Emotion analysis system and method based on artificial intelligence driving

Abstract

The invention provides an emotion analysis system and method based on artificial intelligence driving, and relates to the technical field of artificial intelligence and natural language processing, and the system comprises a voice acquisition module which is used for receiving voice information and converting the voice information into text language information; the keyword information acquisition module is used for extracting keywords from the text language information; the judgment module is used for judging the language type of the text language information; the text evaluation module is used for selecting a corresponding sentiment analysis processing mode according to the language type so as to generate evaluation information; and the tone evaluation module is used for receiving the evaluation information output by the text evaluation module, obtaining corresponding tone information according to the evaluation information, and determining tone index scores corresponding to the keywords based on the tone recognition result so as to generate voice information evaluation information containing the tone evaluation information. The speech emotion can be accurately recognized, multi-language processing is supported, intelligent evaluation is achieved, and an accurate and reliable emotion analysis result is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and natural language processing, and in particular to an artificial intelligence-driven sentiment analysis system and method. Background Art

[0002] Sentiment analysis is a key research direction in the field of artificial intelligence and natural language processing. Its core lies in automatically identifying and understanding users' emotional tendencies by parsing inputs in various forms such as text, voice, and images. As a branch of sentiment analysis, speech sentiment analysis focuses on extracting emotional information from speech, such as intonation, rhythm, timbre, etc., and combines language content to determine emotional state. At present, sentiment analysis technology has been widely used in many fields such as intelligent customer service, user experience analysis, and sentiment computing. With the rise of emerging applications such as intelligent voice assistants and emotional robots, the demand for sentiment analysis technology continues to rise, especially in a multilingual environment. The need to accurately capture user emotional expressions is becoming more and more urgent.

[0003] Traditional speech emotion analysis technologies mainly rely on rule-based emotion classifiers or machine learning-based classifiers. These technologies usually determine the type and intensity of emotion based on basic acoustic features of speech, such as pitch, volume, and speaking speed. However, traditional methods have the following limitations: Limitations based on acoustic features: Traditional sentiment analysis overly relies on a single dimension of acoustic features, such as evaluating sentiment only through pitch and rhythm changes, while ignoring the important influence of language semantics, context, and emotional vocabulary in the text, resulting in insufficient accuracy in sentiment judgment.

[0004] Separation of speech and text processing: Some early technologies only analyze the acoustic features of speech, and do not fully integrate the semantic information after speech is converted into text. This defect is particularly obvious when dealing with multilingual emotion recognition, especially when dealing with languages ​​with complex tones such as Chinese, the system's performance is often unsatisfactory.

[0005] Limited multi-language support: Traditional methods face challenges in processing multi-language sentiment analysis such as Chinese and English. Due to the significant differences in the speech characteristics of Chinese and English, the applicability of existing systems is limited.

[0006] Although the existing technology has achieved certain results in the field of speech emotion analysis, there are still many deficiencies. Specifically, the defects of the existing technology are mainly reflected in the following aspects: Insufficient accuracy in emotion recognition: Traditional sentiment analysis methods rely solely on acoustic features or text vocabulary for sentiment classification, without fully considering the role of tone in expressing emotions in Chinese speech, resulting in low accuracy in emotion recognition.

[0007] Multilingual sentiment processing is difficult: When processing multilingual sentiment analysis such as Chinese and English, existing technologies usually rely on independent processing modules and lack unified cross-language analysis capabilities, making it difficult to conduct comprehensive analysis of speech features in multiple languages.

[0008] Low level of intelligence: Traditional systems lack the ability to automatically learn and evolve, and are unable to make intelligent adjustments based on application scenarios and user emotional expression habits, making it difficult to adapt to diverse user emotional expressions.

[0009] Therefore, there is an urgent need for an artificial intelligence-driven sentiment analysis system that can overcome the limitations of traditional methods and achieve more accurate, efficient and intelligent sentiment recognition. Summary of the invention

[0010] To this end, the embodiments of the present invention provide a sentiment analysis system and method based on artificial intelligence drive, which are used to solve the problems existing in the prior art in terms of sentiment recognition accuracy, multi-language processing capabilities and intelligence level.

