Artificial intelligence-based intention recognition method and device, computer equipment and medium

Abstract

The invention belongs to the technical field of artificial intelligence, relates to an intention recognition method and device based on artificial intelligence, computer equipment and a storage medium, is applied to the field of financial science and technology, and comprises the following steps: acquiring voice call data and call text data corresponding to a target user; performing feature extraction on the voice call data to obtain emotion dimension features; performing feature extraction on the voice call data to obtain attitude dimension features; performing feature extraction on the call text data based on a large language model to obtain cognitive dimension features; generating corresponding basic information features based on the basic information of the target user; performing integration processing on the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a target feature vector; and performing prediction processing on the target feature vector based on an intention recognition model to obtain a product purchase intention result. In addition, the product purchase intent result may be stored in the blockchain. The identification efficiency and the identification precision of the product purchase intention are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and the field of financial technology, and in particular to an intention recognition method, device, computer equipment and storage medium based on artificial intelligence. Background Art

[0002] In the insurance industry, with the increasingly fierce market competition and the diversification of consumer demand, how to efficiently and accurately recommend insurance products to potential customers with high purchase intentions has become a key challenge for insurance companies to improve sales performance, optimize resource allocation, and enhance customer satisfaction. Traditionally, the task of evaluating customers' intention to purchase products mainly relies on the personal experience and subjective judgment of business personnel. However, this method has significant limitations and shortcomings.

[0003] First, the personal judgment of sales personnel is often affected by many factors, including but not limited to personal experience, emotional state, familiarity with the product, and interference from the external environment. These factors may lead to large differences in the assessment of the purchase intention of the same customer by different sales personnel, which lacks uniformity and objectivity.

[0004] Secondly, sales personnel have limited energy and it is difficult to accurately identify potential customers with high purchase intentions among a large customer base. In the absence of an effective screening mechanism, sales personnel may have to invest a lot of time and energy in promoting low-intention customers, which not only wastes valuable sales resources, but also may damage customer experience and reduce brand favorability due to frequent interruptions.

[0005] In summary, the existing insurance industry approach of manually assessing customers' intention to purchase products is inefficient and has low accuracy. Summary of the invention

[0006] The purpose of the embodiments of the present application is to propose an intention recognition method, apparatus, computer equipment and storage medium based on artificial intelligence to solve the technical problem that the existing insurance industry method of relying on manual evaluation of customer intention to purchase products is inefficient and has low accuracy.

[0007] In order to solve the above technical problems, the embodiment of the present application provides an intention recognition method based on artificial intelligence, which adopts the following technical solutions:

[0008] Acquire voice call data corresponding to the target user, and call text data corresponding to the voice call data;

[0009] Extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features;

[0010] Extracting features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features;

[0011] Extracting features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features;

[0012] Acquire basic information of the target user, and generate corresponding basic information features based on the basic information;

[0013] Integrate the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector;

[0014] The target feature vector is predicted based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

[0015] Furthermore, the step of extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features specifically includes:

[0016] Extracting a first audio file containing the user voice of the target user from the voice call data;

[0017] dividing the first audio file into corresponding first audio segments;

[0018] Calculating an average pitch of the first audio segment, and calculating an average sound intensity of the first audio segment;

[0019] generating a corresponding emotional valence level feature based on the average pitch;

[0020] generating a corresponding emotional arousal level feature based on the average sound intensity;

[0021] An emotion dimension feature corresponding to the voice call data is generated based on the emotion valence level feature and the emotion arousal level feature.

[0022] Furthermore, the step of extracting a first audio file containing the user voice of the target user from the voice call data specifically includes:

[0023] Get the preset noise reduction strategy;

[0024] Performing voice noise reduction processing on the voice call data based on the noise reduction strategy to obtain corresponding initial audio;

[0025] Performing voice separation processing on the initial audio based on a preset separation algorithm to obtain corresponding user audio and seat audio;

[0026] The user audio is used as the first audio file.

[0027] Furthermore, the step of extracting features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features specifically includes:

[0028] extracting a second audio file containing the user voice of the target user from the voice call data;

[0029] dividing the second audio file into corresponding second audio segments;

[0030] Calculating an average speaking speed of the second audio segment;

[0031] Attitude dimension features corresponding to the voice call data are generated based on the average speaking speed.

[0032] Furthermore, the step of extracting features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features specifically includes:

[0033] Calling the large language model;

[0034] Encoding the call text data to obtain a corresponding embedding vector;

[0035] Perform cognitive dimension feature extraction processing on the embedding vector based on the large language model to obtain a corresponding first specified feature;

[0036] The cognitive dimension feature is generated based on the first specified feature.

