Emotion analysis method and device, electronic equipment and storage medium

Abstract

The embodiment of the application provides a kind of mood analysis method, device, electronic equipment and storage medium, belong to data processing technical field.The method comprises: the historical conversation information of target user obtained is preprocessed, and the sample conversation information of target user is obtained;Sample conversation information is input into preset mood model to carry out mood classification, and the mood frequency set is obtained;According to mood frequency set, the mood shift matrix of target user is established, and mood model is trained based on mood shift matrix, and the mood shift model is obtained;The target conversation information of target user obtained is input into mood shift model to carry out probability prediction, and the mood fluctuation probability value is obtained.The embodiment of the application can accurately analyze the change of user's mood under the influence of other people's mood.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to an emotion analysis method and device, electronic equipment and storage medium. BACKGROUND

[0002] Characterizing a person's personality has great application prospects and can be applied to scenarios such as social friend pushing, social matching, and psychological counseling. Currently, means such as emotion recognition in conversation (ERC) or questionnaire survey are usually used to predict the emotion label task of each sentence in a conversation. The ERC task is to analyze the emotion label of a single sentence in a conversation, and the emotion expressed by a person in a sentence is understood through the existing ERC task. However, expressing the emotion of a user through only the ERC task is one-sided and cannot accurately calculate the emotion probability of the user affected by others. In addition, the questionnaire survey is easily affected by the subjective emotion of the person filling in the questionnaire and has certain requirements for the number of persons filling in the questionnaire, thereby increasing the labor cost. SUMMARY

[0003] The main purpose of the embodiments of the present application is to provide an emotion analysis method, device, electronic equipment and storage medium, which can accurately analyze the change of the emotion of a user under the influence of the emotion of others.

[0004] To achieve the above-mentioned purpose, a first aspect of the embodiments of the present application provides an emotion analysis method, which comprises:

[0005] The historical conversation information of a target user obtained is preprocessed to obtain sample conversation information of the target user, wherein the sample conversation information carries a plurality of emotion labels;

[0006] The sample conversation information is input into a preset emotion model for emotion classification to obtain an emotion frequency set, wherein the emotion frequency set includes the number of times each emotion label appears in the sample conversation information;

[0007] An emotion transfer matrix of the target user is established according to the emotion frequency set, wherein the emotion transfer matrix is used to represent the probability value of the emotion of the target user being affected by the emotion of others;

[0008] The preset emotion model is trained based on the emotion transfer matrix to obtain an emotion transfer model;

[0009] inputting the obtained target conversation information of the target user into the emotion transfer model to perform probability prediction, to obtain an emotion fluctuation probability value, where the emotion fluctuation probability value is a probability value representing that the target user's emotion changes under the influence of other people's emotion.

[0010] In some embodiments, the historical conversation information includes conversation information of a plurality of users in conversation with the target user; and the preprocessing of the obtained historical conversation information of the target user to obtain sample conversation information of the target user includes:

[0011] determining, in the historical conversation information, the sample conversation information corresponding to the target user;

[0012] annotating the sample conversation information based on a preset annotation model to obtain emotion labels of the sample conversation information.

[0013] In some embodiments, the preset emotion model includes a sentence vector encoder and an emotion classifier; and the inputting of the sample conversation information into the preset emotion model to perform emotion classification to obtain an emotion frequency set includes:

[0014] inputting a sentence in the sample conversation information into the sentence vector encoder to perform sentence segmentation, to obtain a sentence vector of the sample conversation information;

[0015] inputting the sentence vector into the emotion classifier to enable the emotion classifier to perform emotion recognition on the emotion labels in the sentence vector, to obtain a recognition result;

[0016] performing classification statistics on the recognition result to obtain the emotion frequency set.

[0017] In some embodiments, the method further includes:

[0018] establishing a self-emotion matrix according to the emotion frequency set, where the self-emotion matrix is used to represent a probability value of an emotion of a next sentence in conversation under the influence of an emotion of a current conversation of the target user;

[0019] establishing an emotion influence matrix according to the emotion frequency set, where the emotion influence matrix is used to represent a probability value of the target user being influenced by the emotion of the target user in the process of conversation with the target user.

[0020] In some embodiments, the training of the preset emotion model based on the emotion transfer matrix to obtain an emotion transfer model includes:

[0021] constraining the emotion labels in the sample conversation information to obtain a real emotion label sequence;

[0022] input the real emotion label sequence into the emotion classifier for probability calculation to obtain a label probability distribution;

[0023] optimize a preset score value function according to the emotion transition matrix, the self-emotion matrix, the emotion influence matrix, the real emotion label sequence, and the label probability distribution to obtain a predicted sequence probability function;

[0024] train the preset emotion model according to the predicted sequence probability function to obtain the emotion transition model.

[0025] In some embodiments, the preset emotion model includes a conditional random field layer; and the constraining the emotion labels in the sample dialogue information to obtain a real emotion label sequence includes:

[0026] inputting the emotion labels in the sample dialogue information into the conditional random field layer for screening to obtain a real emotion label probability value;

[0027] performing category statistics on the real emotion label probability value to obtain the real emotion label sequence.

[0028] In some embodiments, the predicted sequence probability function includes a likelihood probability function and a loss function; and the training the preset emotion model according to the predicted sequence probability function to obtain the emotion transition model includes:

[0029] obtaining a likelihood probability value of the preset emotion model according to the likelihood probability function;

[0030] inputting the likelihood probability value into the loss function for calculation to obtain a probability loss value;

[0031] obtaining an emotion transition probability value of the preset emotion model according to the loss function;

[0032] when the probability loss value is less than the emotion transition probability value, updating the preset emotion model according to the probability loss value to obtain the emotion transition model.

