LLM-based customer intention recognition and response system, method, device and medium

Abstract

The present invention discloses a customer intent recognition and response system, method, device, and medium based on LLM, which relates to the field of intelligent dialogue technology. The system includes: a multimodal input processing unit for collecting and fusing multi-source interaction data of an intelligent voice outbound call system to form a multi-dimensional input representation; a dynamic context-enhanced LLM unit for performing contextual understanding based on the multi-dimensional input representation and a preset knowledge base to generate a user feature profile; a reinforcement learning dialogue strategy generation unit for dynamically generating a dialogue strategy based on the user feature profile and generating response content based on the dialogue strategy; and a multimodal output unit for converting the response content into multimodal output information and adaptively adjusting the multimodal output information based on real-time scenario parameters. The present invention achieves accurate recognition of user intent and intelligent response, significantly improving the interactive experience and service quality of the intelligent dialogue system.

Description

Technical Field

[0001] The present invention relates to the field of intelligent dialogue technology, and in particular to a customer intent recognition and response system, method, device and medium based on LLM. Background Art

[0002] With the rapid development of Large Language Model (LLM) technology, its application in scenarios such as intelligent customer service and voice outbound calls is becoming increasingly widespread. Existing technologies typically use single-modal speech recognition or text processing to understand user intent and generate responses based on preset dialogue templates. Some systems have introduced emotion recognition modules, attempting to analyze users' voice features to identify their emotional state and optimize dialogue strategies. However, these systems often exhibit limitations when handling actual dialogue scenarios: First, single-modal feature extraction cannot fully capture the user's interaction state, resulting in inaccurate intent understanding; second, response generation based on fixed templates lacks flexibility and is difficult to adapt to complex and changing dialogue scenarios; third, existing systems generally use independent feature processing modules, lacking effective information exchange and collaborative optimization mechanisms between modules, resulting in less than ideal overall system performance.

[0003] More critically, existing technologies suffer from severe contextual understanding bias and attention tracking failures when handling long conversations. Traditional methods typically employ a fixed-window context memory mechanism, which is unable to dynamically adjust information retention strategies based on the user's cognitive state. Furthermore, the generation process of dialogue strategies often overlooks the dynamic changes in user attention, emotional state, and topic preferences, resulting in significant deviations between system responses and actual user needs. Furthermore, existing systems rarely consider the impact of real-time scenario parameters when delivering multimodal output, significantly compromising the naturalness of system responses and the interactive experience. These issues are particularly prominent in scenarios requiring highly personalized and precise services, such as insurance product recommendations.

[0004] Therefore, how to effectively improve the performance of intelligent dialogue systems in actual application scenarios, especially how to accurately identify and promptly respond to changes in users' cognitive states, has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] In view of the above-mentioned problems, the present invention provides a customer intent identification and response system, method, device and medium based on LLM to solve the problems mentioned in the background technology.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a customer intent recognition and response system based on LLM, comprising: a multimodal input processing unit, used to collect and fuse multi-source interaction data of an intelligent voice outbound call system to form a multi-dimensional input representation; a dynamic context-enhanced LLM unit, which performs contextual understanding based on the multi-dimensional input representation and a preset knowledge base to generate a user feature profile; a reinforcement learning dialogue strategy generation unit, used to dynamically generate a dialogue strategy based on the user feature profile, and generate response content based on the dialogue strategy; a multimodal output unit, used to convert the response content into multimodal output information, and adaptively adjust the multimodal output information according to real-time scenario parameters.

[0007] As a preferred solution of the LLM-based customer intention recognition and response system described in the present invention, the multimodal input processing unit obtains multi-source interaction data in the intelligent voice outbound call system in real time through the multi-channel data acquisition module, specifically: obtains the acoustic features of the customer's voice through the audio acquisition channel; transcribes the customer's voice into text information in real time through the voice recognition engine, and extracts the linguistic features in the text; extracts the rhythmic features in the voice through the emotion recognition module; and obtains the background noise features during the call through the environmental acoustic analysis unit.

[0008] As a preferred solution of the LLM-based customer intention recognition and response system described in the present invention, the multimodal input processing unit adopts a multi-level feature fusion strategy to fuse the acquired multi-source interaction data, specifically: normalizing the neural features, linguistic features, prosodic features, and background noise features; weighting the importance of different modal features; mapping the weighted features to a unified feature space through a multimodal feature encoder to generate a feature representation aligned between modalities; and adopting a cross-modal fusion network to organically integrate the features of each modality through the complementarity and synergy between modalities to form a multi-dimensional input representation.

[0009] As a preferred solution of the LLM-based customer intention recognition and response system described in the present invention, the process of generating the user feature portrait includes the following steps: the dynamic context enhanced LLM unit obtains the multi-dimensional input representation and inputs it into the pre-trained feature conversion network to obtain an initial feature vector; retrieves historical interaction records and product information related to the current interaction scenario from the preset knowledge base to generate a knowledge feature vector; and adopts the attention fusion mechanism to combine the initial feature vector with the knowledge feature vector to generate a combined embedding vector.