[0037] Furthermore, the step of integrating the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector specifically includes:

[0038] Get the preset feature fusion strategy;

[0039] Based on the feature fusion strategy, the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature are fused to obtain a fused second designated feature;

[0040] The second designated feature is used as the target feature vector.

[0041] Furthermore, before the step of predicting the target feature vector based on the preset intention recognition model to obtain the product purchase intention result corresponding to the target user, the method further includes:

[0042] Acquire pre-collected initial sample data; wherein the initial sample data includes basic information of the user, voice call data of the user, and corresponding product purchase results;

[0043] Preprocessing the initial sample data to obtain corresponding first sample data;

[0044] Performing feature engineering processing on the first sample data based on a preset feature scoring model and the large language model to obtain corresponding second sample data;

[0045] Call the preset initial model;

[0046] Training and evaluating the initial model based on the second sample data to obtain a specified model that meets the preset construction conditions;

[0047] The designated model is used as the intent recognition model.

[0048] In order to solve the above technical problems, the embodiment of the present application also provides an intention recognition device based on artificial intelligence, which adopts the following technical solution:

[0049] A first acquisition module, used to acquire voice call data corresponding to a target user, and call text data corresponding to the voice call data;

[0050] A first extraction module, configured to extract features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features;

[0051] A second extraction module, configured to extract features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features;

[0052] A third extraction module is used to extract features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features;

[0053] A generating module, used to obtain the basic information of the target user and generate corresponding basic information features based on the basic information;

[0054] An integration module, used for integrating the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector;

[0055] The prediction module is used to perform prediction processing on the target feature vector based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

[0056] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0057] Acquire voice call data corresponding to the target user, and call text data corresponding to the voice call data;

[0058] Extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features;

[0059] Extracting features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features;

[0060] Extracting features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features;

[0061] Acquire basic information of the target user, and generate corresponding basic information features based on the basic information;

[0062] Integrate the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector;

[0063] The target feature vector is predicted based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

[0064] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0065] Acquire voice call data corresponding to the target user, and call text data corresponding to the voice call data;

[0066] Extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features;

[0067] Extracting features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features;

[0068] Extracting features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features;

[0069] Acquire basic information of the target user, and generate corresponding basic information features based on the basic information;

[0070] Integrate the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector;

[0071] The target feature vector is predicted based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

[0072] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0073] The present application first obtains voice call data corresponding to a target user, and call text data corresponding to the voice call data; then, based on a preset emotion extraction strategy, feature extraction is performed on the voice call data to obtain corresponding emotion dimension features; and based on a preset attitude extraction strategy, feature extraction is performed on the voice call data to obtain corresponding attitude dimension features; and feature extraction is performed on the call text data based on a preset large language model to obtain corresponding cognitive dimension features; then, basic information of the target user is obtained, and corresponding basic information features are generated based on the basic information; subsequently, the emotion dimension features, the attitude dimension features, the cognitive dimension features and the basic information features are integrated and processed to obtain a corresponding target feature vector; finally, the target feature vector is predicted and processed based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user. The present application obtains voice call data, call text data and basic information corresponding to the target user, and then extracts the emotion dimension features, attitude dimension features, cognitive dimension features and basic information features corresponding to the target user from four dimensions, namely emotion, attitude, cognition and basic information, based on the emotion extraction strategy, attitude extraction strategy and the use of a large language model. Subsequently, the target feature vector obtained by integrating the four dimensional features is predicted and processed based on the use of an intention recognition model, so that the product purchase intention result corresponding to the target user can be generated quickly and accurately, thereby effectively improving the recognition efficiency of product purchase intention and improving the recognition accuracy of product purchase intention. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0075] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0076] Figure 2 A flowchart of an embodiment of an artificial intelligence-based intention recognition method according to the present application;

[0077] Figure 3is a schematic structural diagram of an embodiment of an intention recognition device based on artificial intelligence according to the present application;

[0078] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0079] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0080] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0081] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0082] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0083] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0084] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.

[0085] The server 103 may be a server that provides various services, such as a background server that provides support for a web page displayed on the terminal device 101 .

[0086] It should be noted that the artificial intelligence-based intention recognition method provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based intention recognition device is generally set in the server / terminal device.

[0087] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0088] Continue to refer Figure 2 , showing a flow chart of an embodiment of an intention recognition method based on artificial intelligence according to the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted. The intention recognition method based on artificial intelligence provided in the embodiment of the present application can be applied to any scenario where product purchase intention recognition is required, and the intention recognition method based on artificial intelligence can be applied to products in these scenarios, for example, the purchase intention recognition of financial products in the field of finance and insurance. The intention recognition method based on artificial intelligence includes the following steps:

[0089] Step S201, obtaining voice call data corresponding to a target user and call text data corresponding to the voice call data.

