms of information that the user sends to the system, and a speech pattern, which is a pre-built knowledge base that contains a number of predefined dialogue patterns, answer pairs to questions, and dialogue strategies for handling specific situations.

After receiving the user input, the system will extract keywords from it to identify key information that can represent the user's intent and topic trend. The keywords are then compared and matched with keyword libraries in the speech atlas, corresponding tags are searched for, and corresponding speech contents are activated. If a match is not found directly, then semantic recognition techniques are further used to determine the closest phonetic label based on semantic similarity.

Notably, the importance of emotional state assessment is particularly emphasized by the present invention. The method utilizes a pre-trained emotion analysis model to quantify the current emotion state of the user and outputs an emotion value reflecting the emotion positive and negative degree of the user. Upon detecting that the user's mood value is below the set negative mood threshold, indicating that the user is in an unsatisfactory or frustrated state, the system records the specific turn in which this occurred. If the number of consecutive negative emotions exceeds the preset upper emotion limit, measures are considered to be taken to alleviate the emotion of the user, such as outputting a pacifying conversation, and the conversation is ended at the right time, so that more serious problems are caused by the accumulation of negative emotions.

At the same time, the invention provides a convergence degree calculation-based method for intelligently judging when to end the dialogue with the user. The degree of change of topics in the dialog can be quantitatively evaluated by carrying out similarity calculation on text vectors extracted from two adjacent rounds of dialog. When the text similarity of successive rounds of dialog exceeds a preset topic convergence threshold, meaning that the dialog content tends to repeat or no longer generate new information, the system will output polite to gracefully end the dialog. This not only helps to improve the efficiency of the AI customer service system, but also ensures that the user gets a good experience because the conversation neither ends prematurely nor is excessively prolonged.

In summary, the big data-based algorithm AI customer service dialogue management method provided by the invention covers technical innovations from multiple aspects of user input analysis, speech surgery matching, priority ordering, emotion monitoring and the like, and achieves more efficient, intelligent and humanized customer service.

Referring to fig. 2, a second embodiment of the present invention provides an algorithm AI customer service session management system based on big data, including:

Preferably, the data acquisition module is configured to:

User input content and a speech pattern are obtained.

Preferably, the speech search module is configured to:

keyword retrieval is carried out according to the user input content and the conversation map, conversation tags are obtained, and corresponding conversation content is activated, and the method comprises the following steps:

Extracting keywords from the user input content to obtain user keywords;

Preferably, the speaking list module is configured to:

Constructing a speaking directed graph according to the speaking tag and the speaking map, and performing greedy optimization to obtain a speaking priority list, wherein the method comprises the following steps:

Acquiring a tag priority list;

Preferably, the speaking output module is configured to:

The semantic confidence is calculated by the following formula:

Preferably, the emotion soothing module is configured to:

carrying out emotion state assessment according to the user input content, outputting a pacifying operation and ending a dialogue when the user emotion negative is detected, wherein the emotion state assessment comprises the following steps:

Preferably, the topic convergence module is configured to:

performing convergence degree calculation according to the user input content, and ending the dialogue when the topic convergence degree is greater than a preset convergence degree threshold value, wherein the method comprises the following steps:

It should be noted that, the big data based algorithm AI customer service dialogue management system provided by the embodiment of the present invention is used for executing all the flow steps of the big data based algorithm AI customer service dialogue management method in the above embodiment, and the working principles and beneficial effects of the two correspond one to one, so that the description is omitted.

The embodiment of the invention also provides electronic equipment. The electronic device comprises a processor, a memory and a computer program, such as a data acquisition program, stored in the memory and executable on the processor. The processor executes the computer program to implement the steps in the embodiments of the big data based algorithm AI customer service dialogue management method described above, for example, step S11 shown in fig. 1. Or the processor, when executing the computer program, performs the functions of the modules/units in the above-described device embodiments, such as a data acquisition module.

