Communication sentence recognition method and device, computer device and readable storage medium

Abstract

This application proposes a communication statement recognition method, apparatus, computer device, and readable storage medium. The method includes: splitting communication text converted from current communication audio into multiple first communication statements; selecting multiple second communication statements from the multiple first communication statements based on attribute information corresponding to each first communication statement; determining the target complaint tendency of each second communication statement based on each second communication statement and a preset classification model; and determining the overall complaint tendency of the current communication audio based on the target complaint tendency of each of the multiple second communication statements. The technical solution of this application can automatically determine the level of complaint tendency in communication audio, reducing the manual and time costs consumed in determining complaint tendency, while improving the accuracy of complaint tendency recognition.

Description

[Technical Field]

[0001] This application relates to the field of semantic recognition technology, and in particular to a method and apparatus for recognizing communication statements, a computer device, and a readable storage medium. [Background Technology]

[0002] In existing online sales systems, during communication between customer service personnel and callers, it's inevitable that callers may experience emotional fluctuations due to unmet expectations, and these fluctuations reflect the caller's tendency to complain. Correspondingly, the language used in communication between customer service personnel and callers can reveal the emotional fluctuations of both parties, thus reflecting the caller's level of complaint tendency to some extent.

[0003] In response, relevant technologies typically require customer service personnel or additional staff to manually mark callers who have a tendency to complain or a high tendency to complain. However, due to the huge workload involved in online sales operating systems, manual marking is time-consuming, labor-intensive, inefficient, and prone to omissions and errors.

[0004] Therefore, efficiently and accurately identifying the level of complaint tendency of callers has become an urgent technical problem to be solved. [Summary of the Invention]

[0005] This application provides a communication statement recognition method and apparatus, computer equipment and readable storage medium, aiming to solve the technical problem that the methods for identifying the level of complaint tendency of callers in related technologies are time-consuming, laborious and inaccurate.

[0006] In a first aspect, embodiments of this application provide a method for recognizing communication statements, including:

[0007] The communication text converted from the current communication audio is split into multiple first communication statements;

[0008] Based on the attribute information corresponding to each first communication statement, multiple second communication statements are selected from the multiple first communication statements. The attribute information includes at least one of the following: complaint keywords, speech rate, maximum volume value within the statement, and average volume value within the statement.

[0009] Based on each second communication statement and a preset classification model, the target complaint tendency of each second communication statement is determined. The preset classification model is trained based on the attribute information of historical communication statements corresponding to historical complaint audio and the target complaint tendency of the historical statements.

[0010] Based on the target complaint tendency of each of the multiple second communication statements, the overall complaint tendency of the current communication audio is determined.

[0011] Secondly, embodiments of this application provide a communication statement recognition device, including:

[0012] The communication statement splitting unit is used to split the communication text obtained by converting the current communication audio into multiple first communication statements;

[0013] A communication statement filtering unit is used to filter out multiple second communication statements from the multiple first communication statements based on the attribute information corresponding to each first communication statement, wherein the attribute information includes at least one of complaint-related keywords, speech rate, maximum volume value within the statement, and average volume value within the statement;

[0014] A communication statement classification unit is used to determine the target complaint tendency of each second communication statement based on each second communication statement and a preset classification model, wherein the preset classification model is trained based on the attribute information of historical communication statements corresponding to historical complaint audio and the target complaint tendency of the historical communication statements;

[0015] The complaint tendency unit is used to determine the overall complaint tendency of the current communication audio based on the target complaint tendency of each of the plurality of second communication statements.

[0016] Thirdly, embodiments of this application provide a computer device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being configured to perform the method described in the first aspect above.

[0017] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions for performing the method described in the first aspect above.

