Intelligent assistance method and system for call center human agents

Abstract

The application relates to the technical field of natural language processing, in particular to an intelligent assistance method and system suitable for call center human agents, which comprises the following steps: obtaining target dialogue text; taking the segmented words after removing stop words in the target dialogue text as target segmented words; obtaining all co-occurrence words of each target segmented word according to the stop word features and emotional intensity of the sentence where the target segmented word is located; obtaining the associated words of each target segmented word according to the co-occurrence frequency of each target segmented word and its co-occurrence words and the appearance frequency of each co-occurrence word; dividing all target segmented words into multiple entity sets according to the similarity of associated words and the similarity of pinyin between target segmented words, then obtaining the entity vocabulary of each entity set, and finally replacing all target segmented words with the entity vocabulary corresponding to the entity set where the target segmented word is located. The application accurately obtains the co-occurrence words of each target segmented word, thereby accurately obtaining the entity vocabulary corresponding to each target segmented word, and improving the error correction accuracy of the dialogue text.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of natural language processing, in particular to an intelligent assistance method and system suitable for call center human agents. BACKGROUND

[0002] In a multi-round dialogue scenario, the call center intelligent assistance system automatically performs speech recognition to generate dialogue record text, avoids the problem of human agent information forgetting in multi-round dialogue, and quickly extracts key information of customer consultation to provide accurate knowledge base recommendation for human agents.

[0003] The accuracy rate of speech recognition of the call center intelligent assistance system is affected by factors such as the speaker and the environment, and the recognition error rate of special nouns, homophones and infrequently used words is high in the process of speech to text, which leads to the existence of error information in the dialogue record text obtained by the call center intelligent assistance system, and the accuracy of providing knowledge assistance for human agents is low. The traditional language model can automatically correct the dialogue record text words through data-driven way, however, the dialogue record text of the call center is seriously colloquial, and the nested entities are easily truncated by stop words or emotional words, and the existing language model has insufficient long-distance dependence, and it is difficult to capture the features between the nested entity words across long-distance word intervals, so there is a problem of low accuracy rate when correcting errors. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide an intelligent assistance method and system suitable for call center human agents, and the technical solutions adopted are as follows:

[0005] In a first aspect, the embodiments of the present application provide an intelligent assistance method suitable for call center human agents, which comprises the following steps:

[0006] Obtaining target dialogue text in the dialogue process between the human agent and the customer;

[0007] According to the number of stop words in each sentence of the target dialogue text and the bit sequence difference of each adjacent stop word, the redundancy feature value of each sentence is obtained, and the oral evaluation value of each sentence is obtained in combination with the emotional intensity in each sentence, and then the co-occurrence interval length of each sentence is obtained; the segmented words after removing the stop words in the target dialogue text are recorded as target segmented words; according to the co-occurrence interval length of the sentence where each target segmented word is located, all co-occurrence words of each target segmented word are obtained;

[0008] According to the co-occurrence frequency of each target word and each co-occurrence word thereof, and the occurrence frequency of each co-occurrence word in the target dialogue text, an association word screening value between each target word and each co-occurrence word thereof is obtained, so as to obtain the association word of each target word; according to the number of same association words in the association words of any two target words, and the similarity degree of pinyin, all target words in the target dialogue text are divided into a plurality of entity sets; according to the average level of the spoken language evaluation value of all sentences in which each target word in each entity set is located, an entity vocabulary of each entity set is obtained, and then all target words in the target dialogue text are replaced by the corresponding entity vocabulary of the entity set in which the target word is located.

[0009] Preferably, the calculation formula of the redundancy feature value of each sentence is: ; in the formula, is the redundancy feature value of the jth sentence, is the number of stop words in the jth sentence, is the total number of all word pieces in the jth sentence, and respectively represent the position sequence value of the ith and (i-1)th stop words in the jth sentence.

[0010] Preferably, the calculation formula of the spoken language evaluation value of each sentence is: ; in the formula, is the spoken language evaluation value of the jth sentence, is the redundancy feature value of the jth sentence, is the emotional intensity value of the jth sentence, and Norm() is a normalization function; wherein, the process of obtaining the emotional intensity value of each sentence is: obtaining the emotional category and category intensity of all word pieces in the target dialogue text; taking the maximum value of the maximum value of the category intensity under all emotional categories of all word pieces of each sentence as the emotional intensity value of each sentence.

