A customer service response interaction method and system based on natural language processing

Abstract

The present application relates to the field of natural language processing technology, and specifically to a customer service response interaction method and system based on natural language processing, the method comprising: obtaining user text information, as well as keywords and associated words therein; obtaining each local vocabulary vector of each keyword; obtaining the matching priority factor of each keyword; and obtaining a comprehensive score for each keyword in combination with the difference in the number of characters between each keyword and the remaining keywords, as well as the degree of position dispersion of each keyword in the user text information; obtaining the intention matching degree of each keyword in the user text information based on the position difference of all to-be-matched words corresponding to each keyword in the user text information, the comprehensive score of each keyword, and the degree of similarity between each keyword and all corresponding to-be-matched words, thereby obtaining priority matching words and responding to questions raised by the user. By extracting priority matching words from user text, the customer service response interaction experience is improved.

Description

Technical Field

[0001] The present application relates to the technical field of natural language processing, and in particular to a customer service response interaction method and system based on natural language processing. Background Art

[0002] Natural Language Processing (NLP), an interdisciplinary field in computer science, artificial intelligence, and linguistics, enables computers to understand, analyze, and generate textual information that is similar to human language. In the field of customer service interaction, NLP-driven intelligent customer service demonstrates significant advantages in response speed, processing power, and cost optimization. NLP-based customer service interaction systems, through technologies such as intent recognition and multi-round dialogue, can effectively improve service efficiency and user experience, and are widely used in finance, e-commerce, and government affairs.

[0003] Existing customer service response interaction methods still have certain technical drawbacks. Intelligent customer response interaction is prone to semantic deviations in complex sentence structures, making it difficult to accurately capture keywords that contain the user's true semantic meaning. Response interactions typically involve multiple rounds of dialogue, and existing natural language processing-based response interaction technologies are severely inadequate in handling contextual dependencies. This means that historical keyword information is lost during multiple rounds of dialogue, potentially causing logical gaps and making it impossible to accurately grasp the user's true intent. Answers made during the customer service response interaction lack conversational coherence, impacting the customer service response interaction experience. Therefore, a method is urgently needed to accurately screen keywords in user text messages that truly reflect user intent. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a customer service response interaction method and system based on natural language processing. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a customer service response interaction method based on natural language processing, the method comprising the following steps:

[0006] Obtain user text information during customer service response interaction;

[0007] Obtain keywords and related words in the user text information; obtain local vocabulary vectors for each keyword based on its adjacent keywords in the user text information; obtain matching priority factors for each keyword in the user text information based on the frequency, number of characters, distance to the nearest related words, and similarity between all corresponding local vocabulary vectors;

[0008] Obtain the inter-word association weight of each keyword based on the difference in the number of characters between each keyword and the other keywords in the user text information; obtain the word position weight of each keyword based on the degree of position dispersion of each keyword in the user text information, and combine the matching priority factor and inter-word association weight of each keyword to obtain the comprehensive score of each keyword;

[0009] Obtain the words to be matched for each keyword in the user text information in the knowledge base; obtain the word span of each keyword in the user text information based on the position difference of all the words to be matched corresponding to each keyword in the user text information; and combine the comprehensive score of each keyword and the similarity between all local word vectors of each keyword and the local word vectors of all corresponding words to be matched to obtain the intention matching degree of each keyword in the user text information, and then obtain the priority matching words to respond to the questions raised by the user.

[0010] Preferably, the process of obtaining the local vocabulary vectors of each keyword is: taking the vector composed of each word and the previous and next keywords in the sentence as the local vocabulary vector of each word, and recording the local vocabulary vector of each word corresponding to each keyword as the local vocabulary vector of each keyword.