[0011] In order to solve the above problems, an embodiment of the present invention provides a sentiment analysis system driven by artificial intelligence, which includes: A voice acquisition module, used for receiving voice information and converting the voice information into text language information; A keyword information acquisition module, used to extract keywords from the text language information; A determination module, configured to receive the text language information and determine the language type of the text language information; A text evaluation module, used to select a corresponding sentiment analysis processing method according to the language type of the text language information to generate evaluation information of the text language information; The tone evaluation module is used to receive the evaluation information output by the text evaluation module, obtain the corresponding tone information according to the evaluation information, and determine the tone index score corresponding to the keyword based on the tone recognition result, so as to generate voice information evaluation information including the tone evaluation information.

[0012] Preferably, the text evaluation module is used to select a corresponding sentiment analysis processing method according to the language type of the text language information to generate evaluation information of the text language information, including: When the language type is Chinese, the emotion intensity and emotion tendency corresponding to the keywords are calculated, and the overall emotion expression intensity and overall emotion tendency corresponding to the text language information are calculated according to the emotion intensity and emotion tendency corresponding to each keyword, so as to obtain evaluation information corresponding to the text language information; The calculating of the sentiment intensity and sentiment tendency corresponding to the keyword specifically includes: The calculation formula for the sentiment intensity corresponding to the keyword is: ; In the formula, For keywords The emotional intensity, For keywords The word frequency, For keywords The weight in the sentiment dictionary, is the correction coefficient of the context dependency factor; The calculation formula for the sentiment tendency corresponding to the keyword is: ; In the formula, For keywords emotional tendencies, and Keywords corresponding positive and negative weights; The calculating the overall emotion expression intensity and overall emotion tendency corresponding to the text language information according to the emotion intensity and emotion tendency corresponding to each keyword includes: The overall emotional expression intensity corresponding to the text language information : ; In the formula, is the total number of keywords; is a summation symbol, which means that for all Perform summation; The overall emotional tendency corresponding to the text language information : ; When the language type is English, the text language information is identified through the text content identification module, and the emotion intensity and emotion tendency corresponding to each word in the identification result are calculated through the language model learning module, and the overall emotion expression intensity and overall emotion tendency corresponding to the text language information are calculated according to the emotion intensity and emotion tendency corresponding to each word, so as to obtain the evaluation information corresponding to the text language information; The calculation of the emotion intensity and emotion tendency corresponding to each word in the recognition result by the language model learning module specifically includes: First, the recognition results are input into the BERT model to generate the embedding vector for each word. ; Then, the sentiment classification layer is used to embed the vector Classify and get the sentiment distribution of each word : ; In the formula, represents the Softmax function, represents the weight matrix of the classification layer, Represents the bias vector of the classification layer; Finally, according to the sentiment distribution , calculate the sentiment intensity corresponding to each word and emotional tendencies : ; ; In the formula, Representing words The probability of belonging to positive sentiment tendency, Representing words The probability of having a negative emotional tendency, represents the argmax function; The calculating the overall emotional expression intensity and overall emotional tendency corresponding to the text language information according to the emotional intensity and emotional tendency corresponding to each word includes: The overall emotional expression intensity corresponding to the text language information : ; In the formula, is a summation symbol, which means that for all To sum, is the total number of words; The overall emotional tendency corresponding to the text language information : ; In the formula, Represents the mode function, which is used to find the mode; It means finding the mode of the most likely sentiment of all words, that is, finding out which sentiment appears most often in the text as the most likely sentiment.

[0013] Preferably, the text content recognition module is used to recognize text language information based on a preset content recognition algorithm when the language type is English, and send the recognition result as the first content information to the language model learning module.

[0014] Preferably, the language model learning module is used to calculate the emotion intensity and emotion tendency corresponding to each word in the first content information based on the learning result of the preset language model when the language type is English, and send the calculation result to the text evaluation module.

[0015] Preferably, the tone evaluation module is used to receive the evaluation information output by the text evaluation module, obtain the corresponding tone information according to the evaluation information, and determine the tone index score corresponding to the keyword based on the tone recognition result, thereby generating voice information evaluation information containing the tone evaluation information, including: Receiving evaluation information output by the text evaluation module, and acquiring corresponding tone information according to the evaluation information; dividing the data of multiple sub-tones in the tone information into multiple tone data classes, and recognizing the tone information corresponding to the voice information based on the tone data corresponding to the tone data class to obtain a tone recognition result; The tone index score corresponding to the keyword is determined based on the tone recognition result, and the voice information evaluation information including the tone evaluation information is generated according to the tone index score corresponding to the keyword.