[0033] To achieve the above object, a second aspect of the embodiment of the present application provides an emotion analysis device, which comprises:

[0034] a dialogue processing module configured to pre-process the obtained historical dialogue information of a target user to obtain sample dialogue information of the target user, wherein the sample dialogue information carries a plurality of emotion labels;

[0035] The emotion classification module is configured to input the sample conversation information into a preset emotion model to perform emotion classification, and obtain an emotion frequency set, wherein the emotion frequency set includes a number of times that each emotion label appears in the sample conversation information.

[0036] The matrix establishing module is configured to establish an emotion transfer matrix of the target user according to the emotion frequency set, wherein the emotion transfer matrix is used to represent a probability value of the emotion of the target user being affected by emotions of other people.

[0037] The model training module is configured to train the preset emotion model based on the emotion transfer matrix, and obtain an emotion transfer model.

[0038] The probability prediction module is configured to input the obtained target conversation information of the target user into the emotion transfer model to perform probability prediction, and obtain an emotion fluctuation probability value, wherein the emotion fluctuation probability value is used to represent a probability value of the emotion of the target user being changed under the influence of emotions of other people.

[0039] To achieve the above object, a third aspect of the embodiments of the present application provides an electronic device, which comprises a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection communication between the processor and the memory, and the program is executed by the processor to realize the emotion analysis method of the first aspect.

[0040] To achieve the above object, a fourth aspect of the embodiments of the present application provides a storage medium, which is a computer readable storage medium, used for computer readable storage, and the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to realize the emotion analysis method of the first aspect.

[0041] The emotion analysis method, apparatus, electronic device, and storage medium proposed in this application first preprocess the historical dialogue information of the target user to obtain sample dialogue information of the target user with emotion tags. Then, the sample dialogue information is input into a preset emotion model for emotion classification, thereby obtaining the frequency of different emotion tags in the sample dialogue information. An emotion frequency set is generated based on the frequency of different emotion tags. Then, an emotion transfer matrix of the target user is established based on the emotion frequency set to facilitate subsequent analysis of the target user's emotion changes. The preset emotion model is then trained based on the emotion transfer matrix to improve the robustness of the preset emotion model, resulting in an emotion transfer model. Finally, the target user's target dialogue information is input into the emotion transfer model for probability prediction. Thus, the emotion transfer model analyzes the probability value of the target user's emotion change under the influence of others' emotions, improving the accuracy of the target user's emotion analysis and enabling accurate analysis of the target user's emotion changes under the influence of others' emotions. Attached Figure Description

[0042] Figure 1 This is a flowchart of the sentiment analysis method provided in the embodiments of this application;

[0043] Figure 2 yes Figure 1 The flowchart of step S101 in the text;

[0044] Figure 3 yes Figure 1 The flowchart of step S102 in the document;

[0045] Figure 4 This is a flowchart of a sentiment analysis method provided in another embodiment of this application;

[0046] Figure 5 yes Figure 1 The flowchart of step S104 in the process;

[0047] Figure 6 yes Figure 5 The flowchart of step S501 in the process;

[0048] Figure 7 yes Figure 5 The flowchart of step S504 in the process;

[0049] Figure 8 This is a schematic diagram of the structure of the emotion analysis device provided in the embodiments of this application;

[0050] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0051] In order to make the purposes, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0052] It should be noted that although the functional modules are divided in the device schematic diagram, and the logical sequence is shown in the flowchart, in some cases, the steps shown or described can be performed in a manner different from the module division in the device or the sequence in the flowchart. The terms "first", "second", and the like in the specification and claims and the above-described drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0053] 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. The terms used herein are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.

[0054] First, the meanings of several terms involved in the present application are analyzed:

[0055] Emotion Recognition in Conversations (ERC): ERC is a major task in the study of conversation emotion, which is used to realize a conversation system with emotional understanding. The task is a classification task, which aims to classify the emotions of all utterances in a conversation. The input of the task is a continuous conversation, and the output is the emotion of all utterances in the conversation.

[0056] Natural Language Processing (NLP): NLP uses computers to process, understand and use human languages (such as Chinese, English, etc.), and NLP is a branch of artificial intelligence, which is an interdisciplinary subject of computer science and linguistics, and is also commonly known as computational linguistics. Natural language processing includes syntax analysis, semantic analysis, discourse understanding, etc. Natural language processing is commonly used in machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining, etc. It involves data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research related to language processing, and language computing related linguistic research.

[0057] Conditional Random Field (CRF): Conditional Random Field is a kind of discriminant model most suitable for prediction tasks, in which adjacent context information or state will affect the current prediction. CRF has applications in named entity recognition, part-of-speech tagging, gene prediction, noise reduction and object detection problems, etc. CRF is an undirected graph model, which is often used for labeling or analyzing sequence data such as natural language text or biological sequences. In recent years, it has achieved good results in sequence labeling tasks such as word segmentation, part-of-speech tagging and named entity recognition.