[0011] Preferably, the calculation formula for the matching priority factor of each keyword in the user text information is: Where, is the matching priority factor of the i-th keyword in the user text information; The product of the frequency of the i-th keyword and the number of characters in the user's text information; is the number of characters between the ith keyword and its nearest related word in the user text information; is a preset constant; is the lexical dependency flexibility of the i-th keyword in the user text information; norm() is the normalization function; wherein, the lexical dependency flexibility of each keyword is obtained by accumulating the distance between all any two local vocabulary vectors of each keyword as the lexical dependency flexibility of each keyword.

[0012] Preferably, the inter-word association weight of each keyword is the absolute value of the difference between the number of characters of each keyword in the user text information and the average number of characters of all other keywords.

[0013] Preferably, the word position weight of each keyword is obtained by obtaining the position order of each word according to its position in the user text information; and taking the variance of the position order of all words corresponding to each keyword as the word position weight of each keyword.

[0014] Preferably, the comprehensive score of each keyword is obtained by taking the matching priority factor, inter-word association weight, and word position weight of each keyword as an evaluation index, and adopting the TOPSIS comprehensive evaluation method to obtain the comprehensive score of each keyword.

[0015] Preferably, the word span of each keyword in the user text information is the range of the position of all to-be-matched words corresponding to each keyword in the user text information in the knowledge base.

[0016] Preferably, the calculation formula for the intention matching degree of each keyword in the user text information is: Where, is the intention matching degree of the i-th keyword in the user's text information; is the word span of the i-th keyword in the user's text information; is the comprehensive score of the i-th keyword in the user's text information; It is the cumulative result of the Levenshtein distance between all local vocabulary vectors of the i-th keyword in the user text information and the local vocabulary vectors of all its corresponding words to be matched.

[0017] Preferably, the specific process of obtaining priority matching words to respond to questions raised by users is as follows: the sentences currently input by the user during a single customer service response interaction and the keywords in the historical response interaction text information whose intention matching degree is greater than or equal to a preset threshold are recorded as priority matching words; all sentences containing all to-be-matched words corresponding to all priority matching words in the knowledge base are used as input to the response generation interaction engine to generate answer sentences.

[0018] In the second aspect, an embodiment of the present application also provides a customer service response interaction system based on natural language processing, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-mentioned customer service response interaction methods based on natural language processing.

[0019] This application has at least the following beneficial effects:

[0020] 1. This application constructs a matching priority factor for each keyword by evaluating the semantic richness and lexical dependency flexibility of the keyword, which more accurately reflects the matching priority of the keyword in the user's response interaction text to the knowledge base content data;

[0021] 2. Based on the characteristics of keywords in user text messages, we obtain inter-word association weights and word position weights, and then generate a comprehensive score. Combined with the semantic contribution of each keyword, we can more accurately capture keywords that reflect the user's independent intentions and true semantics in complex sentence structures in response interactions, effectively avoiding the disadvantage of customers' semantic understanding of complex sentences in response interactions being biased.

[0022] 3. By analyzing the matching success rate between the to-be-matched words in the knowledge base and the user's actual intent, the intent matching degree of each keyword in the user's text information is constructed, and priority matching words are selected to generate customer service response interaction information. This improves the contextual coherence in the semantic understanding process of user response interactions, reduces the logical discontinuity caused by the loss of historical information, more accurately grasps the user's true intent, and improves the user response interaction experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0024] Figure 1 A flowchart of a customer service response interaction method based on natural language processing provided in one embodiment of the present application;

[0025] Figure 2 A flowchart for obtaining the intent matching degree of each keyword in user text information provided in one embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to further illustrate the technical means and effects adopted by this application to achieve the predetermined invention purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation method, structure, features and effects of a customer service response interaction method and system based on natural language processing proposed in this application. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0027] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0028] The following describes in detail a specific solution of a customer service response interaction method and system based on natural language processing provided by this application with reference to the accompanying drawings.

[0029] See also Figure 1 , which shows a flowchart of a customer service response interaction method based on natural language processing provided by an embodiment of the present application, the method comprising the following steps:

[0030] Step 1: Obtain user text information during the customer service response interaction process.