[0018] The above technical solution addresses the problems of time-consuming, labor-intensive, and inaccurate methods for identifying the complaint tendency of callers in related technologies. It can automatically determine the complaint tendency of communication audio, reducing the manual and time costs consumed in identifying complaint tendency. Furthermore, due to the introduction of multiple attribute information, the complaint tendency reflected by multiple dimensions of these attributes can be comprehensively and accurately determined automatically to assess the complaint tendency of communication audio. [Attached Image Description]

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of a communication statement recognition method according to an embodiment of this application is shown;

[0021] Figure 2 A block diagram of a communication statement recognition device according to an embodiment of this application is shown;

[0022] Figure 3 A block diagram of a computer device according to one embodiment of this application is shown;

[0023] Figure 4 A block diagram of a computer device according to one embodiment of this application is shown.

Detailed Implementation Methods

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Figure 1 A flowchart of a communication statement recognition method according to an embodiment of this application is shown.

[0026] like Figure 1 As shown, a communication statement recognition method according to an embodiment of this application includes:

[0027] Step 102: The communication text converted from the current communication audio is split into multiple first communication statements.

[0028] To address the technical problems of time-consuming, labor-intensive, and inaccurate methods for identifying the complaint tendency of callers in existing technologies, the first step is to automate this identification process. To this end, this application proposes converting the audio of the current communication between customer service personnel and callers into text, and then automatically processing this text information to improve the ease of identifying the complaint tendency of callers.

[0029] The current communication audio includes real-time recordings of two- or multi-party calls made through mobile phones, landlines, and other communication devices, as well as audio generated from online voice communication between customers and customer service personnel through real-time chat tools provided by apps or web pages. This audio reflects the content of the communication between the two parties, which is an important basis for identifying the level of customer complaint tendency.

[0030] Specifically, the current audio communication can first be identified as text communication, and then the text communication can be broken down into multiple first-level communication statements. Specifically, the text communication can be broken down into statements based on its semantics and the pauses in the corresponding audio content.

[0031] Step 104: Based on the attribute information corresponding to each first communication statement, select multiple second communication statements from the multiple first communication statements.

[0032] The attribute information includes at least one of the following: complaint keywords, speech rate, maximum volume within a sentence, and average volume within a sentence.

[0033] Complaint-related keywords include, but are not limited to, words directly related to complaints such as "complaint," "report," and "superior," as well as uncivilized language with insulting tendencies, and threatening or suggestive phrases like "you just wait and see." Furthermore, faster speaking speed and higher peak and / or average volume within a sentence indicate a louder voice, greater urgency, and more agitated emotions, thus indicating a higher tendency to complain. In other words, complaint-related keywords, speaking speed, peak and average volume within a sentence can each reflect the level of a customer's complaining tendency in terms of content and tone, and therefore can serve as a preliminary basis for judging the level of complaining tendency.

[0034] Among them, the multiple second communication statements selected from the multiple first communication statements are statements whose attribute information reflects a high tendency to complain.

[0035] In one possible design, the specific filtering process for selecting multiple second communication statements from the plurality of first communication statements is as follows:

[0036] First, based on the attribute information corresponding to each first communication statement and the weight of the attribute information, the initial complaint tendency corresponding to each attribute information of the first communication statement is calculated.

[0037] Optionally, for a single first communication statement, the product of the feature value and the corresponding weight of each attribute information can be used as the initial complaint tendency of that attribute information. Here, the weight of the attribute information reflects the degree of influence of that attribute information on the level of complaint tendency, while the feature value of the attribute information reflects the actual participation degree of that attribute information in that first communication statement. Therefore, the initial complaint tendency of a single attribute information in a single first communication statement reflects the degree of participation of that single attribute information in that single first communication statement based on its own influence on the level of complaint tendency.

[0038] Next, for any of the first communication statements, optionally, if the initial complaint tendency of any attribute information of the first communication statement is greater than or equal to a first threshold, the first communication statement is determined as the second communication statement. The first threshold is the minimum initial complaint tendency required for a single attribute information of a single first communication statement to have a sufficiently high impact on the complaint tendency of that single first communication statement.