[0016] Preferably, the tone evaluation module further includes a tone classification strategy and a tone evaluation strategy, the tone classification strategy includes a preset first tone group, a second tone group and a third tone group, and the tone evaluation strategy includes an evaluation strategy for each tone group for scoring the tone recognition result.

[0017] Preferably, the tone recognition result includes a first recognition result, a second recognition result and a third recognition result, which correspond to the tone contents of Chinese speech, Chinese-English mixed speech and English speech respectively, and different tone evaluation strategies are selected for scoring according to the tone recognition results.

[0018] The embodiment of the present invention further provides a sentiment analysis method based on artificial intelligence driving, the method using the above-mentioned sentiment analysis system based on artificial intelligence driving, specifically comprising: Receiving voice information and converting the voice information into text language information; Extracting keywords from the text language information; receiving the text language information, and determining the language type of the text language information; According to the language type of the text language information, a corresponding sentiment analysis processing method is selected to generate evaluation information of the text language information; The corresponding tone information is acquired according to the evaluation information, and the tone index score corresponding to the keyword is determined based on the tone recognition result, so as to generate voice information evaluation information including the tone evaluation information.

[0019] An embodiment of the present invention also provides an electronic device, which includes a processor, a memory and a bus system, wherein the processor and the memory are connected through the bus system, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the above-mentioned artificial intelligence-driven sentiment analysis method.

[0020] An embodiment of the present invention also provides a computer storage medium, which stores a computer software product. The computer software product includes a number of instructions for enabling a computer device to execute the above-mentioned artificial intelligence-driven sentiment analysis method.

[0021] It can be seen from the above technical solutions that the present invention has the following beneficial effects: (1) High-precision emotion recognition: By combining text evaluation and tone evaluation, the present invention can accurately identify and evaluate speech emotions, thereby improving the accuracy of emotion analysis.

[0022] (2) Multi-language processing capability: The system supports voice input in Chinese and English, and can perform sentiment assessment and tone analysis based on the characteristics of different languages, enhancing its practicality and scope of application.

[0023] (3) Intelligent sentiment evaluation mechanism: Through the learning and application of language models, the system achieves a high degree of automation in the sentiment analysis process and can score the tone of each word according to preset strategies to obtain more comprehensive and accurate sentiment evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them: Figure 1 A block diagram of an artificial intelligence-driven sentiment analysis system provided in an embodiment; Figure 2 It is a flow chart of tone classification and evaluation in the embodiment; Figure 3 The present invention is a flowchart of an artificial intelligence-driven sentiment analysis method provided in an embodiment. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Embodiment 1

[0026] In order to solve the existing problems in emotion recognition accuracy, multi-language processing capabilities and intelligence level. Figure 1 As shown, the embodiment of the present invention proposes an artificial intelligence-driven sentiment analysis system, which includes: A voice acquisition module is used to receive voice information and convert the voice information into text language information; A keyword information acquisition module is used to extract keywords from text language information; A judgment module, used for receiving text language information and judging the language type of the text language information; A text evaluation module is used to select a corresponding sentiment analysis processing method according to the language type of the text language information to generate evaluation information of the text language information; The tone evaluation module is used to receive the evaluation information output by the text evaluation module, obtain the corresponding tone information according to the evaluation information, and determine the tone index score corresponding to the keyword based on the tone recognition result, so as to generate voice information evaluation information containing the tone evaluation information.

[0027] In this embodiment, the voice acquisition module is used to receive voice information and convert the voice information into text language information. The module adopts advanced voice recognition technology and can accurately convert voice information into processable text form.

[0028] In this embodiment, the keyword information acquisition module is used to extract keywords from the text language information. These keywords are crucial for subsequent sentiment analysis because they often carry the main sentiment information in the text.

[0029] In this embodiment, the judgment module is used to receive text language information and judge its language type. This module can recognize multiple languages ​​such as Chinese and English, providing a selection basis for subsequent sentiment analysis processing methods.

[0030] In this embodiment, the text evaluation module is used to select a corresponding sentiment analysis processing method according to the language type of the text language information to generate evaluation information of the text language information.

[0031] Furthermore, when the language type is Chinese, the emotion intensity and emotion tendency corresponding to the keywords are calculated, and the overall emotion expression intensity and overall emotion tendency corresponding to the text language information are calculated based on the emotion intensity and emotion tendency corresponding to each keyword to obtain the evaluation information corresponding to the text language information.

