Intelligent charging decision-making method and device based on large model

Abstract

The invention provides an intelligent charging decision-making method and device based on a large model. The method comprises the steps that target input of a user is converted into text data; identifying the text data based on a preset intention model to obtain a dialogue intention and domain information, and determining a numerical value of the target slot according to the text data, the dialogue intention and the domain information; performing charging decision based on a decision model according to the dialogue intention, the domain information and the numerical value of the target slot position to obtain a charging decision result, and determining a charging service according to the charging decision result; the decision model is obtained by mathematical modeling based on a sample dialogue intention, sample field information, a sample slot numerical value and a charging influence factor as inputs and a charging emotion score as an output. According to the method, the charging decision-making efficiency and the charging efficiency are improved, the user operation is simplified, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a large model-based intelligent charging decision method and device. Background Art

[0002] At a time when every aspect of life is profoundly influenced by artificial intelligence technology, actively embracing artificial intelligence technology to empower business growth is both a general trend and a business necessity.

[0003] In daily business, existing charging decision-making methods focus on the ranking of recommended content, that is, modeling user interests, so as to put the content that users are most interested in first and the content that users are less interested in second. The more recommended content, the more options there are, which makes the user's charging operation complicated and the charging decision more difficult. In addition, before charging, users often consider multiple factors such as user behavior, charging time or price, and query charging methods suitable for different users in different travel scenarios. However, existing charging decision-making methods often cannot reflect the actual charging habits of different users, resulting in inaccurate charging decision results and low charging efficiency. Summary of the invention

[0004] The present invention provides a large-model-based intelligent charging decision method and device, which are used to solve the problem that the prior art focuses on the sorting of charging recommendation content, and the recommended content is applicable to a single charging scenario, which leads to complex charging operations and difficult decisions for users, thereby resulting in low charging decision efficiency, thereby improving the accuracy of charging decisions and charging efficiency.

[0005] The present invention provides a large-model-based intelligent charging decision method, comprising: Converting a user's target input into text data; wherein the target input includes at least one of a touch input and a voice input; Recognize the text data based on a preset intention model to obtain a dialogue intention and domain information, and determine a value of a target slot according to the text data, the dialogue intention and the domain information; Based on the decision model, a charging decision is made according to the conversation intention, the domain information and the value of the target slot to obtain a charging decision result, and a charging service is determined according to the charging decision result; wherein the decision model is based on mathematical modeling with sample conversation intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

[0006] According to a large model-based intelligent charging decision method provided by the present invention, the decision model is obtained through the following steps: Mapping the sample conversation intention, the sample domain information, and the sample slot value to a high-dimensional space vector to obtain a mapped sample; A sample data set is constructed according to the mapped samples and the charging influencing factors, the sample data set is used as a training sample, and the long short-term memory network LSTM is iteratively trained with the numerical labels obtained by converting the sentiment scores as training labels, and the decision model is obtained when a preset number of training times is reached; the sentiment score is obtained by weighted calculation based on at least two of the charging influencing factors.

[0007] According to a large model-based intelligent charging decision method provided by the present invention, after obtaining the decision model, the method further includes: Obtain real-time text or voice recording data, real-time charging decision data, and real-time charging call service data; The decision model is updated online based on the real-time text or voice recording data and the real-time charging decision data to obtain a new decision model; and the mapping relationship between the charging decision result and the charging service is updated according to the real-time charging calling service data.

[0008] According to a large model-based intelligent charging decision method provided by the present invention, the charging service includes a backend calling service and a hardware calling service; Determining the charging service according to the charging decision result includes: Determine the target charging type according to the charging decision result, A calling instruction is determined according to the target charging mode to call a corresponding charging service; different target charging types are associated with different charging services through a mapping dictionary.

[0009] According to a large model-based intelligent charging decision method provided by the present invention, after determining the value of the target slot according to the text data, the conversation intention and the domain information, the method further includes: Performing TF-IDF calculation on the text data to obtain a TF-IDF value; The conversation intention, the domain information and the target slot value are respectively matched with the TF-IDF value to obtain a keyword matching result.

[0010] The present invention also provides a large model-based intelligent charging decision device, comprising: A data acquisition module, used to convert a user's target input into text data; wherein the target input includes at least one of a touch input and a voice input; An intention recognition module, used to recognize the text data based on a preset intention model, obtain the dialogue intention and domain information, and determine the value of the target slot according to the text data, the dialogue intention and the domain information; A charging decision module is used to make a charging decision based on a decision model according to the conversation intention, the domain information and the value of the target slot, obtain a charging decision result, and determine the charging service according to the charging decision result; wherein the decision model is based on mathematical modeling with sample conversation intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

[0011] According to a large model-based intelligent charging decision device provided by the present invention, the device also includes: The intention verification module is used to perform TF-IDF calculation on the text data to obtain a TF-IDF value after determining the value of the target slot according to the text data, the dialogue intention and the domain information; and perform keyword matching on the dialogue intention, the domain information and the value of the target slot with the TF-IDF value respectively to obtain a keyword matching result.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the intelligent charging decision method based on the large model as described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the computer program implements any of the above-mentioned large-model-based intelligent charging decision methods.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned large model-based intelligent charging decision methods.