[0039] Optionally, for any of the first communication statements, if the initial complaint tendency of each attribute information of the first communication statement is greater than or equal to the first threshold, the first communication statement is determined as the second communication statement.

[0040] Optionally, for any of the first communication statements, if the average initial complaint tendency of all attribute information of the first communication statement is greater than or equal to a second threshold, the first communication statement is determined as the second communication statement. The second threshold is the minimum average initial complaint tendency required to achieve a sufficiently high average impact of all attribute information of a single first communication statement on the complaint tendency of that single first communication statement.

[0041] In one possible design, the specific method for calculating the initial complaint tendency corresponding to each attribute information of the first communication statement includes:

[0042] For any attribute information of any first communication statement, the product of the feature value of the attribute information and the weight of the attribute information is obtained as a first parameter; then the ratio of the initial complaint tendency of the attribute information in historical communication statements to the actual target complaint tendency of the historical communication statements is obtained as a second parameter; finally, the first parameter is corrected using the second parameter as a correction coefficient to obtain the initial complaint tendency of the attribute information.

[0043] The first parameter is the product of the feature value and the weight of the attribute information, reflecting the comprehensive performance of the attribute information in two dimensions: its actual participation in the first communication statement and its influence on the level of complaint tendency. The second parameter is the ratio of the initial complaint tendency of the attribute information in historical communication statements to the actual target complaint tendency of the historical communication statements, reflecting the deviation between the initial and actual complaint tendency of the attribute information in historical communication statements. Therefore, the second parameter is used to correct the first parameter, so as to correct the current initial complaint tendency by the deviation between the initial and actual complaint tendency of the attribute information in historical data, making the current initial complaint tendency closer to the final calculated actual complaint tendency.

[0044] Step 106: Based on each second communication statement and the preset classification model, determine the target complaint tendency for each second communication statement.

[0045] The preset classification model is trained based on the attribute information of the historical communication statements corresponding to the historical complaint audio and the target complaint tendency of the historical communication statements. It includes, but is not limited to, any natural language processing model with language classification function such as BERT classification model, GPT-1 model, GPT-2 model, GPT-3 model, fastText model, textCNN model, charCNN model, Bi-LSTM model, Bi-LSTM+Attention model, RCNN model, AdversarialLSTM model, Transformer model, ELMO pre-trained model, etc.

[0046] Optionally, historical communication statements involving historical complaints are used as positive samples to train a preset classification model, and the preset classification model is used to calculate the complaint tendency of the second communication statement in the current communication audio.

[0047] Optionally, while using historical communication statements involving historical complaints as positive samples, historical communication statements without historical complaints can also be used as negative samples for training.

[0048] Furthermore, the methods for training the preset classification model include:

[0049] First, obtain a first set of historical statements, wherein each historical communication statement in the first set of historical statements has complaint-related keywords and complaint tendency indicators.

[0050] Optionally, if a complaint is associated with a historical communication statement, it can be automatically marked with a complaint tendency indicator.

[0051] Optionally, if historical communication statements contain complaint-related keywords, or if other attribute information meets the criteria for historical complaints, a complaint tendency indicator can be automatically added to them.

[0052] Next, the first set of historical statements is expanded into a second set of historical statements, wherein the second set of historical statements includes the first set of historical statements, and the number of statements in the second set of historical statements is greater than the number of statements in the first set of historical statements.

[0053] Optionally, the first set of historical statements can be copied a specified multiple to obtain a second set of historical statements. The specified multiple includes, but is not limited to, 5 times.

[0054] Then, historical communication statements that meet the preset filtering rules are selected from the second set of historical statements to obtain the third set of historical statements.

[0055] The expanded second set of historical sentences often includes sentences that are too similar to the original sentences in the first set of historical sentences, or sentences that are semantically incoherent and illogical. Therefore, the expanded second set of historical sentences can be deredundant.