The computer program may be divided into one or more modules/units, which are stored in the memory and executed by the processor to accomplish the present invention, for example. The one or more modules/units may be a series of computer program instruction segments capable of performing the specified functions, which instruction segments are used for describing the execution of the computer program in the electronic device.

The electronic equipment can be a desktop computer, a notebook computer, a palm computer, an intelligent tablet and other computing equipment. The electronic device may include, but is not limited to, a processor, a memory. It will be appreciated by those skilled in the art that the above components are merely examples of electronic devices and are not limiting of electronic devices, and may include more or fewer components than those described above, or may combine certain components, or different components, e.g., the electronic devices may also include input-output devices, network access devices, buses, etc.

The Processor may be a central processing unit (Central Processing Unit, CPU), other general purpose Processor, digital signal Processor (DIGITAL SIGNAL Processor, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), off-the-shelf Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like that is a control center of the electronic device, connecting various parts of the overall electronic device using various interfaces and lines.

The memory may be used to store the computer program and/or modules, and the processor may implement various functions of the electronic device by running or executing the computer program and/or modules stored in the memory and invoking data stored in the memory. The memory may mainly include a storage program area which may store an operating system, an application program required for at least one function (such as a sound playing function, an image playing function, etc.), etc., and a storage data area which may store data created according to the use of the cellular phone (such as audio data, a phonebook, etc.), etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart memory card (SMART MEDIA CARD, SMC), secure Digital (SD) card, flash memory card (FLASH CARD), at least one disk storage device, flash memory device, or other volatile solid-state storage device.

Wherein the integrated modules/units of the electronic device may be stored in a computer readable storage medium if implemented in the form of software functional units and sold or used as stand alone products. Based on such understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of each of the method embodiments described above. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the computer readable medium contains content that can be appropriately scaled according to the requirements of jurisdictions in which such content is subject to legislation and patent practice, such as in certain jurisdictions in which such content is subject to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

It should be noted that the above-described apparatus embodiments are merely illustrative, and the units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of 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. In addition, in the drawings of the embodiment of the device provided by the invention, the connection relation between the modules represents that the modules have communication connection, and can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

The foregoing embodiments have been provided for the purpose of illustrating the general principles of the present invention, and are not to be construed as limiting the scope of the invention. It should be noted that any modifications, equivalent substitutions, improvements, etc. made by those skilled in the art without departing from the spirit and principles of the present invention are intended to be included in the scope of the present invention.

Claims

1. A method for managing customer service dialogues using an algorithm based on big data, comprising:

Get user input content and speech graph;

Perform keyword search based on the user input content and the speech graph to obtain speech tags and activate corresponding speech content;

A directed graph of speech techniques is constructed according to the speech technique labels and the speech technique graph, and greedy optimization is performed to obtain a speech technique priority list;

Output all activated speech contents according to the speech priority list, and end the conversation after completing the speech output;

Evaluate the emotional state of the user based on the user input content, and when it is detected that the user is in a negative mood, output soothing words and end the conversation;

Calculating the convergence degree according to the user input content, and ending the conversation when the topic convergence degree is greater than a preset convergence degree threshold;

The method of constructing a speech directed graph according to the speech tags and the speech graph, and performing greedy optimization to obtain a speech priority list includes:

Take each speech as a node, establish directed edges between nodes according to the preset initial speech search order, and construct a speech directed graph;

Get the tag priority list;

When traversing the directed graph of speech, the speech corresponding to the speech tag with high priority in the tag priority list is traversed first;

After traversing the directed graph of speech, all activated speech contents are output in the traversal order to obtain a speech priority list;

Wherein, before outputting all activated speech contents according to the speech priority list and ending the conversation after completing the speech output, it also includes:

The semantic confidence is calculated by the following formula:

in, is the semantic confidence, Indicates the number of matching nodes, Represents the total number of nodes, represents the number of matching edges, represents the total number of edges, and is the weight coefficient;

When the sentiment tendency of the user input content and the speech content is consistent, a preset first confidence difference is added to the semantic confidence to obtain an optimized semantic confidence;

When the optimized semantic confidence is less than a preset confidence threshold, the speech content is converted to an inactivated state.