[0090] In this embodiment, the electronic device (e.g. Figure 1The server / terminal device shown in the figure) can obtain voice call data through a wired connection or a wireless connection. It should be pointed out that the above-mentioned wireless connection methods may include but are not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future. The executor of the present application is specifically an intent recognition system, or simply referred to as a system. The present application can be applied to call scenarios under the product marketing business of financial insurance, such as customer service systems, sales platforms, etc. The above-mentioned voice call data may be call data in the form of voice collected by customer service during the business processing of recommending insurance products to target users. Among them, the corresponding call text data can be obtained by performing speech-to-text processing on the above-mentioned voice call data.

[0091] Step S202: extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features.

[0092] In this embodiment, the above-mentioned specific implementation process of extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0093] Step S203: extract features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features.

[0094] In this embodiment, the above-mentioned specific implementation process of extracting features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0095] Step S204: extract features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features.

[0096] In this embodiment, the specific implementation process of extracting features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0097] Step S205: acquiring basic information of the target user, and generating corresponding basic information features based on the basic information.

[0098] In this embodiment, the user database can be queried using the user ID (e.g., user name) of the target user to obtain basic information of the target user. The basic information may include age, gender, occupation, etc., and the basic information is further feature constructed to generate corresponding basic information features.

[0099] Step S206, integrating the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector.

[0100] In this embodiment, the above-mentioned specific implementation process of integrating the emotion dimension features, the attitude dimension features, the cognitive dimension features and the basic information features to obtain the corresponding target feature vector will be further described in detail in subsequent specific embodiments of this application and will not be elaborated here.

[0101] Step S207: predicting the target feature vector based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

[0102] In this embodiment, the target feature vector is input into the above-mentioned intention recognition model, and the target feature vector is predicted and processed with respect to the purchase intention by the intention recognition model, and the corresponding prediction result, i.e., the above-mentioned product purchase intention result, is returned. The product purchase intention result can be a probability distribution or a direct category label (with purchase intention or without purchase intention), which is used to indicate the probability of whether the target user intends to purchase the product. Among them, the specific construction process of the above-mentioned intention recognition model will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.

[0103] The present application obtains voice call data, call text data and basic information corresponding to the target user, and then extracts the emotion dimension features, attitude dimension features, cognitive dimension features and basic information features corresponding to the target user from four dimensions, namely emotion, attitude, cognition and basic information, based on the emotion extraction strategy, attitude extraction strategy and the use of a large language model. Subsequently, the target feature vector obtained by integrating the features of the four dimensions is predicted and processed based on the use of an intention recognition model, so as to quickly and accurately generate the product purchase intention result corresponding to the target user, effectively improving the recognition efficiency of product purchase intention and improving the recognition accuracy of product purchase intention.

[0104] In some optional implementations, step S202 includes the following steps:

[0105] A first audio file including the user voice of the target user is extracted from the voice call data.

[0106] In this embodiment, the specific implementation process of extracting the first audio file containing the user voice of the target user from the voice call data will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0107] The first audio file is divided into corresponding first audio segments.

[0108] In this embodiment, the first audio file may be divided into a plurality of corresponding first audio segments according to a preset division time length, for example, one audio segment every 10 seconds.

[0109] An average pitch of the first audio segment is calculated, and an average intensity of the first audio segment is calculated.

[0110] In this embodiment, the process of calculating the average pitch of the first audio segment includes: for each first audio segment, converting the audio signal from the time domain to the frequency domain by performing a fast Fourier transform on the first audio segment. Then, the audio signal is cut into time frames, and the pitch at the time point is obtained for each time frame, and the average is calculated to obtain the average pitch of the first audio segment. Subsequently, the median of the average pitch of all audio segments containing the user's voice in the voice call data is calculated to obtain a quantitative index of the emotional tendency of the target user in the voice call data.

[0111] The process of calculating the average sound intensity of the first audio segment includes: for each first audio segment, by performing fast Fourier transform on the first audio segment, dividing each first audio segment into time frames, and obtaining the sound intensity of the time frame. The sound intensity is calculated by calculating the square of the amplitude, summing the squares and then taking the square root. Then, the sound intensity corresponding to all frames of a first audio segment is calculated to obtain the average sound intensity of the first audio segment.

[0112] A corresponding emotional valence level feature is generated based on the average pitch.

[0113] In this embodiment, by calculating the median of the average pitch of all audio clips containing the user's voice in the voice call data, a quantitative index of the target user's emotional tendency in the voice call data is obtained and used as the corresponding emotional valence level feature.

[0114] A corresponding emotional arousal level feature is generated based on the average sound intensity.