[0032] The above method based on sentiment dictionary is used to calculate the sentiment intensity and sentiment tendency corresponding to the keywords, which specifically includes: The calculation formula for the sentiment intensity corresponding to the keyword is: ; In the formula, For keywords The emotional intensity, For keywords The word frequency, For keywords The weight in the sentiment dictionary, is the correction factor of the context dependency factor.

[0033] The sentiment tendency corresponding to the keyword is calculated by the difference in the positive or negative weight of the keyword in the sentiment dictionary. The calculation formula is: ; In the formula, For keywords emotional tendencies, and Keywords Corresponding positive and negative weights.

[0034] Furthermore, the overall emotional expression intensity and overall emotional tendency corresponding to the text language information are calculated according to the emotional intensity and emotional tendency corresponding to each keyword, including: The overall emotional expression intensity corresponding to the text language information : ; In the formula, is the total number of keywords; is a summation symbol, which means that for all Perform the summation.

[0035] The overall emotional tendency corresponding to the text language information : .

[0036] When the language type is English, the text language information is identified through the text content recognition module, and the emotion intensity and emotion tendency corresponding to each word in the recognition result are calculated through the language model learning module. The overall emotion expression intensity and overall emotion tendency corresponding to the text language information are calculated based on the emotion intensity and emotion tendency corresponding to each word, so as to obtain the evaluation information corresponding to the text language information.

[0037] Specifically, when the language type is English, the text content recognition module is used to pre-process the text language information, including word segmentation and stop word removal. Subsequently, the language model learning module (such as a pre-trained model based on BERT) is used to calculate the sentiment intensity and sentiment tendency corresponding to each word. The specific process is as follows: First, the recognition results are input into the BERT model to generate the embedding vector for each word. .

[0038] Then, the sentiment classification layer (fully connected layer with softmax activation) is used to embed the vector Classify and get the sentiment distribution of each word : ; In the formula, represents the Softmax function, represents the weight matrix of the classification layer, Represents the bias vector of the classification layer.

[0039] Finally, according to the sentiment distribution , calculate the sentiment intensity corresponding to each word and emotional tendencies : ; ; In the formula, Representing words The probability of belonging to positive sentiment tendency, Representing words The probability of having a negative emotional tendency, Represents the argmax function.

[0040] Furthermore, the calculation of the overall emotional expression intensity and overall emotional tendency corresponding to the text language information according to the emotional intensity and emotional tendency corresponding to each word includes: The overall emotional expression intensity corresponding to the text language information : ; In the formula, is a summation symbol, which means that for all To sum, is the total number of words.

[0041] The overall emotional tendency corresponding to the text language information : ; In the formula, Represents the mode function, which is used to find the mode; It means finding the mode of the most likely sentiment of all words, that is, finding out which sentiment appears most often in the text as the most likely sentiment.

[0042] Specifically, the text content recognition module is used to recognize the text language information based on a preset content recognition algorithm when the language type is English, and send the recognition result as the first content information to the language model learning module. The language model learning module is used to calculate the emotion intensity and emotion tendency corresponding to each word in the first content information based on the learning result of the preset language model when the language type is English, and send the calculation result to the text evaluation module.

[0043] In this embodiment, the tone evaluation module is used to receive the evaluation information output by the text evaluation module, obtain the corresponding tone information according to the evaluation information, and determine the tone index score corresponding to the keyword based on the tone recognition result, thereby generating voice information evaluation information containing the tone evaluation information. This step is crucial to the accuracy of sentiment analysis because tone has rich emotional expression functions in language.

[0044] Specifically, the tone evaluation module is further used to divide the data of multiple sub-tones in the tone information into multiple tone data classes, and recognize the tone information corresponding to the voice information based on the tone data corresponding to the tone data class to obtain the tone recognition result.

[0045] In order to more accurately evaluate the emotional expression in speech, the present invention also proposes a tone classification and evaluation strategy, the flow chart of which is as follows: Figure 2 As shown. The strategy includes a preset tone classification strategy and a tone evaluation strategy. The tone classification strategy includes a preset first tone group, a second tone group and a third tone group; the tone evaluation strategy includes an evaluation strategy for each tone group, which is used to score the tone recognition results, wherein the tone recognition results include a first recognition result, a second recognition result and a third recognition result, which correspond to the tone content of Chinese speech, Chinese-English mixed speech and English speech respectively, and different tone evaluation strategies are selected for scoring according to the tone recognition results. During the sentiment analysis process, the system will identify the tone information in the text according to the tone classification strategy, and evaluate each tone according to the tone evaluation strategy, so as to obtain a more comprehensive and accurate sentiment evaluation result.