[0031] Specifically, the layered architecture of the customer service response interaction system in this application includes:

[0032] User interface layer: supports multi-channel access (webpage, APP, API interface), is responsible for interacting with users, receiving user input, and implementing standardized preprocessing of user input text. In this application, standardized preprocessing of user input text includes special character cleaning, removal of redundant spaces and non-ASCII characters.

[0033] Dialogue management layer: responsible for tracking and collecting interaction information between users and customer service, and implementing multi-round interaction data management by adding timestamps.

[0034] Knowledge base: Responsible for storing common enterprise questions, solutions, product information, etc. It can also be connected to existing open source databases through a network interface. The customer service response interaction system retrieves relevant information from the knowledge base and interacts with users.

[0035] Natural Language Processing (NLP) Engine: As the core of the system, it is responsible for understanding the multi-round interaction information between users and customer service representatives obtained by the dialogue management layer and converting it into structured data that can be processed by machines. This enables sentence structure understanding and sentiment analysis of the interaction text data.

[0036] Response generation interaction engine: Uses the natural language generation module (NLG) to match the understanding of user conversations with the information in the knowledge base and generate natural and fluent responses.

[0037] This application aims to improve the sentence structure understanding within the NLP engine of a customer service response system, enabling accurate capture of complex sentence structures and enhancing the contextual coherence of responses. The aforementioned dialogue management layer captures textual information from the intelligent customer service response interaction process. This textual information includes user text (the text of the user's question) and response sentences (the text of the customer service response system's response).

[0038] Step 2: Obtain keywords and related words in the user text information; obtain each keyword's local vocabulary vector based on the adjacent keywords in the user text information; obtain the matching priority factor of each keyword in the user text information based on the word frequency, number of characters, distance between the keyword and the nearest related word, and the similarity between all corresponding local vocabulary vectors.

[0039] With the continuous development of intelligent interactive response technology, user needs are no longer limited to simple questions, but have become more complex and diverse. Interactive sentences are often interspersed with complex sentence structures and contain a large number of words. Therefore, the decomposition of complex sentence structures and the prioritization of keyword matching with knowledge base content are crucial steps in user interaction semantic recognition. They can effectively capture user needs and improve the interactive response service experience.

[0040] Specifically, in the customer service response interaction process, the longer the number of characters in the user's keyword, the more likely it is to contain richer semantics, and can provide more dimensional intent features in the subsequent knowledge base content matching. The keyword frequency can reflect the customer's key semantic needs; the positional relationship between keywords and related words in complex sentences can also represent the user's true semantics and emotions, that is, the closer the distance between keywords and related words, the stronger the correlation, and the more accurately the user's true intention can be captured; at the same time, the local vocabulary collocation differences of keyword text entities can better reflect the richness of the user's response interaction context.

[0041] All user text information in a single customer service response interaction process is used as input, and the jieba word segmentation tool is used to segment the user text information, obtain all keywords and related words in the user text information, and count the word frequency and number of characters of each keyword.

[0042] Since a keyword may appear multiple times in user text information, a keyword will correspond to multiple words. The vector formed by each word and the previous and next keywords in the sentence is used as the local word vector of each word, and the local word vector of each word corresponding to each keyword is recorded as the local word vector of each keyword. It should be noted that when the position containing the word is at the beginning or end of the sentence, the corresponding local word vector only contains two keywords. For example: the user text information is: "If there is a problem with the quality of the received goods and you want a refund, please inform me of the specific refund process for the goods." Among them, the local word vectors of the text entity corresponding to "goods" are (arrival, goods, quality) and (information, goods, refund); since the word after "quality" is the conjunction "if", the keyword after the conjunction is selected to form the local word vector, so the local word vector of "quality" is (goods, quality, existence).

[0043] As a preferred implementation, the matching priority factor of each keyword in the user text information is obtained based on the word frequency, number of characters, distance to the nearest related words, and similarity between all corresponding local vocabulary vectors of each keyword in the user text information, which is used to characterize the matching priority status of the keywords used in the user interactive response with the knowledge base information in the process of intent understanding.