[0010] A multi-level context window is constructed, and multiple key indicators of the combined embedding vector and the historical interaction information in the window are calculated in each context window. A decreasing attenuation function is designed based on the cross-correlation between the key indicators, and weight coefficients are assigned to context information at different time scales to generate a dynamic weight matrix.

[0011] The combined embedding vector and the dynamic weight matrix are input into a large language model to extract user intent features; the user intent features are multi-headed parsed based on the attention mechanism to obtain an explicit intent vector, an implicit intent vector, and a sentiment tendency vector respectively; the explicit intent vector, the implicit intent vector, and the sentiment tendency vector are fused to generate a user feature portrait.

[0012] As a preferred solution of the LLM-based customer intention identification and response system described in the present invention, the key indicators include semantic similarity, sentiment volatility and topic coherence; the semantic similarity is obtained by calculating the cosine distance between the combined embedding vector and the historical interaction information; the sentiment volatility is obtained by analyzing the variance of sentiment polarity in N consecutive interactions; and the topic coherence is obtained by calculating the similarity of topic vectors in adjacent rounds of interactions.

[0013] Designing a decreasing attenuation function based on the cross-correlation between the key indicators includes: designing a dynamic adjustment rule based on the cross-correlation of the semantic similarity, emotional volatility and topic coherence. Specifically, when the semantic similarity and topic coherence both show an upward trend and the emotional volatility shows a downward trend, it is determined to be a high concentration state, and the attention enhancement factor is calculated. Based on the attention enhancement factor, the shape parameters of the time attenuation curve are updated, the distribution ratio of the weights of each dimension is recalculated using the sliding window method, and the attention state matrix is ​​updated.

[0014] When the sentiment volatility shows an upward trend and the semantic similarity and topic coherence both show a downward trend, it is determined to be an emotion fluctuation state, and an emotion smoothing function is constructed. Based on the emotion smoothing function, the shape parameters of the time decay curve are adjusted, the exponential moving average of historical information is introduced as a stabilizing term, and a temporary weight buffer is established.

[0015] When topic coherence shows a downward trend but semantic similarity remains at a high level and sentiment volatility fluctuates, it is determined to be a topic shift state, and the topic migration coefficient is calculated. Based on the topic migration coefficient, the shape parameters of the time decay curve are corrected, the adaptive learning rate is applied to adjust the weight update speed, and a topic coherence compensation mechanism is established.

[0016] A decreasing attenuation function is generated based on the dynamic adjustment rule; the decreasing attenuation function includes a time weight coefficient, a semantic similarity weight, an emotional weight coefficient and a topic coherence weight, wherein each weight coefficient satisfies the normalization constraint and is a non-negative number; the decreasing attenuation function also includes a time attenuation rate, a baseline attenuation rate and a modulation function corresponding to each state, wherein the modulation function is used to adjust the attenuation curve under different states.

[0017] As a preferred solution of the LLM-based customer intention recognition and response system described in the present invention, wherein: the reinforcement learning dialogue strategy generation unit constructs a multi-objective state-value function based on the user feature portrait and generates a dialogue strategy decision sequence, specifically mapping the explicit intention vector, implicit intention vector and emotional tendency vector into state values ​​of three dimensions: dialogue goal completion, user satisfaction and emotional stability; based on the weight distribution ratio in the dynamic weight matrix, the state values ​​of the three dimensions are weightedly combined to construct a multi-objective state-value function; the Monte Carlo tree search algorithm is used to perform strategy planning for the next N rounds of dialogue, and the action sequence that maximizes the multi-objective state-value function is selected as the dialogue strategy decision sequence.

[0018] Based on the dialogue strategy decision sequence, multiple rounds of dialogue response content are generated, and a dynamic feedback calibration mechanism is established. Specifically, the dialogue strategy decision sequence is matched with product information in a preset knowledge base to generate a candidate response set; the time decay rate in the dynamic weight matrix is ​​used to score the timeliness of historical response templates in the candidate response set; the intensity coefficient of the response tone is adjusted according to the output value of the emotion smoothing function; the rhythm and depth of topic switching are controlled based on the topic migration coefficient; the response content with the highest score is used as the output result of the current round, and the deviation between the response result and the actual interaction effect is fed back to the state value calculation process to optimize the multi-objective state value function.

[0019] As a preferred solution of the LLM-based customer intent recognition and response system described in the present invention, the multimodal output unit receives the response content and converts the text content into multimodal output information in three modes: voice, expression, and body movement based on preset multimodal mapping rules; according to the real-time scene parameters, the multimodal output information is adaptively adjusted, the adjusted multimodal output information is synchronously integrated, and the final interactive response is output; the real-time scene parameters include pitch, speaking speed, and volume; and the multimodal output information includes voice data and visual information.