[0115] In this embodiment, the median of the average sound intensity of all the first audio clips is selected as a quantitative indicator of the overall emotion intensity of the user, and as a corresponding emotion arousal level feature.

[0116] An emotion dimension feature corresponding to the voice call data is generated based on the emotion valence level feature and the emotion arousal level feature.

[0117] In this embodiment, emotions can be considered from two dimensions: valence and arousal. The higher the valence, the more positive and pleasant the individual's emotions are. The higher the arousal, the stronger and more positive the individual's emotions are. Positive emotions tend to indicate that individuals have a higher tendency to buy. Higher pitches usually mean that people's emotions are positive. It was found that high-energy sounds express stronger levels of emotional arousal. In summary, high pitch and high volume usually express positive and strong emotions, while low pitch and low volume often imply relatively negative and bland emotions.

[0118] Specifically, by using a two-dimensional coordinate system, in the two-dimensional coordinate system, with the emotional valence level as the vertical axis and the emotional arousal level as the horizontal axis, the emotional point of each voice call data is plotted. According to the distribution of the voice call data in the emotional space (the emotional valence level characteristics and the emotional arousal level characteristics), the corresponding emotional dimension characteristics are analyzed and obtained. Among them, positive emotions are usually located in the positive direction of the valence axis and the arousal is moderate or high, while negative emotions may be located in the negative direction of the valence axis and the arousal is high or low.

[0119] The present application extracts a first audio file containing the user voice of the target user from the voice call data; then divides the first audio file into corresponding first audio segments; then calculates the average pitch of the first audio segment, and calculates the average sound intensity of the first audio segment; subsequently generates a corresponding emotional valence level feature based on the average pitch; and generates a corresponding emotional arousal level feature based on the average sound intensity; finally, generates an emotional dimension feature corresponding to the voice call data based on the emotional valence level feature and the emotional arousal level feature, thereby achieving accurate generation of emotional dimension features according to the emotional valence level and the emotional arousal level, and ensuring the feature accuracy of the obtained emotional dimension features.

[0120] In some optional implementations of this embodiment, the feature extraction of the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features includes the following steps:

[0121] Get the preset noise reduction strategy.

[0122] In this embodiment, the above-mentioned noise reduction strategy may specifically use methods such as wavelet transform.

[0123] The voice call data is subjected to voice noise reduction processing based on the noise reduction strategy to obtain corresponding initial audio.

[0124] In this embodiment, the voice call data is subjected to voice noise reduction processing by using a noise reduction strategy to improve the accuracy of subsequent processing and obtain corresponding initial audio.

[0125] The initial audio is processed for voice separation based on a preset separation algorithm to obtain corresponding user audio and seat audio.

[0126] In this embodiment, the initial audio may be subjected to Fourier transformation, and the human voice in the initial audio may be separated using a spectral analysis method, thereby separating the user voice and the seat voice, that is, obtaining the above-mentioned user audio and seat audio.

[0127] The user audio is used as the first audio file.

[0128] The present application obtains a preset noise reduction strategy; then performs voice noise reduction processing on the voice call data based on the noise reduction strategy to obtain the corresponding initial audio; then performs human voice separation processing on the initial audio based on a preset separation algorithm to obtain the corresponding user audio and seat audio; and subsequently uses the user audio as the first audio file. The present application obtains the initial audio by performing voice noise reduction processing on the voice call data based on the use of a noise reduction strategy; then performs human voice separation processing on the initial audio based on the use of a preset separation algorithm, thereby realizing intelligent and accurate extraction of the first audio file of the user voice of the target user from the initial audio, thereby improving the extraction efficiency of the first audio file and ensuring the data accuracy of the obtained first audio file.

[0129] In some optional implementations, step S203 includes the following steps:

[0130] A second audio file containing the user voice of the target user is extracted from the voice call data.

[0131] In this embodiment, the implementation process of extracting the second audio file containing the user voice of the target user from the voice call data can refer to the processing process of extracting the first audio file containing the user voice of the target user from the voice call data, and will not be elaborated here.

[0132] The second audio file is divided into corresponding second audio segments.

[0133] In this embodiment, the implementation process of dividing the second audio file into corresponding second audio segments can refer to the processing process of dividing the first audio file into corresponding first audio segments, which will not be described in detail here.

[0134] The average speaking speed of the second audio segment is calculated.

[0135] In this embodiment, the process of calculating the average speaking speed of the second audio segment includes: converting each second audio segment into text, calculating the total number of words in each second audio segment, and dividing the total number of words in each second audio segment by the time of all second audio segments, so as to obtain the average speaking speed of the target user in the second audio segment.

[0136] Attitude dimension features corresponding to the voice call data are generated based on the average speaking speed.