[0058] Information Extraction: Information Extraction is a text processing technology that extracts specified types of entities, relationships, events and other factual information from natural language text and forms structured data output. Information Extraction is a technology that extracts specific information from text data. Text data is composed of specific units such as sentences, paragraphs, and chapters, and text information is composed of specific units such as words, phrases, sentences, paragraphs, or combinations of these specific units. Extracting noun phrases, names, and places from text data is text information extraction, of course, the information extracted by the text information extraction technology can be various types of information.

[0059] Bidirectional Encoder Representations from Transformer (BERT): BERT model is a pre-trained language representation model, which emphasizes that it is no longer necessary to use traditional unidirectional language models or shallowly concatenate two unidirectional language models for pre-training, but to use a new Masked Language Model (MLM) to generate deep bidirectional language representations. The goal of the BERT model is to train a large amount of unlabeled corpus to obtain a representation of the text containing rich semantic information, i.e., semantic representation of the text, and then fine-tune the semantic representation of the text in a specific NLP task, and finally apply it to the NLP task.

[0060] Based on this, the embodiments of the present application provide a mood analysis method and device, electronic equipment and storage medium, aiming to improve the accuracy of mood analysis of the target user, and can accurately analyze the mood change condition of the target user affected by the mood of others.

[0061] The mood analysis method and device, electronic equipment and storage medium provided by the embodiments of the present application are specifically explained by the following embodiments, first, the mood analysis method in the embodiments of the present application is described.

[0062] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. The artificial intelligence (AI) is a theory, method, technology and application system for simulating, extending and expanding human intelligence by using a digital computer or a machine controlled by a digital computer, perceiving an environment, acquiring knowledge and using the knowledge to obtain optimal results.

[0063] The artificial intelligence basic technology generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. The artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0064] The emotion analysis method provided by the embodiments of the present application relates to the field of artificial intelligence. The emotion analysis method provided by the embodiments of the present application can be applied to a terminal, can be applied to a server end, and can also be software running in a terminal or a server end. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc.; the server end can be configured as an independent physical server, can be configured as a server cluster or a distributed system composed of multiple physical servers, can also be configured as a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and basic cloud computing services such as big data and artificial intelligence platforms; and the software can be an application implementing the emotion analysis method, etc., but is not limited to the above forms.

[0065] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0066] It should be noted that in various specific embodiments of the present application, when relevant processing needs to be performed on data related to the identity or characteristics of the user, such as user information, user behavior data, user history data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of such data will comply with relevant laws, regulations, and standards of the relevant countries and regions. In addition, when the embodiments of the present application need to obtain sensitive personal information of the user, the separate permission or separate consent of the user will be obtained through a pop-up window or a jump to a confirmation page, and after obtaining the separate permission or separate consent of the user, the necessary user-related data for enabling the embodiments of the present application to function normally will be obtained.

[0067] Figure 1 is an optional flowchart of the emotion analysis method provided by the embodiments of the present application, Figure 1 The method in can include but is not limited to steps S101-S105.

[0068] Step S101, pre-processing the obtained historical conversation information of the target user to obtain sample conversation information of the target user;

[0069] It should be noted that the sample conversation information carries multiple emotion labels.

[0070] In step S101 of some embodiments, the obtained historical conversation information of the target user is pre-processed to obtain sample conversation information of the target user, which facilitates subsequent analysis of the emotion probability value of the target user based on the sample conversation information of the target user.

[0071] It can be understood that the historical conversation information can be the chat record of the target user with friends or relatives, wherein the chat record can include chat voice, chat message, or chat emoticon record, and the number of historical conversation information is as large as possible, and as much as possible to include conversation information of various time periods and various chat friends. The historical conversation information can be obtained three times a month or once a day, etc., and the present embodiment does not make specific limitations.

[0072] It should be noted that the historical conversation information, chat record, chat voice, and chat emoticon information described above are obtained with the permission of the target user.

[0073] Step S102, inputting the sample conversation information into a preset emotion model for emotion classification to obtain an emotion frequency set;

[0074] It should be noted that the emotion frequency set includes the number of times each emotion label appears in the sample conversation information.

[0075] It can be understood that the emotional tags include but are not limited to happy, angry, sad, excited, sad, excited, etc.

[0076] In step S102 of some embodiments, the sample conversation information of the target user is input into the preset emotion model for emotion classification, to obtain the frequency values of different emotional tags appearing in the sample conversation information, and finally obtain an emotion frequency set, so as to obtain the emotional state proportion of the target user in daily life, facilitating subsequent accurate analysis of the target user's emotions.

[0077] It should be noted that the emotion frequency set also includes the frequency of the related emotional tags appearing in different sentences in the sample conversation information, wherein the different sentences in the sample conversation information can be the conversation content of the target user himself or the conversation content of the target user and others, for example, the conversation content of the target user himself is "I bought a bouquet of flowers today, I am very happy", "Unfortunately, when I get home, the flowers are a little wilted, I am a little sad", the preset emotion model will identify the emotional tags such as "happy" and "sad" in the sample conversation information, combine the context of the conversation sentences, and classify the emotional tags in the specific situation to obtain the emotion frequency set; or the conversation content of the target user and others is that the target user "the weather is good today, let's go for a picnic tomorrow", user A "I have a cold, I feel very uncomfortable now, I won't go tomorrow", and the target user "that's okay, it's a little disappointing, take care of your rest", the preset emotion model identifies the emotional tags in the sample conversation information, combines the conversation content of the target user and user A, and classifies the emotional tags in the conversation situation to obtain the emotion frequency set.