[0056] In one possible design, historical communication statements with a semantic confidence level below a third threshold can be deleted from the second set of historical statements. Semantic confidence level reflects the semantic fluency of historical communication statements, and the third threshold is the minimum semantic confidence level that a historical communication statement must meet to be semantically fluent. Once the semantic confidence level falls below the third threshold, it indicates that the historical communication statement is semantically incoherent and must be deleted.

[0057] In another possible design, historical communication statements with a similarity greater than or equal to a fourth threshold that are in the second historical statement set can be deleted.

[0058] When the similarity between statements is too high, the two statements are considered homogeneous. Therefore, a fourth threshold can be set as the minimum similarity for homogeneity between the two statements. If the similarity between a historical communication statement in the second historical statement set and any statement in the first historical statement set is greater than or equal to the fourth threshold, it indicates that the two are almost identical. To improve the conciseness of the training samples, this historical communication statement in the second historical statement set can be deleted.

[0059] In one possible design, firstly, all statements in the first historical statement set are retained in the second historical statement set, and then the statements in the second historical statement set other than all statements in the first historical statement set are compared one by one with all statements in the first historical statement set.

[0060] In another possible design, similarity calculation and comparison can be performed on every two statements in the second set of historical statements.

[0061] In another possible design, historical communication statements with semantic confidence levels below a third threshold can be deleted from the second historical statement set, while historical communication statements with similarity to any statement in the first historical statement set greater than or equal to a fourth threshold can also be deleted from the second historical statement set.

[0062] The above technical solutions can not only expand the sample data used for training the model, but also effectively remove redundancy from the sample data, improve the rationality of the sample data, and help train a more accurate and reliable pre-defined classification model.

[0063] Finally, based on the attribute information and target complaint tendency of each historical communication statement in the third historical statement set, the preset classification model is trained.

[0064] In one possible design, for each historical communication statement in the third historical statement set, each attribute information of that historical communication statement can be converted into a feature value, and then the feature values ​​are normalized to make them of the same magnitude. Finally, the normalized feature values ​​of all attribute information of each historical communication statement are used as input samples for a preset classification model, and the historical complaint results of each historical communication statement are used as output samples for the preset classification model to train the preset classification model.

[0065] Optionally, if the historical communication statement is in the historical communication audio where a complaint occurred, its historical complaint result is 1; otherwise, if the historical communication statement is in the historical communication audio where no complaint occurred, its historical complaint result is 0.

[0066] Step 108: Determine the overall complaint tendency of the current communication audio based on the target complaint tendency of each of the multiple second communication statements.

[0067] The target complaint tendency of each second communication statement reflects its own relevance to the occurrence of a complaint. Therefore, the overall complaint tendency calculated from the target complaint tendency of all second communication statements can reflect the relevance of the current communication audio to which all second communication statements belong to the occurrence of a complaint. This relevance to the occurrence of a complaint is precisely the manifestation of the possibility of a complaint.

[0068] The above technical solution combines attributes such as complaint-related keywords, speech rate, maximum volume, and average volume within sentences to filter communication statements for complaint tendency calculation. Then, a pre-defined classification model trained using these attributes calculates the complaint tendency of each selected statement in the current communication audio. Finally, the overall complaint tendency of the current communication audio is calculated based on the individual complaint tendency of each selected statement. This automatically determines the level of complaint tendency in communication audio, reducing the manual and time costs associated with complaint tendency assessment. Furthermore, the introduction of multiple attributes allows for a more accurate and comprehensive automatic determination of the complaint tendency of communication audio by integrating the complaint tendency reflected across multiple dimensions of these attributes.

[0069] On the one hand, this technical solution can be applied to real-time communication processes to calculate the complaint tendency of the content communicated in real time, so that customer service personnel can adjust their communication strategies in a timely manner based on the complaint tendency obtained in real time, in order to appease customers' emotions.

[0070] On the other hand, this technical solution can be applied to the analysis of historical communication audio, which makes it easier to accurately identify the level of complaint tendency in historical communication audio, so as to conduct timely after-sales communication, soothe customers' emotions, and improve the service experience.