[0020] To further solve the above technical problems, the present invention provides the following technical solutions: a customer intent recognition and response method based on LLM, comprising: collecting multi-source interaction data and fusing them to generate a multi-dimensional input representation; performing contextual understanding based on the multi-dimensional input representation and a preset knowledge base to generate a user feature profile; dynamically generating a dialogue strategy based on the user feature profile, and generating response content based on the dialogue strategy; converting the response content into multimodal output information, and adaptively adjusting the multimodal output information according to real-time scenario parameters.

[0021] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, it implements the steps of the above-mentioned LLM-based customer intent recognition and response system.

[0022] A computer-readable storage medium stores a computer program thereon, wherein when the computer program is executed by a processor, the computer program implements the steps of the LLM-based customer intent identification and response system as described above.

[0023] Beneficial effects of the present invention: The present invention realizes all-round, dynamic recognition and precise response to customer intentions by constructing a system architecture of a multimodal input processing unit, a dynamic context-enhanced LLM unit, a reinforcement learning dialogue strategy generation unit and a multimodal output unit: through multi-channel data acquisition and multi-level feature fusion, the user's acoustic, linguistic, prosodic and environmental characteristics are fully captured; the dynamic weight matrix and the decreasing attenuation function are used to accurately track the user's attention state and emotional changes; based on the multi-target state value function and the dynamic feedback calibration mechanism, the optimal dialogue strategy and response content are generated; through the adaptive adjustment of multimodal output, the naturalness of the system response and the interactive experience are ensured, thereby significantly improving the performance of the intelligent dialogue system in actual application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 This is a schematic diagram of the structure of the LLM-based customer intention recognition and response system proposed in the present invention;

[0026] Figure 2 This is a computer device diagram of the LLM-based customer intent recognition and response system proposed in the present invention. DETAILED DESCRIPTION

[0027] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0028] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0029] Example 1, reference Figure 1 , which is an embodiment of the present invention, provides a customer intention recognition and response system based on LLM.

[0030] Existing intelligent dialogue systems use a single modal processing method and fixed dialogue templates, which makes it difficult to accurately understand user intentions and generate flexible response content; and traditional systems use a fixed-window context memory mechanism, which cannot dynamically track changes in users' cognitive states and attention distribution, resulting in deviations between dialogue strategies and actual user needs; in addition, existing technologies lack effective fusion and collaborative optimization mechanisms for multimodal features, and rarely consider real-time scene parameters in multimodal output, affecting the naturalness of system responses and the interactive experience.

[0031] This application provides an effective solution to the above-mentioned problems. Next, we will combine multiple embodiments to elaborate on how to implement the LLM-based customer intent recognition and response system.

[0032] Figure 1 This is a structural diagram of the LLM-based customer intent recognition and response system provided by the present invention, which includes a multimodal input processing unit, a dynamic context-enhanced LLM unit, a reinforcement learning dialogue strategy generation unit, and a multimodal output unit.

[0033] The multimodal input processing unit collects and integrates multi-source interaction data of the intelligent voice outbound call system to form a multi-dimensional input representation.

[0034] Specifically, the multimodal input processing unit obtains multi-source interaction data in the intelligent voice outbound call system in real time through the multi-channel data acquisition module, including: obtaining the acoustic characteristics of the customer's voice through the audio acquisition channel, including acoustic parameters such as fundamental frequency, resonance peak, and spectral envelope; transcribing the customer's voice into text information in real time through the speech recognition engine, and extracting linguistic features such as keywords, parts of speech, and syntactic structure in the text; extracting rhythmic features in the voice through the emotion recognition unit, including rhythmic parameters such as pitch, sound intensity, speaking speed, and pauses, which are used to characterize the speaker's emotional state; obtaining background noise characteristics during the call through the environmental acoustic analysis unit, including acoustic environment parameters such as signal-to-noise ratio, noise type, and reverberation degree.

[0035] In order to effectively fuse the above-mentioned multi-source interaction data, the multimodal input processing unit adopts a multi-level feature fusion strategy: first, the various types of original features obtained (acoustic features, linguistic features, prosodic features, and background noise features) are normalized and preprocessed to make features of different dimensions comparable; second, the attention mechanism is used to weight the importance of different modal features, highlighting the features that are more critical to judging the current interaction state; then, the weighted features are mapped to a unified feature space through the multimodal feature encoder to generate inter-modal aligned feature representations; finally, a cross-modal fusion network is used to organically integrate the features of each modality through the complementarity and synergy between modalities to form a multi-dimensional input representation that can comprehensively characterize the current interaction state.

[0036] The specific fusion process can be expressed as:

[0037] 1. Feature normalization. Normalize the original features of each modality to a unified range using methods such as min-max or z-score;

[0038] 2. Attention weighting. Calculate the importance weight of each feature in the current interaction scenario to form a dynamic feature weight matrix;

[0039] 3. Feature encoding. Use a modality-specific encoder to convert raw features into a latent representation of uniform dimension.