[0137] In this embodiment, the slower the speaking speed, the more friendly and patient the other party will think the target user is. The average speaking speed of the target user in the second audio clip can be regarded as a quantitative indicator of the target user's patience level and used as an attitude dimension feature corresponding to the voice call data.

[0138] The present application extracts a second audio file containing the user voice of the target user from the voice call data; then divides the second audio file into corresponding second audio segments; then calculates the average speaking speed of the second audio segments; and subsequently generates attitude dimension features corresponding to the voice call data based on the average speaking speed, thereby achieving accurate generation of corresponding attitude dimension features based on the use of the average speaking speed of the second audio segments, thereby ensuring the feature accuracy of the obtained attitude dimension features.

[0139] In some optional implementations, step S204 includes the following steps:

[0140] The large language model is called.

[0141] In this embodiment, there is no specific limitation on the selection of the above-mentioned large language model. For example, a model such as BERT and GPT can be used, which has the function of performing in-depth analysis on the user's call text data and extracting the user's focus points, questions and other features.

[0142] The call text data is encoded to obtain a corresponding embedding vector.

[0143] In this embodiment, the call text data is encoded to obtain the corresponding embedding vector.

[0144] Based on the large language model, cognitive dimension feature extraction processing is performed on the embedding vector to obtain a corresponding first specified feature.

[0145] In this embodiment, the embedding vector is input into the above-mentioned large language model, and then the built-in classifier of the large language model or a customized NLP task (such as entity recognition, sentiment analysis, etc.) is applied to extract cognitive dimension features, such as focus points, question points, etc.

[0146] The cognitive dimension feature is generated based on the first specified feature.

[0147] In this embodiment, the first specified feature may be used as the cognitive dimension feature. Alternatively, the first specified feature may be converted into a specific quantitative index or label according to actual needs and used as the cognitive dimension feature.

[0148] When users show a more positive attitude, it is often conducive to the final product transaction. For example, users have questions about customer service, need quotes, and have risk awareness. This application constructs features of three cognitive dimensions, and more cognitive dimension features can be designed according to different scenarios or products. Using a large language model, "whether to ask questions to customer service", "whether to need quotes", and "whether to have risk awareness" are obtained from the call text. Exemplarily, for "whether to have risk awareness", by inputting the call text data of the target user into the large language model, if the user hits the intention under the risk awareness system, and the number of hit intentions is greater than the preset quantity threshold, it is considered that the user has risk awareness, which further indicates that the user has a preliminary positive attitude towards the product. Among them, the corresponding risk awareness system can be pre-built according to actual business needs. Exemplarily, the risk awareness system may include: asking about the impact of comprehensive reform premiums, asking about compulsory traffic insurance, asking about traffic accident insurance, and so on. In addition, there is no specific limitation on the value of the above-mentioned quantity threshold, which can be set according to actual needs.

[0149] The present application calls the large language model, then encodes the call text data to obtain a corresponding embedding vector, then performs cognitive dimension feature extraction processing on the embedding vector based on the large language model to obtain a corresponding first specified feature, and subsequently generates the cognitive dimension feature based on the first specified feature. The present application encodes the call text data to obtain a corresponding embedding vector, and then performs cognitive dimension feature extraction processing on the embedding vector based on the use of a large language model, so that the corresponding cognitive dimension features can be extracted quickly and accurately, thereby improving the extraction efficiency of cognitive dimension features and ensuring the accuracy of the obtained cognitive dimension features.

[0150] In some optional implementations of this embodiment, step S206 includes the following steps:

[0151] Get the preset feature fusion strategy.

[0152] In this embodiment, a feature fusion strategy is pre-designed according to actual processing requirements to combine multiple features to form a comprehensive fusion feature representation. The feature fusion strategy includes but is not limited to concatenation, weighted sum, attention mechanism, principal component analysis (PCA), autoencoder and other strategies.

[0153] Based on the feature fusion strategy, the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature are fused to obtain a fused second designated feature.

[0154] In this embodiment, the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature can be fused according to the fusion processing steps included in the above-mentioned feature fusion strategy to obtain the corresponding fused second specified feature.

[0155] The second designated feature is used as the target feature vector.

[0156] The present application obtains a preset feature fusion strategy; then based on the feature fusion strategy, the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature are fused to obtain the fused second designated feature; and the second designated feature is subsequently used as the target feature vector. The present application fuses the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature based on the use of a feature fusion strategy, which can achieve fast and accurate integration processing between features, improve the processing efficiency of feature fusion processing, ensure the data accuracy of the obtained target feature vector, facilitate the subsequent prediction processing of the target feature vector based on a preset intention recognition model, and can achieve accurate generation of product purchase intention results corresponding to the target user, thereby improving the recognition accuracy and reliability of product purchase intentions.