[0071] Furthermore, the step of determining the overall complaint tendency of the current communication audio based on the target complaint tendency of each of the plurality of second communication statements specifically includes:

[0072] First, for each second communication statement, the following additional information is obtained: the number of complaint-related keywords in the second communication statement, the ratio of the speaking speed to the preset moderate speaking speed, the ratio of the highest volume value in the statement to the preset moderate volume, and the ratio of the average volume value in the statement to the preset moderate volume.

[0073] Among them, the additional information can more deeply reflect the degree of influence of the associated attribute information on the level of complaint tendency.

[0074] Next, a complaint tendency matrix is ​​generated based on all attribute information and all additional information of the second communication statement.

[0075] The eigenvalue of the element in the nth row and mth column of the complaint tendency matrix is ​​the square root of the product of the third parameter and the fourth parameter.

[0076] The third parameter is the ratio of the product of the nth attribute information and the mth additional information of the second communication statement to the target complaint tendency of the second communication statement, reflecting the influence level of the combination of a single attribute information and a single additional information on the target complaint tendency of the second communication statement. The fourth parameter is the ratio of the initial complaint tendency of the attribute information in historical communication statements to the actual target complaint tendency of the historical communication statements, reflecting the deviation level between the initial complaint tendency of the attribute information in historical communication statements and the actual complaint tendency.

[0077] Furthermore, the product of the third and fourth parameters can reflect the combined influence of individual attribute information and individual supplementary information on the level of complaint tendency after correcting for historical biases caused by that attribute information. Obtaining the square root of the product of the third and fourth parameters can, to some extent, reduce the error inherent in the product of the third and fourth parameters. This error represents the calculation error of the combined influence of individual attribute information and individual supplementary information on the level of complaint tendency after correcting for historical biases caused by that attribute information.

[0078] Finally, the complaint tendency matrices of all the second communication statements are concatenated into an overall tendency matrix, and the overall complaint tendency of the current communication audio is determined based on the overall tendency matrix.

[0079] Optionally, the rank of the overall tendency matrix can be set to the overall complaint tendency.

[0080] Optionally, if the rank of the overall tendency matrix is ​​greater than or equal to a specified threshold, the overall complaint tendency is set to 1; otherwise, if the rank of the overall tendency matrix is ​​less than the specified threshold, the overall complaint tendency is set to 0.

[0081] Figure 2 A block diagram of a communication statement recognition device according to an embodiment of this application is shown.

[0082] like Figure 2 As shown, a communication statement recognition device 200 according to one embodiment of this application includes:

[0083] The communication statement splitting unit 202 is used to split the communication text obtained by converting the current communication audio into multiple first communication statements;

[0084] The communication statement filtering unit 204 is used to filter out a plurality of second communication statements from the plurality of first communication statements based on the attribute information corresponding to each first communication statement, wherein the attribute information includes at least one of complaint-related keywords, speech rate, maximum volume value within the statement, and average volume value within the statement.

[0085] The communication statement classification unit 206 is used to determine the target complaint tendency of each second communication statement based on each second communication statement and a preset classification model, wherein the preset classification model is trained based on the attribute information of historical communication statements corresponding to historical complaint audio and the target complaint tendency of the historical statements;

[0086] The complaint tendency acquisition unit 208 is used to determine the overall complaint tendency of the current communication audio based on the target complaint tendency of each of the plurality of second communication statements.

[0087] Optionally, in one embodiment of this application, the communication statement filtering unit is used to:

[0088] Based on the attribute information corresponding to each first communication statement and the weight of the attribute information, the initial complaint tendency degree corresponding to each attribute information of the first communication statement is calculated; for any first communication statement, if the initial complaint tendency degree of any attribute information of the first communication statement is greater than or equal to a first threshold, the first communication statement is determined as the second communication statement; or if the initial complaint tendency degree of each attribute information of the first communication statement is greater than or equal to the first threshold, the first communication statement is determined as the second communication statement; or if the average value of the initial complaint tendency degrees of all attribute information of the first communication statement is greater than or equal to a second threshold, the first communication statement is determined as the second communication statement.