[0040] 4. Cross-modal fusion: Through the designed fusion network structure, the correlation and complementarity between modalities are captured to generate the final multi-dimensional representation.

[0041] The optimal multimodal input processing unit takes into account the specificities of intelligent outbound calling scenarios, comprehensively collecting multi-dimensional features such as acoustics, linguistics, emotion, and environment. It also employs a dynamic attention mechanism that adaptively adjusts the importance of features based on the interaction scenario. Furthermore, a specialized cross-modal fusion network is designed to effectively handle the heterogeneity and correlation between modalities. This multi-layered feature extraction and fusion strategy provides rich and accurate input information for subsequent contextual understanding and strategy generation, enhancing the system's intelligent interaction capabilities.

[0042] The dynamic context-enhanced LLM unit performs context understanding based on multi-dimensional input representation and preset knowledge base to generate user feature profiles.

[0043] First, the dynamic context-enhanced LLM unit obtains a multi-dimensional input representation and inputs it into a pre-trained feature conversion network to obtain an initial feature vector. Then, the historical interaction records and product information related to the current interaction scenario are retrieved from the preset knowledge base to generate a knowledge feature vector. Finally, the initial feature vector is combined with the knowledge feature vector using the attention fusion mechanism to generate a combined embedding vector.

[0044] It should be noted that the preset knowledge base includes historical interaction data, product knowledge graphs and the latest market trend information.

[0045] Secondly, a multi-level context window is constructed, and multiple key indicators of the combined embedding vector and the historical interaction information within the window are calculated in each context window. A decreasing attenuation function is designed based on the cross-correlation between the key indicators, and weight coefficients are assigned to the context information at different time scales to generate a dynamic weight matrix.

[0046] Specifically, key metrics include semantic similarity, sentiment volatility, and topic coherence. Semantic similarity is obtained by calculating the cosine distance between the combined embedding vector and historical interaction information; sentiment volatility is obtained by analyzing the variance of sentiment polarity in N consecutive interactions; and topic coherence is obtained by calculating the similarity of topic vectors in adjacent rounds of interactions.

[0047] Furthermore, the design of a decreasing attenuation function based on the cross-correlation between key indicators includes:

[0048] First, a dynamic adjustment rule is designed based on the cross-correlation of semantic similarity, sentiment volatility, and topic coherence. Specifically, when both semantic similarity and topic coherence show an upward trend, and sentiment volatility shows a downward trend, it is determined to be a high-attention state. The attention enhancement factor is calculated, and the shape parameters of the time decay curve are updated based on the attention enhancement factor. The sliding window method is used to recalculate the distribution ratio of the weights of each dimension, and the attention state matrix is ​​updated.

[0049] When the sentiment volatility shows an upward trend and the semantic similarity and topic coherence both show a downward trend, it is determined to be a state of sentiment fluctuation. A sentiment smoothing function is constructed, and the shape parameters of the time decay curve are adjusted based on the sentiment smoothing function. The exponential moving average of historical information is introduced as a stabilizing term, and a temporary weight buffer is established.

[0050] When topic coherence shows a downward trend but semantic similarity remains at a high level and sentiment volatility fluctuates, it is determined to be a topic shift state. The topic migration coefficient is calculated, and the shape parameters of the time decay curve are corrected based on the topic migration coefficient. The adaptive learning rate is applied to adjust the weight update speed, and a topic coherence compensation mechanism is established.

[0051] It should be noted that the attention enhancement factor is calculated by calculating a weighted combination of semantic similarity, the rising slope of topic coherence, and the falling slope of sentiment volatility. It is used to quantify the user's attention focus during the current interaction. The topic migration coefficient is calculated by calculating the ratio of the rate of decline of topic coherence to the stability of semantic similarity. It is used to characterize the user's tendency to switch topics while maintaining semantic understanding. The sentiment smoothing function is calculated by taking an exponentially weighted moving average of sentiment volatility to reduce the impact of sentiment fluctuations on weight distribution.

[0052] Secondly, a decreasing decay function is generated based on dynamic adjustment rules. The decreasing decay function includes a time weight coefficient, a semantic similarity weight coefficient, a sentiment weight coefficient, and a topic coherence weight coefficient. Each weight coefficient satisfies normalization constraints and is non-negative. The decreasing decay function also includes a time decay rate, a baseline decay rate, and a modulation function corresponding to each state. The modulation function is used to adjust the decay curve in different states. Based on the decreasing decay function, weight coefficients are assigned to contextual information at different time scales to generate a dynamic weight matrix.