[0015] The large-model-based intelligent charging decision method and device provided by the present invention recognize the text data corresponding to the user input information through the intention model, and determine the value of the target slot according to the text data, the conversation intention and the domain information, and finally use the decision model to make a charging decision according to the conversation intention, the domain information and the value of the target slot to obtain the charging decision result, and determine the charging service according to the charging decision result, thereby improving the charging decision efficiency and charging efficiency, simplifying the user operation and improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is one of the flow charts of the intelligent charging decision method based on the large model provided by the present invention.

[0018] Figure 2 This is the second flow chart of the large model-based intelligent charging decision method provided by the present invention.

[0019] Figure 3 It is a structural schematic diagram of the large model-based intelligent charging decision-making device provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Combine the following Figure 1-Figure 3 The invention describes a large model-based intelligent charging decision method and device.

[0023] Figure 1 The present invention provides a flow chart of a smart charging decision method based on a large model, such as Figure 1 As shown, the method includes the following: Step 110: Convert the user's target input into text data; wherein the target input includes at least one of touch input and voice input.

[0024] In this step, touch input generally involves operations performed by the user on a touch screen or a touch pad, such as the user selecting a corresponding function service through touch inputs such as clicking, sliding, and long pressing on the APP display interface on the screen.

[0025] For example, when a user types letters, numbers, or symbols on a touch screen keyboard, the device recognizes these typing actions and converts them into corresponding characters. The device then combines the characters entered by the user into words, phrases, or sentences in the order in which they were entered. Finally, these character combinations are integrated into complete text data for subsequent processing or display.

[0026] In this step, the terminal device collects the user's voice signal through a microphone, and performs preprocessing steps such as denoising, filtering, and feature extraction on the collected voice signal to extract key information for recognition. The preprocessed voice signal is then segmented, compared, and decoded using a machine learning or deep learning model to identify the most likely text content and convert it into text data. The identified text data may need to be further corrected and optimized to improve accuracy and readability. This can be achieved through natural language processing (NLP) technology, such as grammar checking, spelling correction, and context understanding.

[0027] In this real example, the user inputs into the terminal device via voice, "Charge for 30 minutes at noon. I need to eat nearby during the charging time. Please give a specific plan." The terminal device recognizes the voice content and generates text data including the above content.

[0028] Step 120: Identify the text data based on a preset intention model to obtain the dialogue intention and domain information, and determine the value of the target slot based on the text data, dialogue intention and domain information.

[0029] In this step, before inputting the text data into the preset intent model for recognition, the text data input by the user can be cleaned to remove irrelevant characters, redundant information and noise data through preprocessing to improve the quality of the text data; in addition, natural language processing tasks such as word segmentation and part-of-speech tagging can be performed on the text data to provide reliable data support for subsequent intent recognition.

[0030] In this embodiment, the model has been trained with text data from a large number of charging scenarios and is able to identify intentions related to charging; for example, the model extracts and classifies features of text data through machine learning or deep learning algorithms, thereby identifying the intent of the conversation, such as "query charging stations", "make an appointment for charging", "pay charging fees", etc.

[0031] In this embodiment, different dialogue intentions correspond to different domain information. For example, the domain information related to "query charging station" includes the IP level of the charging pile, the charging protocol, or fast charging and slow charging, etc.

[0032] In this embodiment, the preprocessed text data is input into the pre-trained intent model, the corresponding dialogue intent is output, and the domain-related information in the text is extracted by combining the identified intent and the domain knowledge base (including professional terms in the charging field, charging station information, charging equipment types, etc.).

[0033] In this embodiment, key slots that need to be filled are defined according to the specific requirements of the charging scenario. These slots may include the charging station name, charging equipment number, charging time, charging cost, etc.

[0034] In this embodiment, the corresponding information is automatically extracted and filled in according to the identified intent and domain information, as well as the defined key slots. For example, if the identified intent is "query charging station", the name and location information of the charging station can be extracted from the text data and filled in the corresponding slot, and the automatically filled slot information is verified to ensure its accuracy and completeness.

[0035] Step 130: Make a charging decision based on the decision model according to the conversation intention, domain information and the value of the target slot, obtain a charging decision result, and determine the charging service based on the charging decision result; wherein the decision model is based on mathematical modeling with sample conversation intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

[0036] In this step, charging influencing factors include but are not limited to average price of historical orders, average order distance, fast or slow charging, average power, charging time, whether or not the user is a platform VIP, gender, age, and frequency of use of the charging assistant app and other user portrait data.