[0046] Specifically, the preset tone classification strategy includes a first tone group, the first tone group includes light tone, yinping tone, yangping tone, rising tone, falling tone, and falling tone; a second tone group, the second tone group includes light tone, yinping tone, yangping tone, rising tone, falling tone, and falling tone; a third tone group, the third tone group includes light tone, yinping tone, yangping tone, rising tone, falling tone, and falling tone.

[0047] Specifically, the tone evaluation strategies include light tone evaluation strategy, yinping tone evaluation strategy, yangping tone evaluation strategy, shangsheng tone evaluation strategy, xiasheng tone evaluation strategy and qusheng tone evaluation strategy.

[0048] Specifically, the tone recognition result includes a first recognition result, a second recognition result, and a third recognition result. When the tone recognition result is the first recognition result, it is determined that the voice information complies with the recognition strategy corresponding to the first tone group, and the voice information is the tone content of Chinese voice; when the tone recognition result is the second recognition result, it is determined that the voice information complies with the strategy corresponding to the second tone group, and the voice information is the tone content of Chinese voice and English voice; when the tone recognition result is the third recognition result, it is determined that the voice information complies with the tone strategy corresponding to the third tone group, and the voice information is the tone content of English voice.

[0049] Furthermore, the recognition strategy corresponding to the first tone group includes: determining the tone information of each voice information in the text respectively; judging whether the tone information is in a preset tone interval set, if it exists in the preset tone interval, the tone information is represented as the first voice information; if it does not exist in the preset tone interval, the tone information is represented as the second voice information; determining that at least one of the tone information of each of the first voice information and the second voice information is a light tone, and the tone contained in the tone information is any one or more of the yinping, yangping, shangsheng, xiasheng, and qusheng; the tone information of the text content is represented as the first recognition result.

[0050] Furthermore, the recognition strategy corresponding to the second tone group includes: determining the tone information of each voice information in the text respectively; judging whether the tone information is in a preset tone interval set, if it exists in the preset tone interval, the tone information is represented as the first voice information; if it does not exist in the preset tone interval, the tone information is represented as the second voice information; determining that at least one of the tone information corresponding to each first voice information is any one or more of light tone, yinping tone, yangping tone, rising tone, falling tone, and falling tone; further determining that at least one of the tone information is not a light tone, and at least one of the tone information that is not a light tone is an English tone; the tone information of the text content is represented as the second recognition result.

[0051] Further, the recognition strategy corresponding to the third tone group includes: determining the tone information of each voice information in the text respectively; judging whether the tone information is in a preset tone interval set, if it exists in the preset tone interval, the tone information is represented as the first voice information; if it does not exist in the preset tone interval, the tone information is represented as the second voice information; determining that at least one tone information in the tone information corresponding to each first voice information is any one or more of light tone, yinping tone, yangping tone, rising tone, lower tone, and falling tone; further determining that at least one tone information in the tone information corresponding to a first voice information is not a light tone, and at least one tone in the tone information that is not a light tone is an English tone; determining that at least one tone information in the tone information corresponding to a first voice information is a light tone, and the tone contained in the tone information is any one or more of yinping tone, yangping tone, rising tone, lower tone, and falling tone; the tone information of the text content is represented as the third recognition result.

[0052] Furthermore, the system also includes: determining whether the tone information of the text content is a first recognition result; if the tone information of the text content is represented as a first recognition result, calculating the tone score corresponding to the text according to the first recognition result; determining whether the tone of the text content is a second recognition result; if the tone information of the text content is represented as a second recognition result, calculating the tone score corresponding to the text according to the second recognition result; determining whether the tone of the text content is a third recognition result; if the tone information of the text content is represented as a third recognition result, calculating the tone score corresponding to the text according to the third recognition result; according to the tone score of the text content, the scores corresponding to the first recognition result, the second recognition result and the third recognition result are merged to obtain the score corresponding to the voice information.

[0053] The system of the present invention can process Chinese and English voice inputs and has good multi-language processing capabilities. According to the characteristics of different languages, the system can perform tone analysis and sentiment evaluation respectively, which improves the practicality and scope of application of the system. For the processing of English text, the system adopts a learning algorithm based on a language model, which can accurately identify the sentiment intensity and sentiment tendency of each word.