[0157] In some optional implementations of this embodiment, before step S207, the electronic device may further perform the following steps:

[0158] Acquire pre-collected initial sample data; wherein the initial sample data includes the user's basic information, the user's voice call data, and the corresponding product purchase results.

[0159] In this embodiment, the basic information of the user, such as age, gender, occupation, etc., can be obtained from the structured data source (user database) by using SQL query. The voice call data between the agent and the user and the corresponding purchase results (transaction or non-transaction) are extracted from the recording system or storage library.

[0160] The initial sample data is preprocessed to obtain corresponding first sample data.

[0161] In this embodiment, the above preprocessing may include processing missing values ​​of the initial sample data (such as filling with the median). And one hot encoding the category data in the initial sample data. And processing the outliers in the initial sample data. Calculate the quartiles (Q1, Q2, Q3, Q4) and obtain the quartile distance: IQR = Q3-Q1. Values ​​less than Q1-1.5*IQR or greater than Q3+1.5*IQR are considered outliers, and the outliers in the data set are directly deleted from the initial sample data to obtain the corresponding first sample data.

[0162] Based on a preset feature scoring model and the large language model, feature engineering processing is performed on the first sample data to obtain corresponding second sample data.

[0163] In the present embodiment, the above-mentioned feature scoring model can specifically adopt the XGBoost model. The XGBoost model can be used to score the importance of the first sample data initially extracted, and the features with scores greater than the preset scoring threshold can be screened out. The features after being screened by the XGBoost model can be used as the basis for subsequent feature engineering. In feature engineering, these screened features can be used to further generate corresponding emotional dimension features, cognitive dimension features, and attitude dimension features. These dimensional features are usually obtained by aggregating, converting, or encoding the screened features. In addition, the basic information of the user is obtained and the basic information features are constructed, and then the user's emotional dimension features, attitude dimension features, cognitive dimension features, basic information features, and purchase results are integrated and processed to obtain the second sample data corresponding to the user.

[0164] Recall the preset initial model.

[0165] In this embodiment, the above-mentioned initial model can be a comprehensive model constructed by combining a language model, an XGBoost model and audio processing technology. The model architecture of the initial model includes: 1) Input layer. Text input: Encode text features (cognitive dimension features, basic information features) through models such as BERT to obtain text embedding vectors. Audio input: Directly input audio features (such as emotional dimension features, attitude dimension features, etc.) as numerical features. 2) Hidden layer. By designing a multi-layer neural network (such as a fully connected layer, a convolutional layer, etc.), it is used to learn the interaction and association between text and audio features. It is possible to consider using an attention mechanism to enhance the model's ability to capture key information. 3) Output layer. Use a softmax function or a sigmoid function as the activation function of the output layer to predict purchase intentions (binary or multi-classification problems) based on business needs.

[0166] The initial model is trained and evaluated based on the second sample data to obtain a specified model that meets preset construction conditions.

[0167] In this embodiment, the training and evaluation process of the above-mentioned initial model includes: (1) Training process: divide the second sample data into a training set and a test set. Use the training set data to train the initial model. Set a suitable loss function (such as cross entropy loss) and optimizer (such as Adam optimizer). Monitor the loss function value and accuracy and other indicators during the training process, and perform early stopping through verification to prevent overfitting. And adjust hyperparameters such as learning rate, batch size, and training rounds as needed. (2) Evaluation process: Use the test set data to evaluate the accuracy, recall rate, F1 score and other indicators of the model. Analyze the confusion matrix to understand the performance of the model in different categories, especially the cases of misclassification and missed classification. And draw ROC curves, PR curves, etc. to intuitively display the model performance, and analyze and optimize the model performance until a specified model is obtained whose model performance meets the actual model construction conditions.

[0168] The designated model is used as the intent recognition model.

[0169] The present application obtains the initial sample data collected in advance; then pre-processes the initial sample data to obtain the corresponding first sample data; then performs feature engineering processing on the first sample data based on the preset feature scoring model and the large language model to obtain the corresponding second sample data; subsequently calls the preset initial model, and trains and evaluates the initial model based on the second sample data to obtain a specified model that meets the preset construction conditions; finally, the specified model is used as the intent recognition model. The present application obtains the first sample data by pre-processing the initial sample data collected in advance; then performs feature engineering processing on the first sample data based on the feature scoring model and the large language model to obtain the second sample data; then trains and evaluates the initial model based on the second sample data to obtain the intent recognition model that meets the construction conditions, thereby completing the rapid construction of the intent recognition model, improving the construction efficiency of the intent recognition model, and ensuring the model prediction effect of the generated intent recognition model.