[0044] In this embodiment, the matching priority factor of the i-th keyword in the user text information is recorded as , and its specific expression is: Where, is the matching priority factor of the i-th keyword in the user text information; The product of the frequency of the i-th keyword and the number of characters in the user's text information; is the number of characters between the ith keyword and its nearest related word in the user text information; is a preset constant, which is taken as 0.1 in this embodiment to prevent the denominator from being 0; is the lexical dependency flexibility of the i-th keyword in the user text information; norm() is the normalization function, so that The value range of is between [0,1]. This embodiment selects the sigmoid() function as the normalization function. The lexical dependency flexibility of keywords is obtained by accumulating the Levenshtein distance between all two local vocabulary vectors corresponding to each keyword as the lexical dependency flexibility of each keyword. When a keyword has only one local vocabulary vector, the lexical dependency flexibility of the keyword is 0.1. The calculation of the Levenshtein distance is a well-known technology, and the specific process will not be repeated here.

[0045] The keyword's matching priority factor reflects the keyword's vocabulary matching priority with the knowledge base content data during the user's response interaction process; lexical dependency flexibility effectively represents the flexibility of the keyword's local vocabulary collocation, further reflecting the richness of the user's response interaction context. The greater the keyword's word frequency and number of characters, and the smaller the distance between the keyword and related words in the text information, that is, the larger and smaller the index, the higher the priority of the keyword when matching with the knowledge base data content; the greater the difference in the keyword's local vocabulary vector, that is, the larger the index, the better the keyword can reflect the response interaction context, and the higher the matching priority should be given to the keyword.

[0046] At this point, the matching priority factor of each keyword in the user text information can be obtained through the above method.

[0047] Step 3: Obtain the inter-word association weight of each keyword based on the difference in the number of characters between each keyword and other keywords in the user text information; obtain the word position weight of each keyword based on the degree of position dispersion of each keyword in the user text information, and combine the matching priority factor and inter-word association weight of each keyword to obtain the comprehensive score of each keyword.

[0048] The absolute difference between the number of characters in each keyword and the average number of characters in all other keywords in the user's text is recorded as the inter-word association weight of each keyword. The inter-word association weight reflects the independent intention of the keyword during the user's response interaction. The larger the inter-word association weight, the more likely the keyword contains the user's true meaning.

[0049] The positional order of each word is determined based on its position in the user's text. When a keyword corresponds to multiple words, it also corresponds to multiple positions. The more chaotic the positions corresponding to a keyword are, the more flexible its position distribution is within the user's text, indicating a higher semantic contribution. Therefore, the variance of the positions of all words corresponding to each keyword is used as the word position weight for each keyword. The word position weight reflects the higher semantic contribution of a keyword within the user's text.

[0050] The matching priority factor, inter-word association weight, and word position weight of each keyword are used as evaluation indicators respectively, and the TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) comprehensive evaluation method is used to obtain the comprehensive score of each keyword. The higher the comprehensive score, the more the keyword can represent the user's real interaction needs and the more it should be given priority in matching the knowledge base information. Among them, the matching priority factor, inter-word association weight, word position weight and comprehensive score are all positively correlated, and the weights of the three can be determined by the entropy weight method. Since the TOPSIS comprehensive evaluation method and the entropy weight method are both well-known technologies, the specific acquisition process will not be elaborated on.

[0051] Step 4: Obtain the words to be matched for each keyword in the user text information in the knowledge base; obtain the word span of each keyword in the user text information based on the position difference of all the words to be matched corresponding to each keyword in the user text information; and combine the comprehensive score of each keyword and the similarity between all local word vectors of each keyword and the local word vectors of all corresponding words to be matched to obtain the intention matching degree of each keyword in the user text information, and then obtain the priority matching words to respond to the questions raised by the user.