[0054] Through the learning and application of language models, this system can automatically generate evaluation information based on the emotional intensity and tendency of the text, achieving a high degree of automation in the emotional analysis process. In addition, the system also has an intelligent emotional evaluation mechanism that can score the tone of each word according to the preset tone classification strategy and evaluation strategy. This scoring mechanism not only takes into account the characteristics of the tone itself, but also combines the emotional expression of the entire text, thereby obtaining a more comprehensive and accurate emotional evaluation result.

[0055] In summary, the present invention provides an artificial intelligence-driven sentiment analysis system that can accurately identify and evaluate voice sentiment, supports multi-language processing, and has an intelligent sentiment evaluation mechanism. Through the collaborative work of various modules, the system can provide users with accurate and reliable sentiment analysis results. Embodiment 2

[0056] like Figure 3 As shown, the present invention provides a sentiment analysis method based on artificial intelligence driving, which uses the sentiment analysis system based on artificial intelligence driving of the above embodiment 1, and specifically includes: Receive voice information and convert the voice information into text language information; Extract keywords from text language information; Receiving text language information, and determining the language type of the text language information; According to the language type of the text language information, a corresponding sentiment analysis processing method is selected to generate evaluation information of the text language information; Corresponding tone information is acquired according to the evaluation information, and the tone index score corresponding to the keyword is determined based on the tone recognition result, thereby generating voice information evaluation information including the tone evaluation information.

[0057] An artificial intelligence-driven sentiment analysis method of this embodiment is used to use the aforementioned artificial intelligence-driven sentiment analysis system. Therefore, the specific implementation method of the artificial intelligence-driven sentiment analysis method can be found in the embodiment section of the artificial intelligence-driven sentiment analysis system above. In order to avoid redundancy, it will not be repeated here. Embodiment 3

[0058] An embodiment of the present invention provides an electronic device, which includes a processor, a memory and a bus system. The processor and the memory are connected through the bus system. The memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the above-mentioned artificial intelligence-driven sentiment analysis method. Embodiment 4

[0059] An embodiment of the present invention provides a computer storage medium, which stores a computer software product. The computer software product includes a number of instructions for enabling a computer device to execute the above-mentioned artificial intelligence-driven sentiment analysis method.

[0060] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0061] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0062] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0063] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from them are still within the protection scope of the invention.

Claims

1. An artificial intelligence-driven sentiment analysis system, characterized in that: include: A voice acquisition module, used for receiving voice information and converting the voice information into text language information; A keyword information acquisition module, used to extract keywords from the text language information; A determination module, configured to receive the text language information and determine the language type of the text language information; A text evaluation module, used to select a corresponding sentiment analysis processing method according to the language type of the text language information to generate evaluation information of the text language information; The tone evaluation module is used to receive the evaluation information output by the text evaluation module, obtain the corresponding tone information according to the evaluation information, and determine the tone index score corresponding to the keyword based on the tone recognition result, so as to generate voice information evaluation information including the tone evaluation information.