[0170] In some optional implementations, the user information obtained is subject to the user's consent and complies with relevant laws and policies.

[0171] In addition, the non-Company software tools or components that appear in the embodiments of the present application are merely examples and do not represent actual use.

[0172] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0173] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned product purchase intention results, the above-mentioned product purchase intention results can also be stored in a node of a blockchain.

[0174] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0175] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0176] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0177] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0178] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0179] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of an intention recognition device based on artificial intelligence, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0180] like Figure 3 As shown, the artificial intelligence-based intention recognition device 300 described in this embodiment includes: a first acquisition module 301, a first extraction module 302, a second extraction module 303, a third extraction module 304, a generation module 305, an integration module 306 and a prediction module 307.

[0181] in:

[0182] A first acquisition module 301 is used to acquire voice call data corresponding to a target user and call text data corresponding to the voice call data;

[0183] A first extraction module 302, configured to extract features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features;

[0184] The second extraction module 303 is used to extract features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features;

[0185] The third extraction module 304 is used to extract features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features;

[0186] A generating module 305 is used to obtain the basic information of the target user and generate corresponding basic information features based on the basic information;

[0187] An integration module 306 is used to integrate the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector;

[0188] The prediction module 307 is used to perform prediction processing on the target feature vector based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

[0189] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based intention recognition method in the aforementioned implementation mode, and will not be repeated here.

[0190] In some optional implementations of this embodiment, the first extraction module 302 includes:

[0191] A first extraction submodule, configured to extract a first audio file containing a user voice of the target user from the voice call data;

[0192] A first segmentation module, used for segmenting the first audio file into corresponding first audio segments;

[0193] A first calculation submodule, configured to calculate an average pitch of the first audio segment and an average sound intensity of the first audio segment;

[0194] A first generating submodule, configured to generate a corresponding emotional valence level feature based on the average pitch;

[0195] A second generating submodule, configured to generate a corresponding emotional arousal level feature based on the average sound intensity;

[0196] The third generating submodule is used to generate an emotion dimension feature corresponding to the voice call data based on the emotion valence level feature and the emotion arousal level feature.

[0197] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based intention recognition method in the aforementioned implementation mode, and will not be repeated here.

[0198] In some optional implementations of this embodiment, the first extraction submodule includes:

[0199] An acquisition unit, used to acquire a preset noise reduction strategy;

[0200] A processing unit, configured to perform voice noise reduction processing on the voice call data based on the noise reduction strategy to obtain corresponding initial audio;

[0201] A separation unit, configured to perform voice separation processing on the initial audio based on a preset separation algorithm to obtain corresponding user audio and seat audio;

[0202] A determining unit is configured to use the user audio as the first audio file.

[0203] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based intention recognition method in the aforementioned implementation mode, and will not be repeated here.

[0204] In some optional implementations of this embodiment, the second extraction module 303 includes:

[0205] A second extraction submodule, configured to extract a second audio file containing the user voice of the target user from the voice call data;

[0206] A second division submodule, used to divide the second audio file into corresponding second audio segments;

[0207] A second calculation submodule, used to calculate the average speaking speed of the second audio segment;

[0208] The fourth generating submodule is used to generate attitude dimension features corresponding to the voice call data based on the average speaking speed.

[0209] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based intention recognition method in the aforementioned implementation mode, and will not be repeated here.

[0210] In some optional implementations of this embodiment, the third extraction module 304 includes:

[0211] A calling submodule, used for calling the large language model;

[0212] An encoding submodule, used for encoding the call text data to obtain a corresponding embedding vector;

[0213] A third extraction submodule is used to perform cognitive dimension feature extraction processing on the embedding vector based on the large language model to obtain a corresponding first specified feature;

[0214] The fifth generating submodule is used to generate the cognitive dimension feature based on the first specified feature.

[0215] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based intention recognition method in the aforementioned implementation mode, and will not be repeated here.

[0216] In some optional implementations of this embodiment, the integration module 306 includes:

[0217] An acquisition submodule is used to acquire a preset feature fusion strategy;

[0218] A fusion submodule, used for fusing the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature based on the feature fusion strategy to obtain a fused second specified feature;

[0219] A determination submodule is used to use the second specified feature as the target feature vector.

[0220] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based intention recognition method in the aforementioned implementation mode, and will not be repeated here.

[0221] In some optional implementations of this embodiment, the artificial intelligence-based intention recognition device further includes:

[0222] A second acquisition module is used to acquire pre-collected initial sample data; wherein the initial sample data includes basic information of the user, voice call data of the user, and corresponding product purchase results;

[0223] A preprocessing module, used for preprocessing the initial sample data to obtain corresponding first sample data;

[0224] A first processing module, configured to perform feature engineering processing on the first sample data based on a preset feature scoring model and the large language model to obtain corresponding second sample data;

[0225] A calling module is used to call the preset initial model;

[0226] A second processing module, used for training and evaluating the initial model based on the second sample data to obtain a specified model that meets the preset construction conditions;

[0227] A determination module is used to use the specified model as the intent recognition model.