[0053] Preferably, the present invention can accurately capture the dynamic changes in the user's attention state in actual application scenarios by constructing a multi-level context window and designing a decreasing decay function based on the cross-correlation of three key indicators: semantic similarity, emotional volatility, and topic coherence. In particular, by combining the three typical states of high concentration, emotional fluctuations, and topic deviation with corresponding dynamic adjustment mechanisms, it effectively solves the technical problem of the existing technology that it is difficult to accurately identify and promptly respond to the user's cognitive state transition. This solution not only significantly improves the adaptability to complex scenarios such as distracted attention and emotional fluctuations during long-term user consultations, but also ensures the coherence and stability of the entire process of the dialogue system in actual application through the dynamic weight distribution mechanism of the decreasing decay function.

[0054] Finally, the combined embedding vector and the dynamic weight matrix are input into the large language model to extract the user intent features. Based on the attention mechanism, the user intent features are multi-headedly parsed to obtain the explicit intent vector, implicit intent vector and sentiment tendency vector respectively. The explicit intent vector, implicit intent vector and sentiment tendency vector are fused to generate a user feature profile.

[0055] The reinforcement learning dialogue strategy generation unit dynamically generates dialogue strategies based on user feature profiles and generates response content based on the dialogue strategies.

[0056] First, the reinforcement learning dialogue strategy generation unit constructs a multi-objective state-value function based on the user feature profile and generates a dialogue strategy decision sequence.

[0057] Specifically, the explicit intent vector, implicit intent vector, and emotional tendency vector are mapped into state values ​​of three dimensions: dialogue goal completion, user satisfaction, and emotional stability. Based on the weight distribution ratio in the dynamic weight matrix, the state values ​​of the three dimensions are weightedly combined to construct a multi-objective state value function. The Monte Carlo tree search algorithm is used to plan strategies for the next N rounds of dialogue, and the action sequence that maximizes the multi-objective state value function is selected as the dialogue strategy decision sequence.

[0058] Secondly, multi-round dialogue response content is generated based on the dialogue strategy decision sequence, and a dynamic feedback calibration mechanism is established.

[0059] Specifically, the dialogue strategy decision sequence is matched with the product information in the preset knowledge base to generate a set of candidate responses; the time decay rate in the dynamic weight matrix is ​​used to score the timeliness of the historical response templates in the candidate response set; the intensity coefficient of the response tone is adjusted according to the output value of the emotion smoothing function; the rhythm and depth of topic switching are controlled based on the topic migration coefficient; the response content with the highest score is used as the output result of the current round, and the deviation between the response result and the actual interaction effect is fed back to the state value calculation process to optimize the multi-objective state value function.

[0060] Optimally, the reinforcement learning dialogue strategy generation unit organically integrates the dynamic weight matrix and user feature profile output by the dynamic context-enhanced LLM unit into the reinforcement learning framework, enabling dynamic tracking and response to the user's cognitive state during the dialogue strategy generation process. A multi-objective state-value function is used to uniformly model the dialogue goal, user satisfaction, and emotional stability, and a dynamic weight matrix is ​​used to achieve an adaptive trade-off between these three goals. Furthermore, in intelligent outbound call scenarios, the system can automatically adjust the dialogue strategy based on the user's attention state, emotional changes, and topic preferences, making the product recommendation process more natural and smooth.

[0061] Table 1 Performance comparison and evaluation of intelligent outbound call systems based on dynamic context enhancement

[0062] Evaluation Metrics Traditional fixed strategy The present invention Improvement Average conversation duration (seconds) 127 159 25.2％ Transaction conversion rate (%) 4.2 4.83 15.0％ User interruption rate (%) 28.5 19.9 －30.2％ User satisfaction rating (1-5 points) 3.6 4.1 13.9％ Topic completeness (%) 62.3 78.6 26.2％ Negative emotion recognition accuracy (%) 71.2 86.7 21.8％ Dynamic response delay (ms) 850 720 －15.3％ Cost per call (yuan) 2.8 2.95 5.4％

[0063] As shown in Table 1, compared with the traditional fixed strategy, this solution increased the average user conversation duration by 25.2%, boosted the transaction conversion rate by 15.0%, and reduced the user interruption rate by 30.2%. This improvement is primarily due to the system's ability to promptly capture and respond to user state changes, demonstrating the synergistic effect of dynamic context enhancement and reinforcement learning strategy generation.

[0064] The multimodal output unit converts the response content into multimodal output information and adaptively adjusts the multimodal output information according to real-time scene parameters.

[0065] First, the multimodal output unit receives the response content and converts the text content into output information in three modes: speech, expression, and body movement based on the preset multimodal mapping rules.

[0066] Specifically, a pre-trained Chinese TTS model is used to convert text into speech signals; based on the semantic and emotional labels of the response content, a matching expression sequence is selected from a preset expression action library; and based on the tone intensity coefficient of the response content, corresponding body movement control instructions are generated.

[0067] Then, the multimodal output information is adaptively adjusted according to the real-time scene parameters.

[0068] Specifically, the voice output volume is dynamically adjusted based on the ambient noise level; the frequency of facial expression changes is adjusted according to the user's attention state; and the appropriate combination of body movements is selected according to the type of call scenario.

[0069] Finally, the adjusted multimodal output information is synchronously integrated to output the final interactive response.