[0078] It can be understood that the proportion of each emotion of the target user is the emotional distribution of the target user, and the distribution value of each emotion is 0%-100%, and the total proportion of all emotions is 100%, for example, the happy emotion of the target user is 30%, the sad emotion is 20%, the angry emotion is 40%, and the excited emotion is 10%, etc., which is not limited in the embodiment.

[0079] In step S103, an emotion transfer matrix of the target user is established according to the emotion frequency set.

[0080] It should be noted that the emotion transfer matrix is used to represent the probability value of the target user's emotion being affected by the emotion of others.

[0081] In step S103 of some embodiments, the emotion transfer matrix of the target user is established according to the emotion frequency set, so as to facilitate subsequent training of the preset emotion model according to the emotion transfer matrix, thereby improving the prediction accuracy of the change of the target user's emotion in the preset emotion model when the target user's emotion is affected.

[0082] In some embodiments, the emotion transfer matrix is established according to the number of emotion labels corresponding to the emotion frequency set, wherein the emotion transfer matrix is a two-dimensional matrix. For example, in the context of the target user conversing with others, the number of emotion labels in the emotion frequency set in which the target user is happy is c1, and the number of emotion labels in which the other person is sad is c2, and the emotion transfer matrix is (c1, c2), indicating whether the target user's emotion will be affected, from happy to sad or remain happy, when the other person is sad.

[0083] In step S104, the preset emotion model is trained based on the emotion transfer matrix to obtain an emotion transfer model.

[0084] In step S104 of some embodiments, the preset emotion model is trained based on the emotion transfer matrix, so that the emotion transfer model can accurately calculate the probability value of the target user's emotion being affected, thereby enhancing the robustness of the emotion transfer model.

[0085] In step S105, the target conversation information of the target user obtained is input into the emotion transfer model for probability prediction to obtain an emotion fluctuation probability value.

[0086] It should be noted that the emotion fluctuation probability value is a probability value representing the change in the target user's emotion under the influence of the other person's emotion.

[0087] In step S105 of some embodiments, the target conversation information of the target user is first obtained, and then the target conversation information is input into the trained emotion transfer model for probability prediction to obtain the emotion fluctuation probability value of the target user under the influence of the other person, thereby accurately obtaining the probability of the target user's emotion changing.

[0088] It can be understood that the target conversation information is also the chat record of the target user with friends or relatives, which is not described here.

[0089] The steps S101 to S105 shown in the embodiments of the present application first pre-process the obtained historical conversation information of the target user to obtain sample conversation information of the target user carrying emotion labels, then input the sample conversation information into a preset emotion model for emotion classification, so as to obtain the number of times of occurrence of different emotion labels in the sample conversation information, generate an emotion frequency set according to the number of times of occurrence of different emotion labels, and then establish an emotion transfer matrix of the target user according to the emotion frequency set, so as to facilitate subsequent analysis of the emotion change of the target user, train the preset emotion model according to the emotion transfer matrix, improve the robustness of the preset emotion model, obtain an emotion transfer model, and finally input the obtained target conversation information of the target user into the emotion transfer model for probability prediction, so as to analyze the probability value of the emotion change of the target user under the influence of the emotion of others through the emotion transfer model, improve the accuracy of the emotion analysis of the target user, and accurately analyze the emotion change condition of the target user influenced by the emotion of others.

[0090] Please refer to Figure 2 In some embodiments, step S101 can include but is not limited to steps S201 to S202:

[0091] It should be noted that the historical conversation information includes the conversation information of a plurality of users in conversation with the target user.

[0092] Step S201, determining sample conversation information corresponding to the target user in the historical conversation information;

[0093] Step S202, labeling the sample conversation information based on a preset labeling model to obtain emotion labels of the sample conversation information.

[0094] In step S201 of some embodiments, the sample conversation information corresponding to the target user is determined in the historical conversation information, so as to facilitate subsequent emotion analysis.

[0095] It should be noted that the sample conversation information corresponding to the target user can be determined by the identity of the target user, or manually selected sample conversation information of the target user, etc., and the present embodiment does not make specific limitation.

[0096] In step S202 of some embodiments, the sample conversation information is labeled based on a preset labeling model to obtain a plurality of emotion labels of the sample conversation information, so as to be able to analyze different emotion labels.

[0097] It should be noted that the preset emotion model is a trained conversation emotion recognition model, which automatically predicts and labels the sample conversation information through the trained conversation emotion recognition model, improves the accuracy of emotion label labeling, and reduces the labeling cost.

[0098] It can be understood that the sample dialogue information can also be labeled by manual labeling, and the embodiment is not limited specifically.

[0099] Please refer to Figure 3 In some embodiments, step S102 can include but is not limited to steps S301-S303.

[0100] It should be noted that the preset emotion model includes a sentence vector encoder and an emotion classifier.

[0101] Step S301, inputting a sentence in the sample dialogue information into the sentence vector encoder for sentence segmentation to obtain a sentence vector of the sample dialogue information;

[0102] Step S302, inputting the sentence vector into the emotion classifier to enable the emotion classifier to perform emotion recognition on the emotion label in the sentence vector to obtain a recognition result;

[0103] Step S303, classifying and counting the recognition result to obtain an emotion frequency set.

[0104] In step S301 of some embodiments, each text sentence in the sample dialogue information is input into the sentence vector encoder, so that the sentence vector encoder performs segmentation on the sentence according to the field and the preset delimiter, and finally takes the vector at the 0th position as the final output sentence vector, thereby obtaining the complete sentence vector of the sample dialogue information, and improving the accuracy of emotion label recognition in the sample dialogue information.