[0228] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based intention recognition method in the aforementioned implementation mode, and will not be repeated here.

[0229] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0230] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0231] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0232] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the intention recognition method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0233] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the artificial intelligence-based intention recognition method.

[0234] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0235] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the artificial intelligence-based intention recognition method as described above.

[0236] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0237] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. An intention recognition method based on artificial intelligence, characterized in that: The steps include: Acquire voice call data corresponding to the target user, and call text data corresponding to the voice call data; Extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features; Extracting features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features; Extracting features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features; Acquire basic information of the target user, and generate corresponding basic information features based on the basic information; Integrate the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector; The target feature vector is predicted based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

2. The intention recognition method based on artificial intelligence according to claim 1, characterized in that: The step of extracting features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features specifically includes: Extracting a first audio file containing the user voice of the target user from the voice call data; dividing the first audio file into corresponding first audio segments; Calculating an average pitch of the first audio segment, and calculating an average sound intensity of the first audio segment; generating a corresponding emotional valence level feature based on the average pitch; generating a corresponding emotional arousal level feature based on the average sound intensity; An emotion dimension feature corresponding to the voice call data is generated based on the emotion valence level feature and the emotion arousal level feature.

3. The intention recognition method based on artificial intelligence according to claim 2, characterized in that: The step of extracting a first audio file containing the user voice of the target user from the voice call data specifically includes: Get the preset noise reduction strategy; Performing voice noise reduction processing on the voice call data based on the noise reduction strategy to obtain corresponding initial audio; Performing voice separation processing on the initial audio based on a preset separation algorithm to obtain corresponding user audio and seat audio; The user audio is used as the first audio file.

4. The intention recognition method based on artificial intelligence according to claim 1, characterized in that: The step of extracting features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features specifically includes: extracting a second audio file containing the user voice of the target user from the voice call data; dividing the second audio file into corresponding second audio segments; Calculating an average speaking speed of the second audio segment; Attitude dimension features corresponding to the voice call data are generated based on the average speaking speed.

5. The intention recognition method based on artificial intelligence according to claim 1, characterized in that: The step of extracting features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features specifically includes: Calling the large language model; Encoding the call text data to obtain a corresponding embedding vector; Perform cognitive dimension feature extraction processing on the embedding vector based on the large language model to obtain a corresponding first specified feature; The cognitive dimension feature is generated based on the first specified feature.

6. The intention recognition method based on artificial intelligence according to claim 1, characterized in that: The step of integrating the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector specifically includes: Get the preset feature fusion strategy; Based on the feature fusion strategy, the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature are fused to obtain a fused second designated feature; The second designated feature is used as the target feature vector.

7. The intention recognition method based on artificial intelligence according to claim 1, characterized in that: Before the step of predicting the target feature vector based on the preset intention recognition model to obtain the product purchase intention result corresponding to the target user, the step further includes: Acquire pre-collected initial sample data; wherein the initial sample data includes basic information of the user, voice call data of the user, and corresponding product purchase results; Preprocessing the initial sample data to obtain corresponding first sample data; Performing feature engineering processing on the first sample data based on a preset feature scoring model and the large language model to obtain corresponding second sample data; Call the preset initial model; Training and evaluating the initial model based on the second sample data to obtain a specified model that meets the preset construction conditions; The designated model is used as the intention recognition model.

8. An intention recognition device based on artificial intelligence, characterized in that: include: A first acquisition module, used to acquire voice call data corresponding to a target user, and call text data corresponding to the voice call data; A first extraction module, configured to extract features from the voice call data based on a preset emotion extraction strategy to obtain corresponding emotion dimension features; A second extraction module, configured to extract features from the voice call data based on a preset attitude extraction strategy to obtain corresponding attitude dimension features; A third extraction module is used to extract features from the call text data based on a preset large language model to obtain corresponding cognitive dimension features; A generating module, used to obtain the basic information of the target user and generate corresponding basic information features based on the basic information; An integration module, used for integrating the emotion dimension feature, the attitude dimension feature, the cognitive dimension feature and the basic information feature to obtain a corresponding target feature vector; The prediction module is used to perform prediction processing on the target feature vector based on a preset intention recognition model to obtain a product purchase intention result corresponding to the target user.

9. A computer device, characterized in that: It includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the artificial intelligence-based intention recognition method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based intention recognition method as described in any one of claims 1 to 7.

Images