[0070] Specifically, a timing alignment algorithm is used to ensure the output timing coordination of speech, expression and action; the preset modal fusion rules are used to eliminate conflicts between different modalities; and the integrated output information is passed to the corresponding execution module.

[0071] In summary, the present invention realizes all-round and dynamic recognition and precise response to customer intentions by constructing a system architecture of a multimodal input processing unit, a dynamic context-enhanced LLM unit, a reinforcement learning dialogue strategy generation unit and a multimodal output unit: through multi-channel data acquisition and multi-level feature fusion, the user's acoustic, linguistic, prosodic and environmental characteristics are fully captured; the dynamic weight matrix and the decreasing attenuation function are used to accurately track the user's attention state and emotional changes; based on the multi-target state value function and the dynamic feedback calibration mechanism, the optimal dialogue strategy and response content are generated; through the adaptive adjustment of multimodal output, the naturalness of the system response and the interactive experience are ensured, thereby significantly improving the performance of the intelligent dialogue system in actual application scenarios.

[0072] Example 2, an embodiment of the present invention, provides a method for identifying and responding to customer intent based on LLM, including:

[0073] S1: Collect multi-source interaction data and fuse them to generate multi-dimensional input representation;

[0074] S2: Context understanding is performed based on the multi-dimensional input representation and the preset knowledge base to generate a user feature profile;

[0075] S3: Dynamically generate a dialogue strategy based on the user feature profile, and generate response content based on the dialogue strategy;

[0076] S4: Convert the response content into multimodal output information, and adaptively adjust the multimodal output information according to real-time scene parameters.

[0077] Example 3, reference Figure 2 , is an embodiment of the present invention, which differs from the previous embodiment in that: if the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program code.

[0078] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0079] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0080] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0081] Example 4 is an embodiment of the present invention, which provides a customer intention recognition and response system based on LLM. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0082] To verify the performance of this invention in real-world applications, the intelligent outbound call system of an insurance company's customer service center was selected as the test platform. 1,000 customers were randomly selected and divided equally into five groups. Comparative tests were conducted using different system configurations. The test environment was a standard call center soundproof room, with ambient noise controlled below 45dB. The system hardware configuration utilized an Intel Xeon E5-2680v4 processor, 384GB of memory, and an NVIDIA A100 GPU for deep learning model acceleration.

[0083] In the multimodal input processing unit, customer voice signals are collected using a high-fidelity microphone array with a 16kHz sampling rate. The Wav2Vec 2.0 model is used to extract acoustic features, including 13-dimensional acoustic parameters such as pitch, energy, and spectral centroid. Speech recognition utilizes a real-time transcription engine based on the Transformer architecture, keeping the word error rate (WER) below 4.2%. The emotion recognition module extracts prosodic features based on the ResNet-18 network, including speech rate, pauses, and pitch changes. Environmental acoustic analysis utilizes an adaptive noise suppression algorithm to achieve over 95% accuracy in extracting background noise features.

[0084] The feature fusion strategy adopts a three-layer attention network. The first layer performs intra-modal feature normalization and maps features of different scales to the interval [-1, 1]. The second layer calculates feature weights through a self-attention mechanism, with acoustic features having a weight of 0.3, language features having a weight of 0.4, emotional features having a weight of 0.2, and environmental features having a weight of 0.1. The third layer uses a cross-attention mechanism to achieve inter-modal feature alignment and generate a unified feature vector of 768 dimensions.

[0085] The dynamic context-enhanced LLM unit uses an improved GPT-3 model as its base architecture, with the context window size dynamically adjusted between [128, 512]. In the decreasing decay function, the baseline decay rate is set to 0.85, and the time decay curve uses a modified exponential function. The attention state determination thresholds are set as follows: semantic similarity > 0.75, sentiment volatility < 0.15, and topic coherence > 0.8 for a high attention state.

[0086] Reinforcement learning policy generation uses an improved PPO algorithm, with reward function weights assigned as follows: dialogue goal completion 0.4, user satisfaction 0.4, and sentiment stability 0.2. Monte Carlo tree search depth is set to 4 rounds, with 100 samples per round. Response generation uses kernel sampling with a temperature coefficient of 0.7.

[0087] Table 2 Comparison of performance indicators under different system configurations

[0088]

[0089] Table 3 System adaptability test results in different scenarios

[0090]

[0091] Analysis of the experimental data reveals the following conclusions: First, in terms of core performance indicators, the complete system of this invention achieves significant improvements over both traditional single-modal systems and basic LLM systems. Intent recognition accuracy increased by 21.8 percentage points to 94.3%, thanks to the effectiveness of the multimodal feature fusion strategy. Response relevance increased by 24.5 percentage points to 92.8%, validating the importance of the dynamic context enhancement mechanism. User satisfaction and conversation completion rates increased by 24.7 and 23 percentage points, respectively, fully demonstrating the advantages of reinforcement learning strategy generation.