[0011] Preferably, the calculation formula of the co-occurrence interval length of each sentence is: ; in the formula, is the co-occurrence interval length of the jth sentence, is a preset initial interval length, is a rounding up function, is the spoken language evaluation value of the jth sentence, is a preset correction parameter.

[0012] Preferably, all co-occurrence words of each target word refer to all target words in the sentence in which the target word is located and the distance between which and the target word is not greater than the co-occurrence interval length of the sentence in which the target word is located.

[0013] Preferably, the calculation formula of the association word screening value between each target word and each co-occurrence word thereof is: In the formula, is a correlation word screening value between the target word x and its co-occurrence word y, is a co-occurrence frequency of the target word x and its co-occurrence word y, is a co-occurrence frequency of the target word x and its co-occurrence word y, and is a frequency of the co-occurrence word y in the target dialogue text; wherein the co-occurrence frequency of the target word x and its co-occurrence word y refers to a ratio of a co-occurrence number of the target word x and its co-occurrence word y to a total number of occurrences of the target word x in the target dialogue text.

[0014] Preferably, the correlation word of each target word refers to a co-occurrence word whose correlation word screening value is greater than a preset segmentation threshold value among all co-occurrence words of each target word.

[0015] Preferably, the specific process of dividing all target words in the target dialogue text into a plurality of entity sets is as follows:

[0016] A set of all correlation words of each target word is denoted as a correlation word set of each target word; a pronunciation string of each target word is obtained;

[0017] If the number of correlation words in the intersection of the correlation word sets of any two target words is greater than a preset first threshold value, the two target words satisfy a first judgment condition; if the edit distance between the pronunciation strings of any two target words is less than a preset second threshold value, the two target words satisfy a second judgment condition.

[0018] If any two target words simultaneously satisfy the first judgment condition and the second judgment condition, the two target words are determined to have the same entity relationship, and the two target words are attributed to the same entity set; all target words in the target dialogue text that have the same entity relationship are recorded as an entity set, and all entity sets in the target dialogue text are obtained.

[0019] Preferably, the process of obtaining the entity vocabulary of each entity set is as follows: the average of the colloquial evaluation values of all sentences in which each target word is located is taken as a recognition disturbance value of each target word; the target word with the smallest recognition disturbance value in each entity set is taken as the entity vocabulary of each entity set.

[0020] In a second aspect, the embodiments of the present application further provide an intelligent assistance system suitable for a call center human operator, which comprises a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the intelligent assistance method suitable for the call center human operator when executing the computer program.

[0021] The present application has at least the following beneficial effects:

[0022] The application analyzes the distribution characteristics of stop words and emotion grade characteristics in each sentence, constructs a spoken language evaluation value, and then adaptively adjusts the co-occurrence interval length of each sentence, gives a larger co-occurrence interval length to the sentence with strong emotion and significant stop word redundancy, effectively avoids the obvious spoken language representation characteristics in the dialogue process, improves the accuracy of co-occurrence word acquisition, and reduces the co-occurrence interval length for the sentence with insignificant spoken language characteristics, avoids unnecessary calculation complexity caused by too large co-occurrence interval length, solves the problem that the traditional language model is insufficient in long distance dependence and difficult to capture sentence nested entity vocabulary features across long distance vocabulary interval.

[0023] The application determines whether two target segmented words are the same entity set according to the similarity of the association word set and the edit distance of the pronunciation, and then obtains the entity vocabulary of each entity set, thereby correcting the errors of the record text with serious spoken language, effectively solving the problem of speech transcription errors caused by significant spoken language characteristics in the dialogue record text, improving the accuracy of the dialogue record text, and avoiding recognition errors to induce human agent retrieval errors, providing more accurate dialogue information for human agents. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0025] Figure 1 The step flow chart of the intelligent assistance method for human agents of call center provided by an embodiment of the present application is shown in the figure.

[0026] Figure 2 The acquisition flow chart of the entity vocabulary of each entity set provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0027] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following describes the intelligent assistance method and system for human agents of call center according to the present application, its specific implementation, structure, features and effects in detail, as follows. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0029] The specific scheme of the intelligent assistance method and system for manual agents of a call center provided in the present application will be specifically described below in combination with the drawings.

[0030] Please refer to Figure 1 which shows the step flowchart of the intelligent assistance method for manual agents of a call center provided in an embodiment of the present application, which comprises the following steps:

[0031] Step one: obtaining the target dialogue text in the dialogue process between the manual agent and the customer.