[0089] Optionally, in one embodiment of this application, the communication statement filtering unit is used to:

[0090] For any attribute information of any first communication statement, obtain the product of the feature value of the attribute information and the weight of the attribute information as a first parameter; obtain the ratio of the initial complaint tendency of the attribute information in historical communication statements to the actual target complaint tendency of the historical communication statements as a second parameter; use the second parameter as a correction coefficient to correct the first parameter to obtain the initial complaint tendency of the attribute information.

[0091] Optionally, in one embodiment of this application, the communication statement recognition device 200 further includes:

[0092] A preset classification model training unit is used to obtain a first set of historical statements, wherein each historical communication statement in the first set of historical statements has complaint-related keywords and complaint tendency indicators; expand the first set of historical statements into a second set of historical statements, wherein the second set of historical statements includes the first set of historical statements, and the number of statements in the second set of historical statements is greater than the number of statements in the first set of historical statements; filter historical communication statements that meet preset filtering rules from the second set of historical statements to obtain a third set of historical statements; and train the preset classification model based on the attribute information and target complaint tendency of each historical communication statement in the third set of historical statements.

[0093] In one embodiment of this application, optionally, the preset classification model training unit is used for:

[0094] Delete historical communication statements in the second historical statement set whose semantic confidence is lower than a third threshold; and / or delete historical communication statements in the second historical statement set whose similarity to any statement in the first historical statement set is greater than or equal to a fourth threshold.

[0095] In one embodiment of this application, optionally, the complaint tendency acquisition unit 208 is used for:

[0096] For each second communication statement, the following additional information is obtained: the number of complaint-related keywords, the ratio of speech rate to a preset moderate speech rate, the ratio of the highest volume value within the statement to a preset moderate volume, and the ratio of the average volume value within the statement to the preset moderate volume. Based on all attribute information and all additional information of the second communication statement, a complaint tendency matrix is ​​generated. The feature value of the element in the nth row and mth column of the complaint tendency matrix is ​​the square root of the product of the third parameter and the fourth parameter. The third parameter is the ratio of the product of the nth attribute information and the mth additional information of the second communication statement to the target complaint tendency of the second communication statement. The fourth parameter is the ratio of the initial complaint tendency of the attribute information in historical communication statements to the actual target complaint tendency of the historical communication statements. The complaint tendency matrices of all the second communication statements are concatenated into an overall tendency matrix, and the overall complaint tendency of the current communication audio is determined based on the overall tendency matrix.

[0097] The communication statement recognition device 200 uses the solution described in any one of the above embodiments, and therefore has all the above-mentioned technical effects, which will not be repeated here.

[0098] In another embodiment, this application provides a computer device, which may be a server, and its internal structure diagram may be as follows. Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used for communication with external clients via a network connection. When the computer program is executed by the processor, it can implement the communication statement recognition method described in any of the above embodiments.

[0099] In one embodiment, this application also provides a computer device, which can be a client, and its internal structure diagram can be as follows: Figure 4As shown, the computer device includes a processor, memory, network interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it can implement the communication statement recognition method described in any of the above embodiments.

[0100] Any of the computer devices described in the embodiments of this application exist in various forms, including but not limited to:

[0101] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include: smartphones (e.g., iPhones), multimedia phones, feature phones, and low-end phones, etc.

[0102] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include PDAs, MIDs, and UMPCs, such as the iPad.

[0103] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes: audio and video players (such as iPods), handheld game consoles, e-books, as well as smart toys, wearable devices, and portable car navigation devices.

[0104] (4) Server: A device that provides computing services. The components of a server include a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but because they need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0105] (5) Other electronic devices with data interaction functions.