[0105] It should be noted that the sentence vector encoder is a bert model.

[0106] In step S302 of some embodiments, various sentence vectors obtained in step S301 are input into the emotion classifier to enable the emotion classifier to perform emotion recognition on the emotion label appearing in the sentence vector to obtain a recognition result, thereby determining the probability distribution of different emotion labels in the sentence vector.

[0107] In step S303 of some embodiments, the recognition result obtained in step S302 is classified and counted to determine the probability distribution of different emotion labels in the sentence vector, and the frequency of occurrence of each emotion label is obtained, and finally the emotion frequency set is obtained.

[0108] Please refer to Figure 4 , Figure 4 is an optional flowchart of an emotion analysis method provided by another embodiment of the present application, Figure 4 The method in can include but is not limited to steps S401-S402.

[0109] Step S401, establishing a self-emotion matrix according to the emotion frequency set;

[0110] It should be noted that the self-emotion matrix is used to represent the probability value of the next sentence emotion under the influence of the current conversation emotion of the target user.

[0111] In step S402, an emotion influence matrix is established according to the emotion frequency set.

[0112] It should be noted that the emotion influence matrix is used to represent the probability value of the target user's emotion influence on others in the conversation process.

[0113] In step S401 of some embodiments, an emotion frequency set is established according to the frequency of the emotion label in the conversation under different conversation situations, and an emotion matrix is established to represent the probability of the target user's emotion from the current self-emotion to the next sentence emotion, wherein the self-emotion matrix is a two-dimensional matrix.

[0114] It can be understood that since there are frequency values of various emotion labels under various conversation situations in the emotion frequency set, the self-emotion matrix can be established according to the frequency of the emotion label in the conversation under different conversation situations, for example, the selected conversation situation is the target user's own conversation, and the frequency of the emotion label appearing in the target user's own conversation in the emotion frequency set is assumed to be the frequency a1 of the current emotion of the target user, and the frequency c1 of the emotion of the next sentence, and the established self-emotion matrix is (a1, c1).

[0115] In step S402 of some embodiments, an emotion influence matrix is established according to the frequency of the emotion label in the conversation under different conversation situations, and an emotion influence matrix is established to represent the probability value of the target user's emotion influence on others in the conversation process, wherein the emotion influence matrix is a two-dimensional matrix.

[0116] It can be understood that since there are frequency values of various emotion labels under various conversation situations in the emotion frequency set, the conversation situation can be first selected as the conversation between the target user and others, and then the current emotion of others is obtained as the frequency b1, and the current emotion of the target user is obtained as the frequency a2, and the established emotion influence matrix is (b1, a2).

[0117] Please refer to Figure 5 In some embodiments, step S104 includes but is not limited to steps S501 to S504:

[0118] Step S501, constraint the emotion label in the sample conversation information to obtain the real emotion label sequence;

[0119] Step S502, inputting the real emotion label sequence into the emotion classifier to perform probability calculation to obtain a label probability distribution;

[0120] Step S503, optimizing a preset score value function according to the emotion transition matrix, the self-emotion matrix, the emotion influence matrix, the real emotion label sequence, and the label probability distribution to obtain a predicted sequence probability function;

[0121] Step S504, training a preset emotion model according to the predicted sequence probability function to obtain an emotion transition model.

[0122] In step S501 of some embodiments, the emotion labels in the sample dialogue information are constrained to obtain the real emotion label sequence, which facilitates subsequent analysis of the emotion of the target user according to the real emotion label sequence.

[0123] In step S502 of some embodiments, the real emotion label sequence obtained in step S501 is input into the emotion classifier to perform probability calculation, so that the emotion classifier predicts the emotion labels in the real label sequence to obtain a label probability distribution.

[0124] In step S503 of some embodiments, the preset score value function is optimized according to the various matrices, the real emotion label sequence, and the label probability distribution to obtain a predicted sequence probability function for predicting the emotion change of the target user.

[0125] It should be noted that the maximum likelihood optimization method is used to optimize the preset score value function in the embodiments of the present application.

[0126] In step S504 of some embodiments, the preset emotion model is finally trained according to the predicted sequence probability function to enhance the robustness of the predicted sequence probability function, and an emotion transition model is obtained, which facilitates subsequent prediction of the emotion change of the target user.

[0127] Please refer to Figure 6 In some embodiments, step S501 includes but is not limited to steps S601 to S602:

[0128] It should be noted that the preset emotion model includes a conditional random field layer.

[0129] Step S601, inputting the emotion labels in the sample dialogue information into the conditional random field layer to perform screening to obtain real emotion label probability values;

[0130] Step S602, performing category statistics on the real emotion label probability values to obtain a real emotion label sequence.

[0131] In some embodiments, the emotion label in the sample dialogue information is input into a conditional random field layer in the preset emotion model for screening to obtain a real emotion label probability value after maximum elimination, and then the real emotion label probability value is classified to obtain a real emotion label sequence, thereby improving the prediction accuracy.

[0132] Referring to Figure 7 In some embodiments, the step S504 can include, but is not limited to, steps S701-S704.

[0133] It should be noted that the prediction sequence probability function includes a likelihood probability function and a loss function.

[0134] In step S701, a likelihood probability value of the preset emotion model is obtained according to the likelihood probability function.

[0135] In step S702, the likelihood probability value is input into the loss function for calculation to obtain a probability loss value.