[0032] The call center microphone synchronously collects the voice of the customer and the manual agent, and the call center intelligent assistance system first filters the background noise through noise reduction processing, and then performs voice recognition on the dialogue between the customer and the manual agent in real time to obtain the dialogue record text.

[0033] In the dialogue process between the customer and the manual agent, every preset time is regarded as an error correction moment, and all dialogue record texts from the start time of the dialogue to the current error correction moment are recorded as the target dialogue text of the current error correction moment, and the target dialogue text corresponding to each error correction moment is corrected. The value range of the preset time is 10s-60s, and 20s is taken in the present embodiment. The target dialogue text is processed by using the Chinese word segmentation library jieba. Specifically, after importing the jieba Chinese word segmentation library, the jieba.cut function is used for word segmentation processing, wherein the HMM (Hidden Markov Model) parameter of the jieba.cut function is set to the default value True, and the cut_all parameter of the jieba.cut function is set to False, which is used to judge whether to use the full mode (the cut_all parameter is False by default, and cut_all=False represents the accurate mode of jieba segmentation). After jieba word segmentation processing, all the words in the target dialogue text are obtained.

[0034] Step two: obtaining the redundancy feature value of each sentence according to the number of stop words in each sentence of the target dialogue text and the bit sequence difference of each adjacent stop word, and obtaining the spoken language evaluation value of each sentence in combination with the emotional intensity in each sentence, and then obtaining the co-occurrence interval length of each sentence; the word segmentation after removing the stop words in the target dialogue text is recorded as the target word segmentation; obtaining all the co-occurrence words of each target word segmentation according to the co-occurrence interval length of the sentence where each target word segmentation is located.

[0035] The accuracy of voice recognition of the intelligent assistant system of the call center is affected by the speaker and the environment, etc. When the artificial seat of the call center provides services, the customer's daily colloquialization is serious, and the customer's emotion, accent, swallowing, unclear pronunciation, etc. will all cause the transcription error of voice recognition. Moreover, the call center application scenarios are mostly product after-sales, insurance consultation, e-commerce and other industries, there are some special vocabularies in some fields, and some special vocabularies belong to long word combination nested entities, customers are difficult to express completely and clearly, which causes the fragmentation of entities when customers express, which seriously interferes with the voice recognition of the intelligent assistant system of the call center.

[0036] The degree of redundancy of each sentence can be judged by the stop words in the sentence, therefore, whether all the segmented words in the target dialogue text belong to stop words is marked using the stop word table of Harbin Institute of Technology.

[0037] As a preferred embodiment, the redundancy feature value of each sentence is obtained according to the number of stop words in each sentence of the target dialogue text and the difference of the bit sequence of each adjacent stop word, which is used to represent the degree of redundancy of each sentence.

[0038] In this embodiment, the redundancy feature value of the jth sentence is denoted as , and its specific expression is: ; in the formula, is the redundancy feature value of the jth sentence, is the number of stop words in the jth sentence, is the total number of all segmented words in the jth sentence, and respectively represent the bit sequence value of the ith and the i-1th stop word in the jth sentence. It should be noted that when the jth sentence contains only one stop word, the redundancy feature value of the jth sentence is set to .

[0039] wherein, is used to reflect the proportion of stop words in the jth sentence, the greater the value, the more the proportion of stop words in the sentence, and the more redundant the sentence is; is used to reflect the distribution characteristics of stop words in the jth sentence, the greater the value, the more the stop words are concentrated in the whole sentence, the more the fragmented nested entities caused by the redundant stop words covering the corresponding nested entities of the special vocabulary, and the greater the degree of redundancy of the sentence.

[0040] Further, in the artificial sitting service process, the customer dialogue may contain more emotional words, which are often interspersed between sentence entities. The stronger the emotion, the stronger the influence on speech recognition. However, emotional words are often not in the stop table. All word segmentation of the target dialogue text is input into the sentence matching algorithm based on the Chinese emotion vocabulary ontology library, wherein the data label categories in the Chinese emotion vocabulary ontology library are respectively joy, good, anger, sadness, fear, evil and surprise. The category intensity under each emotion category can be divided into five levels of 1, 3, 5, 7 and 9. 9 represents the maximum intensity, and 1 represents the minimum intensity. The emotion category and category intensity of all word segmentation in the target dialogue text are obtained. The maximum value of the maximum value of the category intensity under all emotion categories corresponding to all word segmentation of each sentence is taken as the emotional intensity value of each sentence.