[0052] When matching keywords with knowledge base information, there are multiple possible matches for the words being matched. Incorrect matches can lead to ineffective responses to user requests, impacting the interactive experience. Therefore, strict attention must be paid to the word span and local vocabulary vectors of the words being matched in the knowledge base to improve the customer service response system's ability to grasp the user's true intent and maintain conversational coherence.

[0053] Retrieve the vocabulary from the knowledge base that matches a single keyword in the user's text and record it as the matching vocabulary for the single keyword in the user's text. A single keyword in the user's interactive text corresponds to multiple matching vocabulary in the knowledge base. The range of the positional difference between all matching vocabulary corresponding to each keyword in the user's text in the knowledge base is recorded as the word span of each keyword in the user's text.

[0054] In the text information stored in the knowledge base, the higher the word span of the to-be-matched vocabulary, the more the to-be-matched vocabulary can reflect the theme of the text, the greater the semantic contribution of the to-be-matched vocabulary to the knowledge base information, and the more likely it is to contain the user's true intention logic; and the more similar the vocabulary dependency conditions are, the higher the success rate of matching the knowledge base information with the user's interaction needs. When the relevant information of the to-be-matched vocabulary in the knowledge base is used as the customer service response, it is more effective to avoid the interaction logic fault and accurately respond to the user's interaction needs.

[0055] As a preferred embodiment, the intention matching degree of each keyword in the user text information is obtained based on the word span, comprehensive score of each keyword in the user text information, and the similarity between all local vocabulary vectors of each keyword and the local vocabulary vectors of all corresponding words to be matched, which is used to represent the user intention matching success status of each keyword in the process of matching with the knowledge base content. The flowchart for obtaining the intention matching degree of each keyword in the user text information is as follows: Figure 2 shown.

[0056] In this embodiment, the intention matching degree of the i-th keyword in the user text information is recorded as , its specific expression is: Where, is the intention matching degree of the i-th keyword in the user's text information; is the word span of the i-th keyword in the user's text information; is the comprehensive score of the i-th keyword in the user's text information; It is the cumulative result of the Levenshtein distance between all local vocabulary vectors of the i-th keyword in the user text information and the local vocabulary vectors of all its corresponding words to be matched.

[0057] The intent match reflects the success of the match between the to-be-matched vocabulary in the knowledge base and the user's actual intent. The word span characterizes the semantic contribution of the to-be-matched vocabulary in the knowledge base. The difference between the local word vectors corresponding to the keyword and the to-be-matched vocabulary reflects the lexical dependency similarity of the user's actual intent. The more closely the keywords in the user's text message reflect the user's true intent, the more prioritized they are for matching with the knowledge base content as the customer response. The larger the word span between the keywords and the to-be-matched vocabulary in the knowledge base, the greater the calculation index, the higher the keyword's overall score, and the smaller the difference between the local word vectors of the keywords and the to-be-matched vocabulary.

[0058] The intent matching degree of all keywords in the user text information is obtained through the above method.

[0059] Furthermore, the intent matching scores of all keywords in the user's current input and historical interaction text during a single customer service response are used as input. The maximum inter-class variance is used to determine the segmentation threshold. This threshold is recorded as the preset threshold, and keywords with an intent matching score greater than or equal to the preset threshold are recorded as priority matching words. The maximum inter-class variance is a well-known technique, and the specific acquisition process is not further described.

[0060] The sentences containing all the matching words in the knowledge base corresponding to the priority matching words are used as input to the response generation engine of the customer service response interaction system. Natural language generation (NLG) models are used to generate natural and fluent response sentences. The acquisition of NLG models is a well-known technology, and the specific process will not be elaborated on here.

[0061] Based on the same inventive concept as the above method, an embodiment of the present application also provides a customer service response interaction system based on natural language processing, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned customer service response interaction methods based on natural language processing are implemented.