[0136] In step S703, an emotion transition probability value of the preset emotion model is obtained according to the loss function.

[0137] In step S704, when the probability loss value is less than the emotion transition probability value, the preset emotion model is updated according to the probability loss value to obtain an emotion transition model.

[0138] In some embodiments, the prediction sequence probability function includes the likelihood probability function and the loss function. First, the preset emotion model is maximum likelihood optimized according to the likelihood probability function to obtain a likelihood probability value of the preset emotion model. Then, the likelihood probability value is input into the loss function for calculation to obtain a probability loss value. The emotion transition probability value of the current preset emotion model is obtained according to the loss function. Finally, the probability loss value and the transition probability value are compared. When the probability loss value is less than the emotion transition probability value, the preset emotion model is updated according to the probability loss value to obtain an emotion transition model.

[0139] It should be noted that when the probability loss value is less than or equal to the transition probability value, the training of the preset emotion model can be stopped.

[0140] Referring to Figure 8 The embodiments of the present application also provide an emotion analysis device, which can implement the above-mentioned emotion analysis method. The device includes:

[0141] The dialogue processing module 801 is configured to pre-process the obtained historical dialogue information of the target user to obtain sample dialogue information of the target user, wherein the sample dialogue information carries a plurality of emotion labels.

[0142] The emotion classification module 802 is used to input sample dialogue information into a preset emotion model for emotion classification and obtain an emotion frequency set, wherein the emotion frequency set includes the number of times each emotion label appears in the sample dialogue information.

[0143] The matrix building module 803 is used to build an emotion transfer matrix for the target user based on the set of emotion frequencies. The emotion transfer matrix is ​​used to represent the probability value of the target user's emotions being affected by the emotions of others.

[0144] The model training module 804 is used to train a preset emotion model based on the emotion transfer matrix to obtain an emotion transfer model.

[0145] The probability prediction module 805 is used to input the target dialogue information of the target user into the emotion transfer model for probability prediction and obtain the emotion fluctuation probability value. The emotion fluctuation probability value is used as the probability value of the target user's emotion changing under the influence of the emotions of others.

[0146] The specific implementation of this emotion analysis device is basically the same as the specific implementation of the emotion analysis method described above, and will not be repeated here.

[0147] To illustrate the process of sentiment analysis more clearly, a specific example will be provided below.

[0148] Example 1:

[0149] Suppose there are M speakers in a dialogue, p0, p1, ..., p2. M-1 p0 is the target user to be analyzed, and p is the speaker. i A total of len were said i Sentences. Suppose a dialogue X contains N sentences, and these N sentences are the sample dialogue information, i.e., x0, x1, ..., xn. N-1 Among them, x in the sample dialogue information i It is the speaker p speaker(i) Speaking of which, speaker(i) is x i The corresponding speaker's subscript. And, other speakers p i The words spoken were utterance(p i [j] is the speaker p i The index of the j-th sentence spoken. The ERC task is to predict the emotion label for each sentence, including happy, angry, neutral, sad, excited, and furious.

[0150] First, extract a single text sentence x from the sample dialogue information. t The input is a sentence vector encoder, which segments the sample dialogue information into sentences, dividing the text into sentences x. tSplit into token sequence [[cls], x t,0 ,x t,1 …… x t,L-1 ,[sep]] according to words, so that there are L words in x t , x t,i represents the i-th word in x t ,[cls] and [sep] represent the beginning and end of the sentence separator respectively, and then take the representation of the [cls] position in [[cls], x t,0 ,x t,1 …… x t,L-1 ,[sep]] as the sentence vector identifier, and obtain u t , u t is a vector of size d, and the formula (1) of the sentence vector u t is as follows:

[0151] u t = BERT([[cls], x t,0 ,x t,1 …… x t,L-1 ,[sep]]) [0] (1)

[0152] It should be noted that [0] in formula (1) is the vector at the 0th position, and the vector at the 0th position is taken as the final output sentence vector.

[0153] Then input the sentence vector u t into the emotion classifier for emotion recognition, and obtain the frequency distribution of each emotion, so as to obtain the emotion frequency set, and the formula (2) of the emotion frequency set is as follows:

[0154] p t =W·u t +b (2)

[0155] Then, according to the emotion frequency set, establish the target self-emotion matrix M1, the emotion influence matrix M2 and the emotion transfer matrix M3, wherein M1, M2 and M3 are all two-dimensional matrices.

[0156] Input the sample dialogue information into the conditional random field layer for screening, and obtain the true emotion label probability value, and the formula (3) of the true emotion label probability value is as follows:

[0157] max(P(y|X)) (3)

[0158] Classify the true emotion label probability value to obtain the true emotion label sequence y = [y0, y1…… y N-1 ], wherein y i is the true emotion label of the i-th sentence. X = [p0, p1…… p N-1is the label probability distribution calculated by the emotion classifier, the true label of the ith sentence is y i , and the label probability distribution predicted by the emotion classifier for the ith sentence is p i .

[0159] Finally, the preset score value function in the preset emotion model is optimized according to the emotion transition matrix, the self-emotion matrix, and the emotion influence matrix, and the preset score value function formula (4) is as follows:

[0160]

[0161] In the model training process, the maximum likelihood optimization method is used for optimization, and the likelihood probability value of the predicted sequence is as shown in formula (5):

[0162]

[0163] The likelihood probability value is input into the loss function for calculation to obtain the probability loss value, and the probability loss value is as shown in formula (6):

[0164] loss = -log(P(y|X)) (6)

[0165] It should be noted that in the training process, the preset emotion model is processed in batches, the probability loss value of the preset emotion model is obtained first, then the gradient is calculated in reverse, the preset emotion model is updated according to the probability loss value, and the final emotion transition model is obtained.