[0041] As a preferred embodiment, according to the redundancy feature value of each sentence and the emotional intensity in each sentence, the spoken language evaluation value of each sentence is obtained, which is used to represent the degree of spoken language of each sentence.

[0042] In this embodiment, the spoken language evaluation value of the jth sentence is denoted as , and the specific expression is: ; in the formula, is the spoken language evaluation value of the jth sentence, is the redundancy feature value of the jth sentence, is the emotional intensity value of the jth sentence, and Norm( ) is a normalization function.

[0043] The spoken language evaluation value reflects the degree of spoken language of the jth sentence. The larger the value, the higher the frequency of the appearance of emotional words, pause words and other stop words in the sentence, and the stronger the emotion. In the scenario where emotional modifiers and stop words truncate the sentence, the neighborhood length of the subsequent search co-occurrence word needs to be expanded as much as possible to avoid the problem that there is no co-occurrence word in a small search range.

[0044] Further, the target dialogue text is processed to remove stop words, and all word segmentation after removing stop words is denoted as target word segmentation. In the speech conversion text process, due to the strong spoken language of the speech, some target word segmentation in the converted text may be wrong, such as a target word segmentation in the target dialogue text is “share”, but the correct word in the speech is actually “box”, so the entity word of “share” is “box”. In order to correct the wrong target word segmentation in the target dialogue text to the correct word, the entity word of each target word segmentation can be obtained according to the similarity of the co-occurrence word and the similarity of the pronunciation of each target word segmentation.

[0045] Firstly, in order to accurately obtain all co-occurrence words of each target word in the target dialogue text, as a preferred embodiment, the co-occurrence interval length of each sentence is obtained according to the spoken language evaluation value of each sentence, which is used to represent the farthest interval length between each target word and its co-occurrence word.

[0046] In this embodiment, the co-occurrence interval length of the jth sentence is denoted as , and its expression is: ; in the formula, is the co-occurrence interval length of the jth sentence, is a preset initial interval length, which is 3 in this embodiment, is a rounding function, is the spoken language evaluation value of the jth sentence, is a preset correction parameter, which is 0.5 in this embodiment.

[0047] If is greater than 1, the dialogue sentence with strong emotion and significant stop word redundancy is given a larger co-occurrence interval length, which effectively avoids the influence of the obvious spoken language representation feature on the speech recognition accuracy in the dialogue process, expands the search length of the co-occurrence word, and is beneficial to mining the sentence features of the word segmentation and accurately obtaining the co-occurrence word of each target word; if is less than 1, the increase of the calculation complexity caused by the too large search length is avoided.

[0048] Further, taking the target word x in the target dialogue text as an example, all target words in the sentence where the target word x is located and whose interval distance with the target word x is not greater than the co-occurrence interval length of the sentence where the target word x is located are taken as the co-occurrence word of the target word x.

[0049] Similarly, the co-occurrence words of all target words in the target dialogue text are obtained.

[0050] Step three: according to the co-occurrence frequency between each target word and its co-occurrence word, and the occurrence frequency of each co-occurrence word in the target dialogue text, the association word screening value between each target word and its co-occurrence word is obtained, so as to obtain the association word of each target word; according to the number of the same association word in the association words of any two target words, and the similarity degree of the pinyin, all target words in the target dialogue text are divided into multiple entity sets; according to the average level of the spoken language evaluation value of all sentences where each target word in each entity set is located, the entity vocabulary of each entity set is obtained, and then all target words in the target dialogue text are replaced by the entity vocabulary corresponding to the entity set where the target word is located.

[0051] Further, according to the co-occurrence frequency of each target word and each co-occurrence word thereof, and the occurrence frequency of each co-occurrence word in the target dialogue text, an association word screening value of each target word and each co-occurrence word thereof is obtained, which is used to represent the association degree between each target word and each co-occurrence word thereof. The ratio of the co-occurrence times of the target word x and the co-occurrence word y in the target dialogue text to the total number of occurrences of the target word x is taken as the co-occurrence frequency of the target word x and the co-occurrence word y, denoted as .

[0052] In this embodiment, the association word screening value between the target word x and the co-occurrence word y is denoted as , and the specific expression is: ; in the formula, is the association word screening value between the target word x and the co-occurrence word y, is the co-occurrence frequency of the target word x and the co-occurrence word y, is the occurrence frequency of the co-occurrence word y in the target dialogue text.