[0106] Additionally, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which are used to perform the following steps:

[0107] The communication text converted from the current communication audio is split into multiple first communication statements;

[0108] Based on the attribute information corresponding to each first communication statement, multiple second communication statements are selected from the multiple first communication statements. The attribute information includes at least one of the following: complaint keywords, speech rate, maximum volume value within the statement, and average volume value within the statement.

[0109] Based on each second communication statement and a preset classification model, the target complaint tendency of each second communication statement is determined. The preset classification model is trained based on the attribute information of historical communication statements corresponding to historical complaint audio and the target complaint tendency of the historical statements.

[0110] Based on the target complaint tendency of each of the multiple second communication statements, the overall complaint tendency of the current communication audio is determined.

[0111] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0112] The technical solution of this application has been described in detail above with reference to the accompanying drawings. This technical solution can automatically determine the level of complaint tendency in communication audio, reducing the manual and time costs associated with identifying complaint tendency. Furthermore, by comprehensively considering multiple dimensions of attribute information, it can more accurately and comprehensively determine the level of complaint tendency in communication audio, improving the accuracy of complaint tendency identification.

[0113] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0114] It should be understood that although the terms "first," "second," etc., may be used to describe communication statements in the embodiments of this application, these communication statements should not be limited to these terms. These terms are only used to distinguish communication statements from each other. For example, without departing from the scope of the embodiments of this application, a first communication statement may also be referred to as a second communication statement, and similarly, a second communication statement may also be referred to as a first communication statement.

[0115] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0116] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0117] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0118] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in a combination of hardware and software functional units.

[0119] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0120] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for recognizing communication statements, characterized in that, include: The communication text converted from the current communication audio is split into multiple first communication statements; Based on the attribute information corresponding to each first communication statement, multiple second communication statements are selected from the multiple first communication statements. The attribute information includes at least one of the following: complaint keywords, speech rate, maximum volume value within the statement, and average volume value within the statement. Based on each second communication statement and a preset classification model, the target complaint tendency of each second communication statement is determined. The preset classification model is trained based on the attribute information of historical communication statements corresponding to historical complaint audio and the target complaint tendency of the historical statements. Based on the target complaint tendency of each of the multiple second communication statements, the overall complaint tendency of the current communication audio is determined; The step of filtering out multiple second communication statements from the plurality of first communication statements based on the attribute information corresponding to each first communication statement includes: Based on the attribute information corresponding to each first communication statement and the weight of the attribute information, the initial complaint tendency degree corresponding to each attribute information of the first communication statement is calculated, wherein... For any attribute information of any of the first communication statements, obtain the product of the feature value of the attribute information and the weight of the attribute information, and use it as the first parameter; The ratio of the initial complaint tendency of the attribute information in historical communication statements to the actual target complaint tendency of the historical communication statements is obtained as a second parameter; Using the second parameter as a correction coefficient, the first parameter is corrected to obtain the initial complaint tendency of the attribute information.

2. The communication statement recognition method according to claim 1, characterized in that, The step of filtering out multiple second communication statements from the plurality of first communication statements based on the attribute information corresponding to each first communication statement includes: For any of the first communication statements If the initial complaint tendency of any attribute information of the first communication statement is greater than or equal to the first threshold, the first communication statement is determined as the second communication statement; or If the initial complaint tendency of each attribute information of the first communication statement is greater than or equal to the first threshold, the first communication statement is determined as the second communication statement; or If the average initial complaint tendency of all attribute information of the first communication statement is greater than or equal to the second threshold, the first communication statement is determined as the second communication statement.

3. The communication statement recognition method according to claim 1 or 2, characterized in that, The methods for training the preset classification model include: Obtain a first set of historical statements, wherein each historical communication statement in the first set of historical statements has complaint-related keywords and complaint tendency identifiers; The first set of historical statements is expanded into a second set of historical statements, wherein the second set of historical statements includes the first set of historical statements, and the number of statements in the second set of historical statements is greater than the number of statements in the first set of historical statements. The third set of historical statements is obtained by filtering out historical communication statements that meet the preset filtering rules from the second set of historical statements. The preset classification model is trained based on the attribute information and target complaint tendency of each historical communication statement in the third historical statement set.