[0092] Furthermore, the present invention achieved a breakthrough in system response latency, reducing the average response latency from 850ms to 450ms while increasing system resource utilization by only 23 percentage points, demonstrating a significant advantage in optimizing algorithm efficiency. In tests of adaptability across different scenarios, the system maintained high feature fusion accuracy (86.9%) and strategy generation efficiency (85.8%), even in complex background environments, demonstrating the present invention's excellent environmental adaptability and robustness.

[0093] A cross-analysis of the data in Tables 2 and 3 reveals that the present invention exhibits significantly less performance degradation than the control group when handling complex scenarios such as high noise and emotional fluctuations. For example, when transitioning from a quiet indoor environment to a complex background, feature fusion accuracy only drops by 8.3 percentage points, while traditional systems typically experience a 15-20 percentage point performance drop under similar conditions. This fully demonstrates the significant technical advantages of the dynamic context enhancement mechanism and adaptive adjustment strategy proposed in this invention. System stability consistently remains above 95%, demonstrating the high reliability and practical value of this invention in engineering implementation.

[0094] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. The customer intention recognition and response system based on LLM is characterized by: include: Multimodal input processing unit, used to collect and fuse multi-source interaction data of the intelligent voice outbound calling system to form a multi-dimensional input representation; A dynamic context-enhanced LLM unit, which performs context understanding based on the multi-dimensional input representation and a preset knowledge base to generate a user feature profile; A reinforcement learning dialogue strategy generation unit, configured to dynamically generate a dialogue strategy based on the user feature profile and generate response content based on the dialogue strategy; a multimodal output unit, configured to convert the response content into multimodal output information and adaptively adjust the multimodal output information according to real-time scene parameters; The process of generating the user feature profile includes the following steps: The dynamic context enhanced LLM unit obtains the multi-dimensional input representation and inputs it into a pre-trained feature conversion network to obtain an initial feature vector; retrieves historical interaction records and product information related to the current interaction scenario from the preset knowledge base to generate a knowledge feature vector; and uses an attention fusion mechanism to combine the initial feature vector with the knowledge feature vector to generate a combined embedding vector; Constructing a multi-level context window, and calculating multiple key indicators of the combined embedding vector and the historical interaction information within the window in each context window; designing a decreasing decay function based on the cross-correlation between the key indicators, assigning weight coefficients to context information at different time scales, and generating a dynamic weight matrix; Inputting the combined embedding vector and the dynamic weight matrix into a large language model to extract user intent features; Performing multi-head analysis on the user intention features based on the attention mechanism to obtain an explicit intention vector, an implicit intention vector, and an emotional tendency vector respectively; fusing the explicit intention vector, the implicit intention vector, and the emotional tendency vector to generate a user feature profile; The key indicators include semantic similarity, sentiment volatility and topic coherence; Designing a decreasing decay function based on the cross-correlation between the key indicators includes: Designing dynamic adjustment rules based on the cross-correlation of semantic similarity, sentiment volatility, and topic coherence; A decreasing decay function is generated based on the dynamic adjustment rule.

2. The LLM-based customer intention recognition and response system according to claim 1, characterized in that: The multimodal input processing unit acquires multi-source interaction data in the intelligent voice outbound call system in real time through the multi-channel data acquisition module, specifically: Acquire the acoustic characteristics of the customer's voice through the audio collection channel; The speech recognition engine transcribes customer speech into text information in real time and extracts linguistic features from the text; Extract prosodic features from speech through the emotion recognition module; The background noise characteristics during the call are obtained through the environmental acoustic analysis unit.

3. The LLM-based customer intention recognition and response system according to claim 2, characterized in that: The multimodal input processing unit adopts a multi-level feature fusion strategy to fuse the acquired multi-source interaction data, specifically: Normalizing the semantic features, linguistic features, prosodic features, and background noise features; Importance weighting of different modal features; The weighted features are mapped to a unified feature space through a multimodal feature encoder to generate inter-modality aligned feature representations. A cross-modal fusion network is used to organically integrate the features of each modality through the complementarity and synergy between modalities to form a multi-dimensional input representation.