2. The algorithmic AI customer service dialogue management method based on big data according to claim 1 is characterized in that the keyword search is performed according to the user input content and the speech graph to obtain the speech tag and activate the corresponding speech content, including:

Extracting keywords from the user input content to obtain user keywords;

Compare and match the user keywords in the speech graph to find speech tags corresponding to the user keywords;

When the corresponding speech tag is found, the speech tag is recorded and the corresponding speech content is activated;

When no directly corresponding speech tag is found, semantic recognition is performed on the user input content to obtain the speech tag with the closest semantics, and the corresponding speech content is activated.

3. The method for managing customer service dialogues using an algorithm based on big data according to claim 1 is characterized in that the emotional state of the user is evaluated based on the user input content, and when a negative user emotion is detected, a soothing speech is output and the dialogue is ended, including:

Inputting the user input content into a pre-trained sentiment analysis model, and outputting a user sentiment value;

When the user's emotion value is lower than the preset negative emotion threshold, it indicates that the user has negative emotions, and the current round is recorded;

When the number of negative emotions exceeds the preset emotional limit, output soothing words and end the conversation;

The training process of the sentiment analysis model includes:

Build a sentiment analysis model based on historical user texts, train the model, and determine that the training is complete after detecting that the model's loss function meets the conditions, and obtain the trained model;

The user input content is input into the trained sentiment analysis model to obtain the user emotion value.

4. The method for managing AI customer service dialogue based on big data algorithm according to claim 1 is characterized in that the convergence degree is calculated according to the user input content, and when the topic convergence degree is greater than a preset convergence degree threshold, the dialogue is ended, comprising:

Extract key content according to the user input content to obtain user keywords;

Vectorize the user keywords to obtain a text vector;

Calculate the similarity of the text vectors of two adjacent rounds to obtain text similarity;

When the text similarity is greater than a preset topic convergence threshold, it indicates that the topic has converged, and the current round is recorded;

When the number of consecutive rounds of topic convergence exceeds the preset convergence limit, polite words are output and the conversation ends.

5. The method for managing customer service dialogues based on big data algorithms according to claim 1 is characterized in that, when the emotional tendency of the user input content and the speech content is consistent, before the semantic confidence is added with a preset first confidence difference to obtain an optimized semantic confidence, it also includes:

Input the speech content into a pre-trained sentiment analysis model, and output a human-machine sentiment value;

When the emotion difference between the user emotion value and the human-machine emotion value is less than a preset emotion threshold, it indicates that the emotion tendencies are consistent.

6. An algorithmic AI customer service dialogue management system based on big data, characterized in that it is used to implement the algorithmic AI customer service dialogue management method based on big data as described in any one of claims 1 to 5, comprising:

Data acquisition module, used to obtain user input content and speech graph;

A speech retrieval module, used to perform keyword retrieval based on the user input content and the speech graph, obtain speech tags and activate corresponding speech content;

A speech list module is used to construct a speech directed graph according to the speech label and the speech graph, and perform greedy optimization to obtain a speech priority list;

A speech output module, used to output all activated speech contents according to the speech priority list, and end the conversation after completing the speech output;

An emotional soothing module is used to evaluate the emotional state of the user based on the user input content, and when it is detected that the user is in a negative mood, output soothing words and end the conversation;

The topic convergence module is used to calculate the convergence degree according to the user input content, and end the conversation when the topic convergence degree is greater than a preset convergence degree threshold.

7. An electronic device, characterized in that it includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the big data-based algorithm AI customer service dialogue management method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the big data-based algorithm AI customer service dialogue management method as described in any one of claims 1 to 5.