[0062] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0063] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0064] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A customer service response interaction method based on natural language processing, characterized in that: The method comprises the following steps: Obtain user text information during customer service response interaction; Obtain keywords and associated words in the user text information; obtain local vocabulary vectors of each keyword based on adjacent keywords of each keyword in the user text information; obtain a matching priority factor of each keyword in the user text information based on the word frequency, number of characters, distance between the keyword and the nearest associated word, and similarity between all corresponding local vocabulary vectors. The calculation formula for the matching priority factor of each keyword in the user text information is: Where, is the matching priority factor of the i-th keyword in the user text information; The product of the frequency of the i-th keyword and the number of characters in the user's text information; is the number of characters between the ith keyword and its nearest related word in the user text information; is a preset constant; is the lexical dependency flexibility of the i-th keyword in the user text information; norm() is the normalization function; wherein, the lexical dependency flexibility of each keyword is obtained by accumulating the distance between all two local vocabulary vectors of each keyword as the lexical dependency flexibility of each keyword; Obtain the inter-word association weight of each keyword based on the difference in the number of characters between each keyword and the other keywords in the user text information; obtain the word position weight of each keyword based on the degree of position dispersion of each keyword in the user text information, and combine the matching priority factor and inter-word association weight of each keyword to obtain the comprehensive score of each keyword; Obtain the words to be matched in the knowledge base for each keyword in the user text information; obtain the word span of each keyword in the user text information based on the positional differences of all the words to be matched corresponding to each keyword in the user text information; and combine the comprehensive score of each keyword and the similarity between all local word vectors of each keyword and the local word vectors of all corresponding words to be matched to obtain the intention matching degree of each keyword in the user text information, and then obtain the priority matching words to respond to the questions raised by the user.

2. A customer service response interaction method based on natural language processing according to claim 1, characterized in that: The process of obtaining the local vocabulary vectors of each keyword is as follows: the vector formed by each word and the previous and next keywords in the sentence is used as the local vocabulary vector of each word, and the local vocabulary vector of each word corresponding to each keyword is recorded as the local vocabulary vector of each keyword.

3. The customer service response interaction method based on natural language processing according to claim 1, characterized in that: The inter-word association weight of each keyword is the absolute value of the difference between the number of characters of each keyword in the user text information and the average number of characters of all other keywords.

4. The customer service response interaction method based on natural language processing according to claim 1, characterized in that: The word position weight of each keyword is obtained by obtaining the position order of each word according to its position in the user text information; and taking the variance of the position order of all words corresponding to each keyword as the word position weight of each keyword.

5. The customer service response interaction method based on natural language processing according to claim 1, characterized in that: The comprehensive score of each keyword is obtained by taking the matching priority factor, inter-word association weight, and word position weight of each keyword as an evaluation index, and adopting the TOPSIS comprehensive evaluation method to obtain the comprehensive score of each keyword.

6. The customer service response interaction method based on natural language processing according to claim 1, characterized in that: The word span of each keyword in the user text information is the range of the position of all to-be-matched words corresponding to each keyword in the user text information in the knowledge base.

7. The customer service response interaction method based on natural language processing according to claim 1, characterized in that: The calculation formula for the intention matching degree of each keyword in the user text information is: Where, is the intention matching degree of the i-th keyword in the user's text information; is the word span of the i-th keyword in the user's text information; is the comprehensive score of the i-th keyword in the user's text information; It is the cumulative result of the Levenshtein distance between all local vocabulary vectors of the i-th keyword in the user text information and the local vocabulary vectors of all its corresponding words to be matched.

8. The customer service response interaction method based on natural language processing according to claim 1, characterized in that: The specific process of obtaining the priority matching words to answer the questions raised by the user is as follows: the keywords whose intention matching degree of the current sentence input by the user and the historical answer interaction text information is greater than or equal to a preset threshold during the single customer service answer interaction are recorded as the priority matching words; All sentences containing all the to-be-matched words corresponding to all the priority matching words in the knowledge base are used as input to the response generation interaction engine to generate answer sentences.

9. A customer service response interaction system based on natural language processing, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the customer service response interaction method based on natural language processing as described in any one of claims 1 to 8 are implemented.

Images