[0166] In some embodiments, by analyzing the emotions of the target user, the daily emotion distribution of the target user can be obtained, for example, the emotion with a higher proportion in the distribution represents the emotional state of the target user. For example, the higher the proportion of happy emotion distribution of a person, the happier the person is in daily chat; the emotional change of the target user, for example, the self-emotion matrix M1 of the target user represents the emotional change characteristics of the target user, which is obtained through model training and learning. In the matrix, M1 i,j represents that the user changes from the ith emotion to the jth emotion, and the larger the value represents that the user changes from the ith emotion to the jth emotion more frequently in the chat; the influence of the target user on the emotions of others, for example, the emotion influence matrix M2 of the target user represents the characteristics of the influence of the target user on the emotions of others, which is obtained through model training and learning. In the matrix, M2 i,j represents the possibility that the user is the ith emotion and the other party is the jth emotion, and M2 i,j is larger, the more this kind of emotion transition occurs; the influence of the target user on the emotions of others, for example, the emotion transition matrix M3 of the target user represents the characteristics of the influence of the user on the emotions of others, which is obtained through model training and learning. In the matrix, M3 i,jThe greater the possibility that the user is in the jth emotion when the opposite party is in the ith emotion, the more the transfer of this kind of emotion occurs, thereby enabling multi-dimensional and comprehensive analysis of the personality characteristics of the target user.

[0167] The embodiment of the present application further provides an electronic device, which comprises a memory, a processor, a program stored in the memory and capable of running on the processor, and a data bus for realizing connection communication between the processor and the memory, and the program is executed by the processor to realize the emotion analysis method. The electronic device can be any intelligent terminal such as a tablet computer or a vehicle-mounted computer.

[0168] Please refer to Figure 9 , Figure 9 The hardware structure of the electronic device of another embodiment is illustrated, and the electronic device comprises:

[0169] The processor 901 can be implemented in the form of a general CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, and is used to execute related programs to realize the technical solutions provided by the embodiments of the present application;

[0170] The memory 902 can be implemented in the form of a ROM (Read Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory). The memory 902 can store an operating system and other application programs, and when the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 902 and are called and executed by the processor 901 to realize the emotion analysis method of the embodiments of the present application;

[0171] The input / output interface 903 is used to realize information input and output;

[0172] The communication interface 904 is used to realize the communication interaction between the device and other devices, and can realize communication through a wired manner (such as a USB, a network cable, etc.) or a wireless manner (such as a mobile network, WIFI, Bluetooth, etc.);

[0173] The bus 905 transmits information between various components (such as the processor 901, the memory 902, the input / output interface 903, and the communication interface 904) of the device;

[0174] The processor 901, the memory 902, the input / output interface 903, and the communication interface 904 are communicatively connected to each other within the device through the bus 905.

[0175] The embodiment of the present application further provides a storage medium, which is a computer readable storage medium, used for computer readable storage, and stores one or more programs, which can be executed by one or more processors to implement the above-mentioned emotion analysis method.

[0176] The memory, as a non-transitory computer readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory can include a high-speed random access memory and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and these remote memories can be connected to the processor through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0177] The emotion analysis method, the emotion analysis device, the electronic device, and the storage medium provided by the embodiment of the present application first pre-process the historical conversation information of the target user obtained to obtain sample conversation information of the target user carrying an emotion label, then input the sample conversation information into a preset emotion model for emotion classification, so as to obtain the number of times of occurrence of different emotion labels in the sample conversation information, generate an emotion frequency set according to the number of times of occurrence of different emotion labels, and then establish an emotion transfer matrix of the target user according to the emotion frequency set, so as to facilitate subsequent analysis of the emotion change of the target user, train the preset emotion model according to the emotion transfer matrix, improve the robustness of the preset emotion model, obtain an emotion transfer model, finally input the target conversation information of the target user obtained into the emotion transfer model for probability prediction, so as to analyze the probability value of the emotion change of the target user under the influence of the emotion of others through the emotion transfer model, improve the accuracy of the emotion analysis of the target user, and accurately analyze the emotion change condition of the target user influenced by the emotion of others.

[0178] The embodiments described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0179] Those skilled in the art can understand that, Figures 1-7The technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and can include more or fewer steps than shown, or combine certain steps, or different steps.

[0180] The apparatus embodiments described above are merely illustrative, and units described as separate components can or can not be physically separate, i.e., can be located in one place, or can be distributed over multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purposes of the embodiments.

[0181] Those skilled in the art can understand that all or some of the steps in the above disclosed method, the functional modules / units in the system and the device can be implemented as software, firmware, hardware and their appropriate combinations.

[0182] The terms "first", "second", "third", "fourth" and the like (if any) in the description and the above drawings of the present application are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in other than the order illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily limit to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0183] It should be understood that in the present application, "at least one" means one or more, and "multiple" means two or more. "And / or" is used to describe the association between the associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases: only A, only B, and A and B at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0184] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented by other manners. For example, the apparatus embodiments described above are merely illustrative, for example, the division of the above units is merely a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, apparatuses or units, and can be electrical, mechanical or other forms.

[0185] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they can be located in one place or distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0186] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0187] When the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program storage media.