4. The communication statement recognition method according to claim 3, characterized in that, The step of filtering historical communication statements that meet preset filtering rules from the second set of historical statements to obtain a third set of historical statements includes: Delete historical communication statements with a semantic confidence level lower than the third threshold from the second set of historical statements; and / or Delete historical communication statements in the second historical statement set that have a similarity to any statement in the first historical statement set that is greater than or equal to a fourth threshold.

5. The communication statement recognition method according to claim 1, characterized in that, The determination of the overall complaint tendency of the current communication audio based on the target complaint tendency of each of the plurality of second communication statements includes: For each second communication statement, obtain the number of complaint-related keywords, the ratio of speech rate to preset moderate speech rate, the ratio of the highest volume value within the statement to the preset moderate volume, and the ratio of the average volume value within the statement to the preset moderate volume, as additional information. Based on all attribute information and all additional information of the second communication statement, a complaint tendency matrix is ​​generated. The feature value of the element in the nth row and mth column of the complaint tendency matrix is ​​the square root of the product of the third parameter and the fourth parameter. The third parameter is the ratio of the product of the nth attribute information and the mth additional information of the second communication statement to the target complaint tendency of the second communication statement. The fourth parameter is the ratio of the initial complaint tendency of the attribute information in the historical communication statements to the actual target complaint tendency of the historical communication statements. The complaint tendency matrices of all the second communication statements are concatenated into an overall tendency matrix, and the overall complaint tendency of the current communication audio is determined based on the overall tendency matrix.

6. A communication statement recognition device, characterized in that, include: The communication statement splitting unit is used to split the communication text obtained by converting the current communication audio into multiple first communication statements; A communication statement filtering unit is used to filter out multiple second communication statements from the multiple first communication statements based on the attribute information corresponding to each first communication statement, wherein the attribute information includes at least one of complaint-related keywords, speech rate, maximum volume value within the statement, and average volume value within the statement; A communication statement classification unit is used to determine the target complaint tendency of each second communication statement based on each second communication statement and a preset classification model, wherein the preset classification model is trained based on the attribute information of historical communication statements corresponding to historical complaint audio and the target complaint tendency of the historical communication statements; The complaint tendency unit is used to determine the overall complaint tendency of the current communication audio based on the target complaint tendency of each of the plurality of second communication statements; The communication statement filtering unit is used to: calculate the initial complaint tendency degree corresponding to each attribute information of the first communication statement based on the attribute information corresponding to each first communication statement and the weight of the attribute information, wherein, for any attribute information of any first communication statement, the product of the feature value of the attribute information and the weight of the attribute information is obtained as a first parameter; the ratio of the initial complaint tendency degree of the attribute information in historical communication statements to the actual target complaint tendency degree of the historical communication statements is obtained as a second parameter; and the first parameter is corrected using the second parameter as a correction coefficient to obtain the initial complaint tendency degree of the attribute information.

7. The communication statement recognition device according to claim 6, characterized in that, The communication statement filtering unit is used for: For any of the first communication statements, if the initial complaint tendency of any attribute information of the first communication statement is greater than or equal to the first threshold, the first communication statement is determined as the second communication statement; or if the initial complaint tendency of each attribute information of the first communication statement is greater than or equal to the first threshold, the first communication statement is determined as the second communication statement; or if the average of the initial complaint tendency of all attribute information of the first communication statement is greater than or equal to the second threshold, the first communication statement is determined as the second communication statement.

8. A computer device, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, the instructions being configured to perform the method according to any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that, The device stores computer-executable instructions for performing the method as described in any one of claims 1 to 5.

Images