[0053] The purpose of calculating is to prevent the co-occurrence word y from being easily misjudged as the association word of the target word x when the co-occurrence word y is a commonly used word; The greater the value is, the more significant the co-occurrence characteristics between the target word x and the co-occurrence word y are, and the lower the possibility of the co-occurrence word being a commonly used word is. Therefore, the co-occurrence word y is more likely to belong to the nested entity in the call center scenario, and the co-occurrence word y plays an important role in text correction for the target word x. Therefore, the co-occurrence word y should be taken as the segmentation association word of the target word x, and the association word screening value is greater.

[0054] Further, the association word screening values between each target word and all co-occurrence words thereof are taken as the input of the maximum inter-class variance OSTU algorithm to obtain a segmentation threshold, denoted as a preset segmentation threshold. The co-occurrence words with an association word screening value greater than the preset segmentation threshold are taken as the association words of each target word, and a set composed of all association words of each target word is denoted as the association word set of each target word.

[0055] In the process of converting voice to text in the call center, due to the influence of the customer's mood, accent, swallowing, unclear pronunciation, etc., the dialogue record text has the characteristics of strong emotional intention and stop word redundancy, and the dialogue record text has significant colloquial features, which can easily interfere with the voice recognition process, resulting in partial voice being incorrectly transcribed. In the dialogue record text, the same nested entity is incorrectly identified as a combination of multiple different words with similar pronunciation.

[0056] The Chinese character pinyin is composed of initial, final and tone. In the call center customer service scene, the customer's voice expression is more daily, and the voice height change is not sensitive. Therefore, the application takes each target word as input, uses the pypinyin library in the programming language Python, and adopts the NORMAL mode (normal style, composed of initial and final, without tone) for pinyin style, and outputs the pronunciation string of each target word.

[0057] Further, if the number of associated words in the intersection of the associated word sets of any two target words is greater than a preset first threshold value, the two target words are more likely to be words in the same nested entity, and the two target words satisfy the first judgment condition; wherein the value range of the preset first threshold value is [2, 4], and the embodiment is set to 2. If the edit distance between the pronunciation strings of any two target words is less than a preset second threshold value, the two target words are more likely to be homophones, and the two target words satisfy the second judgment condition; wherein the value range of the preset second threshold value is [2, 5], and the embodiment is set to 3.

[0058] If any two target words satisfy the first judgment condition and the second judgment condition at the same time, it is determined that the two target words have the same entity relationship, and then the two target words are attributed to the same entity set. It should be noted that if any two target words a and b have the same entity relationship, when the target word x also has the same entity relationship with the target word a or the target word b, the target word x has the same entity relationship with the target word a and the target word b, and the target words a, b and x belong to the same entity set.

[0059] Whether all arbitrary two target words in the target dialogue text have the same entity relationship is calculated, all target words with the same entity relationship in the target dialogue text are recorded as an entity set, and all entity sets in the target dialogue text are obtained.

[0060] For example, in the call center product after-sales field, the customer's real dialogue is "box fuse", and in the voice recognition transcription process, due to the serious colloquialization, "box" is converted into "box", "fly to" and "share" in the dialogue record text, then "box", "fly to" and "share" are an entity set for text correction, and the corresponding entity word in the entity set is "box".

[0061] The average of the colloquial evaluation values of all sentences containing the target word x is taken as the recognition interference value of the target word x, which reflects the degree of interference of the target word x in the speech recognition transcription process. The smaller the recognition interference value of the target word, the weaker the interference degree of the target word in the speech recognition transcription process caused by emotion, accent, elision, unclear pronunciation, etc. The higher the accuracy of the speech recognition transcription, the higher the confidence of the target word after the recognition transcription, and the error correction of the word usage in the call center scenario can be performed.

[0062] The target word with the smallest recognition interference value in each entity set is taken as the entity vocabulary of each entity set, and all target words in the target dialogue text are replaced with the entity vocabulary corresponding to the entity set where the target word is located. The flowchart for obtaining the entity vocabulary of each entity set is shown in Figure 2 In this way, the entity vocabulary with partial recognition errors in the target dialogue text in the multi-round dialogue scenario of the call center is corrected by the text correction, the accuracy of the target dialogue text is improved, the problem of artificial agent retrieval error induced by recognition error is avoided, and accurate intelligent assistance is provided for the artificial agent.

[0063] Based on the same inventive concept as the above method, the embodiments of the present application also provide an intelligent assistance system suitable for artificial agents in a call center, which includes a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the method described in the above intelligent assistance method suitable for artificial agents in a call center when executing the computer program.