[0188] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, but this does not limit the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. A sentiment analysis method, characterized in that, The method includes: The historical dialogue information of the target user is preprocessed to obtain sample dialogue information of the target user, wherein the sample dialogue information carries multiple emotion tags. The sample dialogue information is input into a preset emotion model for emotion classification to obtain an emotion frequency set, wherein the emotion frequency set includes the number of times each emotion tag appears in the sample dialogue information; An emotion transfer matrix for the target user is established based on the set of emotion frequencies, wherein the emotion transfer matrix is ​​used to characterize the probability value of the target user's emotions being influenced by the emotions of others; The preset emotion model is trained based on the emotion transfer matrix to obtain the emotion transfer model; The obtained target dialogue information of the target user is input into the emotion transfer model for probability prediction to obtain an emotion fluctuation probability value, wherein the emotion fluctuation probability value is a probability value that represents the target user's emotion changes under the influence of others' emotions. The step of training the preset emotion model based on the emotion transfer matrix to obtain the emotion transfer model includes: Constraining the emotion labels in the sample dialogue information yields a sequence of real emotion labels. The real emotion label sequence is input into the emotion classifier of the preset emotion model for probability calculation to obtain the label probability distribution; The preset score function is optimized based on the emotion transfer matrix, the self-emotion matrix, the emotion influence matrix, the real emotion label sequence, and the label probability distribution to obtain the prediction sequence probability function; the self-emotion matrix is ​​used to characterize the probability value of the current dialogue emotion of the target user influencing the emotion of the next dialogue, and the emotion influence matrix is ​​used to characterize the probability value of others being influenced by the target user's emotion during the dialogue with the target user; The preset emotion model is trained based on the predicted sequence probability function to obtain the emotion transfer model.

2. The sentiment analysis method according to claim 1, characterized in that, The historical dialogue information includes dialogue information of multiple users who have had conversations with the target user; The step of preprocessing the acquired historical dialogue information of the target user to obtain sample dialogue information of the target user includes: The sample dialogue information corresponding to the target user is determined from the historical dialogue information; The sample dialogue information is labeled based on a preset labeling model to obtain the emotion label of the sample dialogue information.

3. The sentiment analysis method according to claim 1, characterized in that, The preset emotion model includes a sentence vector encoder and an emotion classifier; the step of inputting the sample dialogue information into the preset emotion model for emotion classification to obtain an emotion frequency set includes: The sentences in the sample dialogue information are input into the sentence vector encoder for sentence segmentation to obtain the sentence vectors of the sample dialogue information. The sentence vector is input into the emotion classifier, so that the emotion classifier can perform emotion recognition on the emotion label in the sentence vector and obtain the recognition result. The identification results are classified and statistically analyzed to obtain the set of emotion frequencies.

4. The sentiment analysis method according to claim 3, characterized in that, Also includes: Establish your own emotion matrix based on the set of emotion frequencies; An emotion influence matrix is ​​established based on the set of emotion frequencies.

5. The emotion analysis method according to claim 1, characterized in that, The preset emotion model includes a conditional random field layer; constraining the emotion labels in the sample dialogue information to obtain a sequence of real emotion labels includes: The emotion labels in the sample dialogue information are input into the conditional random field layer for filtering to obtain the probability values ​​of the true emotion labels. The probability values ​​of the true emotion labels are statistically analyzed to obtain the sequence of true emotion labels.

6. The sentiment analysis method according to claim 1, characterized in that, The predicted sequence probability function includes a likelihood probability function and a loss function; the step of training the preset emotion model based on the predicted sequence probability function to obtain the emotion transfer model includes: The likelihood probability value of the preset emotion model is obtained based on the likelihood probability function. The likelihood value is input into the loss function to calculate the probability loss value; The probability value of emotion transfer of the preset emotion model is obtained according to the loss function; When the probability loss value is less than the emotion transfer probability value, the preset emotion model is updated according to the probability loss value to obtain the emotion transfer model.

7. An emotion analysis device, characterized in that, The apparatus for applying the sentiment analysis method according to any one of claims 1 to 6 comprises: The dialogue processing module is used to preprocess the historical dialogue information of the target user to obtain sample dialogue information of the target user, wherein the sample dialogue information carries multiple emotion tags. The emotion classification module is used to input the sample dialogue information into a preset emotion model for emotion classification to obtain an emotion frequency set, wherein the emotion frequency set includes the number of times each emotion tag appears in the sample dialogue information; A matrix building module is used to build an emotion transfer matrix for the target user based on the set of emotion frequencies, wherein the emotion transfer matrix is ​​used to characterize the probability value of the target user's emotions being influenced by the emotions of others; The model training module is used to train the preset emotion model based on the emotion transfer matrix to obtain the emotion transfer model; The probability prediction module is used to input the target dialogue information of the target user into the emotion transfer model for probability prediction to obtain an emotion fluctuation probability value, wherein the emotion fluctuation probability value is used as the probability value of the target user's emotion changing under the influence of others' emotions.

8. An electronic device, characterized in that, The electronic device includes a memory, a processor, a program stored in the memory and capable of running on the processor, and a data bus for enabling communication between the processor and the memory. When the program is executed by the processor, it implements the steps of the sentiment analysis method as described in any one of claims 1 to 6.

9. A storage medium, said storage medium being a computer-readable storage medium for computer-readable storage, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of the sentiment analysis method according to any one of claims 1 to 6.

Images