2. The artificial intelligence-driven sentiment analysis system according to claim 1, characterized in that: The text evaluation module is used to select a corresponding sentiment analysis processing method according to the language type of the text language information to generate evaluation information of the text language information, including: When the language type is Chinese, the emotion intensity and emotion tendency corresponding to the keywords are calculated, and the overall emotion expression intensity and overall emotion tendency corresponding to the text language information are calculated according to the emotion intensity and emotion tendency corresponding to each keyword, so as to obtain evaluation information corresponding to the text language information; The calculating of the sentiment intensity and sentiment tendency corresponding to the keyword specifically includes: The calculation formula for the sentiment intensity corresponding to the keyword is: ； In the formula, For keywords The emotional intensity, For keywords The word frequency, For keywords The weight in the sentiment dictionary, is the correction coefficient of the context dependency factor; The calculation formula for the sentiment tendency corresponding to the keyword is: ； In the formula, For keywords emotional tendencies, and Keywords corresponding positive and negative weights; The calculating the overall emotion expression intensity and overall emotion tendency corresponding to the text language information according to the emotion intensity and emotion tendency corresponding to each keyword includes: The overall emotional expression intensity corresponding to the text language information : ； In the formula, is the total number of keywords; is a summation symbol, which means that for all Perform summation; The overall emotional tendency corresponding to the text language information : ； When the language type is English, the text language information is identified through the text content identification module, and the emotion intensity and emotion tendency corresponding to each word in the identification result are calculated through the language model learning module, and the overall emotion expression intensity and overall emotion tendency corresponding to the text language information are calculated according to the emotion intensity and emotion tendency corresponding to each word, so as to obtain the evaluation information corresponding to the text language information; The calculation of the emotion intensity and emotion tendency corresponding to each word in the recognition result by the language model learning module specifically includes: First, the recognition results are input into the BERT model to generate the embedding vector for each word. ; Then, the sentiment classification layer is used to embed the vector Classify and get the sentiment distribution of each word : ； In the formula, represents the Softmax function, represents the weight matrix of the classification layer, Represents the bias vector of the classification layer; Finally, according to the sentiment distribution , calculate the sentiment intensity corresponding to each word and emotional tendencies : ； ； In the formula, Representing words The probability of belonging to positive sentiment tendency, Representing words The probability of having a negative emotional tendency, represents the argmax function; The calculating the overall emotional expression intensity and overall emotional tendency corresponding to the text language information according to the emotional intensity and emotional tendency corresponding to each word includes: The overall emotional expression intensity corresponding to the text language information : ； In the formula, is a summation symbol, which means that for all To sum, is the total number of words; The overall emotional tendency corresponding to the text language information : ； In the formula, Represents the mode function, which is used to find the mode; It means finding the mode of the most likely sentiment of all words, that is, finding out which sentiment appears most often in the text as the most likely sentiment.

3. The artificial intelligence-driven sentiment analysis system according to claim 2, characterized in that: The text content recognition module is used to recognize text language information based on a preset content recognition algorithm when the language type is English, and send the recognition result as the first content information to the language model learning module.

4. The artificial intelligence-driven sentiment analysis system according to claim 3, characterized in that: The language model learning module is used to calculate the emotion intensity and emotion tendency corresponding to each word in the first content information based on the learning result of the preset language model when the language type is English, and send the calculation result to the text evaluation module.

5. The artificial intelligence-driven sentiment analysis system according to claim 1, characterized in that: The tone evaluation module is used to receive the evaluation information output by the text evaluation module, obtain the corresponding tone information according to the evaluation information, and determine the tone index score corresponding to the keyword based on the tone recognition result, so as to generate the voice information evaluation information including the tone evaluation information, including: Receiving evaluation information output by the text evaluation module, and acquiring corresponding tone information according to the evaluation information; dividing the data of multiple sub-tones in the tone information into multiple tone data classes, and recognizing the tone information corresponding to the voice information based on the tone data corresponding to the tone data class to obtain a tone recognition result; The tone index score corresponding to the keyword is determined based on the tone recognition result, and the voice information evaluation information including the tone evaluation information is generated according to the tone index score corresponding to the keyword.

6. The artificial intelligence-driven sentiment analysis system according to claim 1, characterized in that: The tone evaluation module further includes a tone classification strategy and a tone evaluation strategy. The tone classification strategy includes a preset first tone group, a second tone group and a third tone group. The tone evaluation strategy includes an evaluation strategy for each tone group, which is used to score the tone recognition result.

7. The artificial intelligence-driven sentiment analysis system according to claim 6, characterized in that: The tone recognition results include a first recognition result, a second recognition result and a third recognition result, which correspond to the tone contents of Chinese speech, Chinese-English mixed speech and English speech respectively. Different tone evaluation strategies are selected for scoring according to the tone recognition results.

8. An artificial intelligence-driven sentiment analysis method, characterized in that: The method uses the artificial intelligence-driven sentiment analysis system according to any one of claims 1 to 7, and specifically includes: Receiving voice information and converting the voice information into text language information; Extracting keywords from the text language information; receiving the text language information, and determining the language type of the text language information; According to the language type of the text language information, a corresponding sentiment analysis processing method is selected to generate evaluation information of the text language information; The corresponding tone information is acquired according to the evaluation information, and the tone index score corresponding to the keyword is determined based on the tone recognition result, so as to generate voice information evaluation information including the tone evaluation information.

9. An electronic device, characterized in that: The electronic device includes a processor, a memory and a bus system, the processor and the memory are connected through the bus system, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory to implement the artificial intelligence-driven sentiment analysis method described in claim 8.

10. A computer storage medium, characterized in that: The computer storage medium stores a computer software product, which includes a number of instructions for enabling a computer device to execute the artificial intelligence-driven sentiment analysis method described in claim 8.

Images