[0064] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0065] Each embodiment in the present specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the differences from other embodiments.

[0066] The above is only the preferred embodiment of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent assistance method suitable for human agents in call centers, characterized in that, The method includes the following steps: Obtain the target dialogue text during the conversation between the human agent and the customer; Based on the number of stop words in each sentence of the target dialogue text and the positional differences of adjacent stop words, the redundancy feature value of each sentence is obtained. Combined with the intensity of emotion in each sentence, the spoken evaluation value of each sentence is obtained, and then the co-occurrence interval length of each sentence is obtained. The segmented words in the target dialogue text after removing stop words are recorded as target segments. All co-occurring words of each target segment are obtained based on the co-occurrence interval length of the sentences in which each target segment is located. Based on the co-occurrence frequency of each target word and its co-occurring words, and the occurrence frequency of each co-occurring word in the target dialogue text, the related word filtering value between each target word and its co-occurring words is obtained, thereby obtaining the related words of each target word; based on the number of identical related words in the related words of any two target words, and the similarity of their pinyin, all target words in the target dialogue text are divided into multiple entity sets; based on the average level of the spoken evaluation value of all sentences containing each target word in each entity set, the entity vocabulary of each entity set is obtained, and then all target words in the target dialogue text are replaced with the entity vocabulary corresponding to their respective entity sets.

2. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, The formula for calculating the redundancy feature value of each statement is as follows: In the formula, The redundant feature value of the j-th statement. Let j be the number of stop words in the j-th statement. Let be the total number of all words in the j-th statement. and These represent the position values ​​of the i-th and (i-1)-th stop words in the j-th statement, respectively.

3. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, The formula for calculating the oral assessment score of each statement is as follows: In the formula, Let j be the spoken evaluation value of the j-th statement. The redundant feature value of the j-th statement. Let be the sentiment intensity value of the j-th statement, and Norm() be the normalization function. The process of obtaining the sentiment intensity value of each statement is as follows: obtain the sentiment category and category intensity of all words in the target dialogue text; take the maximum value of the maximum category intensity of all sentiment categories corresponding to all words of each statement as the sentiment intensity value of each statement.

4. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, The formula for calculating the co-occurrence interval length of each statement is as follows: In the formula, Let j be the co-occurrence interval length of the j-th statement. It is the preset initial interval length. It is the floor function. Let j be the spoken evaluation value of the j-th statement. These are preset correction parameters.

5. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, All co-occurring words of each target word refer to all target words within the sentence in which each target word is located, whose distance from the target word is no greater than the co-occurrence interval length of the sentence in which it is located.

6. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, The formula for calculating the correlation word filtering value between each target word segment and its co-occurring words is as follows: In the formula, Filter the associated words between the target word segment x and its co-occurring word y. The co-occurrence frequency of the target word segment x and its co-occurring word y. It is the frequency of co-occurrence of word y in the target dialogue text; where the co-occurrence frequency of target word x and its co-occurrence word y is the ratio of the number of times target word x and its co-occurrence word y co-occur in the target dialogue text to the total number of times target word x appears.

7. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, The associated words of each target word segment refer to the co-occurring words among all co-occurring words of each target word segment whose associated word filtering value is greater than the preset segmentation threshold.

8. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, The specific process of dividing all target words in the target dialogue text into multiple entity sets is as follows: The set of all related words for each target word is denoted as the related word set for each target word; the pronunciation string of each target word is obtained. If the number of related words in the intersection of the related word sets of any two target word segments is greater than a preset first threshold, then the two target word segments satisfy the first judgment condition; if the edit distance between the pronunciation strings of any two target word segments is less than a preset second threshold, then the two target word segments satisfy the second judgment condition. If any two target word segments simultaneously satisfy the first and second judgment conditions, then the two target word segments are determined to be related to the same entity, and these two target word segments are assigned to the same entity set; all target word segments in the target dialogue text that are related to the same entity are recorded as an entity set, thus obtaining the entity set of all targets in the target dialogue text.

9. The intelligent assistance method for call center human agents as described in claim 1, characterized in that, The process of obtaining the entity vocabulary of each entity set is as follows: the average of the spoken evaluation values ​​of all sentences containing each target word is used as the recognition disturbance value of each target word; the target word with the smallest recognition disturbance value in each entity set is used as the entity vocabulary of each entity set.

10. An intelligent auxiliary system suitable for call center human agents, 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, it implements the steps of the intelligent assistance method for call center human agents as described in any one of claims 1-9.

Images