4. The LLM-based customer intention recognition and response system according to claim 3, characterized in that: The semantic similarity is obtained by calculating the cosine distance between the combined embedding vector and the historical interaction information; The sentiment fluctuation rate is obtained by analyzing the variance of sentiment polarity in N consecutive interactions; The topic coherence is obtained by calculating the similarity of topic vectors of adjacent rounds of interaction; Designing a decreasing decay function based on the cross-correlation between the key indicators includes: When both the semantic similarity and topic coherence show an upward trend, and the emotional volatility shows a downward trend, it is determined to be a high-attention state, and an attention enhancement factor is calculated. Based on the attention enhancement factor, the shape parameters of the time decay curve are updated, the distribution ratio of the weights of each dimension is recalculated using the sliding window method, and the attention state matrix is ​​updated; When the sentiment fluctuation rate shows an upward trend and the semantic similarity and topic coherence both show a downward trend, it is determined to be a mood fluctuation state, and a mood smoothing function is constructed. Based on the mood smoothing function, the shape parameters of the time decay curve are adjusted, the exponential moving average of historical information is introduced as a stabilizing term, and a temporary weight buffer is established; When topic coherence shows a downward trend but semantic similarity remains high and sentiment volatility fluctuates, it is determined to be a topic drift state, and the topic migration coefficient is calculated. The shape parameters of the time decay curve are modified based on the topic migration coefficient, the weight update speed is adjusted using an adaptive learning rate, and a topic coherence compensation mechanism is established; The decreasing attenuation function includes a time weight coefficient, a semantic similarity weight, an emotional weight coefficient and a topic coherence weight, wherein each weight coefficient satisfies the normalization constraint and is non-negative; the decreasing attenuation function also includes a time attenuation rate, a baseline attenuation rate and a modulation function corresponding to each state, wherein the modulation function is used to adjust the attenuation curve under different states.

5. The LLM-based customer intention recognition and response system according to claim 4, characterized in that: The reinforcement learning dialogue strategy generation unit constructs a multi-objective state-value function based on the user feature profile and generates a dialogue strategy decision sequence. Specifically, the explicit intention vector, implicit intention vector, and emotional tendency vector are mapped into state values ​​of three dimensions: dialogue goal completion, user satisfaction, and emotional stability. Based on the weight distribution ratio in the dynamic weight matrix, the state values ​​of the three dimensions are weighted and combined to construct a multi-objective state-value function. A Monte Carlo tree search algorithm is used to perform strategy planning for the next N rounds of dialogue, and an action sequence that maximizes the multi-objective state-value function is selected as the dialogue strategy decision sequence. Based on the dialogue strategy decision sequence, multiple rounds of dialogue response content are generated, and a dynamic feedback calibration mechanism is established. Specifically, the dialogue strategy decision sequence is matched with product information in a preset knowledge base to generate a candidate response set; the time decay rate in the dynamic weight matrix is ​​used to score the timeliness of historical response templates in the candidate response set; adjusting the intensity coefficient of the response tone according to the output value of the emotion smoothing function; Based on the topic migration coefficient, the rhythm and depth of topic switching are controlled; the response content with the highest score is used as the output result of the current round, and the deviation between the response result and the actual interaction effect is fed back to the state value calculation process to optimize the multi-objective state value function.

6. The LLM-based customer intention recognition and response system according to claim 5, characterized in that: The multimodal output unit receives the response content and converts the text content into multimodal output information in three modes: voice, expression, and body movement based on a preset multimodal mapping rule; adaptively adjusts the multimodal output information according to the real-time scene parameters; synchronously integrates the adjusted multimodal output information, and outputs a final interactive response; The real-time scene parameters include pitch, speech speed and volume; The multimodal output information includes voice data and visual information.

7. A method for customer intent recognition and response based on LLM, based on the customer intent recognition and response system based on LLM according to any one of claims 1 to 6, characterized in that: include, Collect multi-source interaction data and fuse them to generate multi-dimensional input representation; Perform context understanding based on the multi-dimensional input representation and the preset knowledge base to generate a user feature profile; Dynamically generate a dialogue strategy based on the user feature profile, and generate response content based on the dialogue strategy; Converting the response content into multimodal output information, and adaptively adjusting the multimodal output information according to real-time scene parameters; The process of generating the user feature profile includes the following steps: The dynamic context enhanced LLM unit obtains the multi-dimensional input representation and inputs it into a pre-trained feature conversion network to obtain an initial feature vector; retrieves historical interaction records and product information related to the current interaction scenario from the preset knowledge base to generate a knowledge feature vector; and uses an attention fusion mechanism to combine the initial feature vector with the knowledge feature vector to generate a combined embedding vector; Constructing a multi-level context window, and calculating multiple key indicators of the combined embedding vector and the historical interaction information within the window in each context window; designing a decreasing decay function based on the cross-correlation between the key indicators, assigning weight coefficients to context information at different time scales, and generating a dynamic weight matrix; Inputting the combined embedding vector and the dynamic weight matrix into a large language model to extract user intent features; Performing multi-head analysis on the user intention features based on the attention mechanism to obtain an explicit intention vector, an implicit intention vector, and an emotional tendency vector respectively; fusing the explicit intention vector, the implicit intention vector, and the emotional tendency vector to generate a user feature profile; The key indicators include semantic similarity, sentiment volatility and topic coherence; Designing a decreasing decay function based on the cross-correlation between the key indicators includes: Designing dynamic adjustment rules based on the cross-correlation of semantic similarity, sentiment volatility, and topic coherence; A decreasing decay function is generated based on the dynamic adjustment rule.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the LLM-based customer intention identification and response method described in claim 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the LLM-based customer intention identification and response method according to claim 7 are implemented.

Images