[0037] In this embodiment, the decision model performs sentiment analysis based on the intention information, field information, slot values ​​and multiple charging influencing factors corresponding to each charging station as input, and outputs sentiment scores such as distance score, price score, fast charging score, supporting service score and the total score after weighted calculation of the above scores. The range of each score is between 0 and 1. Accordingly, each score is mapped to the decision dimensions of distance, price, fast charging, supporting services and comprehensive evaluation. The higher the score, the less sensitive the dimension, and the lower the score, the more sensitive the dimension.

[0038] For example, a user goes to A Square to eat and suddenly finds that his car is out of power and needs to find a charging station near A Square. The user clicks the Charging Rabbit icon on the Charging Assistant APP page to enter the Charging Rabbit interactive page. For the two charging stations A and B, the user expresses his needs through voice or text input. If it is voice input, the user's voice will be recognized as text.

[0039] In this embodiment, if the user inputs the voice: "I want to go to A Square to eat, and my car is low on power. Please help me find a cheap charging station. I hope I can charge my car while I eat"; the voice data is converted into text data through a text conversion tool, and the text data is input into the intention model to obtain the domain to which the conversation belongs and the user's intention; (domain: charging, intention: find a charging station, charge), and then use the intention model to analyze the domain, intention and user input information to obtain the value of the key slot corresponding to the user's intention (key slot: Wanda Plaza, cheap price, charge while eating); the key slot is numerically represented (for example, represented by a numerical value such as 0, 1, 2), and then the trained decision model is used to make a decision based on the domain, intention, and slot corresponding information obtained above to obtain a feasible decision result (path information for navigating to the target charging station).

[0040] Finally, the charging service is called according to the decision result (including but not limited to calling the camera to check the road conditions or parking space information before arriving at the target charging station, triggering voice reminders such as the start time of charging, the end time of charging, and the end of charging, etc.).

[0041] The large-model-based intelligent charging decision method provided by the present invention recognizes text data corresponding to user input information through an intention model, and determines the value of a target slot according to the text data, conversation intention and domain information, and finally uses the decision model to make a charging decision according to the conversation intention, domain information and the value of the target slot to obtain a charging decision result, and determines the charging service according to the charging decision result, thereby improving the charging decision efficiency and charging efficiency, simplifying user operations and improving user experience.

[0042] In some embodiments, the decision model is obtained by the following steps: (1) Map the sample conversation intention, sample domain information, and sample slot value to a high-dimensional space vector to obtain the mapped sample.

[0043] In this embodiment, user input record data is collected from channels such as actual conversations, user behavior records, and charging records and converted into text to obtain sample text data, which is then processed through an intent model to obtain corresponding sample conversation intents, sample domain information, and the number of sample slots.

[0044] In this embodiment, an encoding tool may be used to vectorize the sample conversation intent, sample domain information, and sample slot number.

[0045] For example, the encoding tool can be a Word2Vec model, a BERT model, or a GPT model.

[0046] In this embodiment, in order to obtain a unified dialogue representation, the intention vector, domain vector and slot vector obtained above can be further combined; wherein, the combination method may include: vector concatenation, vector weighted average or linear fitting of multiple vectors through a neural network for unified representation; finally, a vector in a high-dimensional space is obtained, i.e., a mapped sample; the mapped sample represents the intention, domain information and slot value of the dialogue sample, which is used for subsequent machine learning or deep learning model for sentiment analysis.

[0047] (2) A sample data set is constructed based on the mapped samples and charging influence factors. The sample data set is used as training samples, and the numerical labels obtained by converting the sentiment scores are used as training labels to iteratively train the long short-term memory network (LSTM). When the preset number of training times is reached, a decision model is obtained. The sentiment score is obtained by weighted calculation based on at least two of the charging influence factors.

[0048] In this embodiment, the operation mechanism of LSTM (Long Short-Term Memory) includes: combining long memory with segment memory to remember the context and prevent gradient vanishing or gradient exploding; specifically, LSTM selectively remembers or forgets information through a forget gate, an input gate, and an output gate, and prevents overfitting in training by adding a dropout layer; a fully connected layer is connected after the LSTM layer to map the output vector of the LSTM to a specific range (such as between 0 and 1) to represent the sentiment score.

[0049] In this embodiment, after obtaining the sample data, the sample data can be labeled as follows: including conversation intention (such as querying charging stations, making an appointment for charging, etc.), domain information (such as charging domain), target slot values ​​(such as expected charging amount, expected charging time, etc.), and then setting weights for different charging influencing factors, scoring each labeled sample in combination with the weights, and converting the score into a numerical label as a training label for each sample; finally, the sample data and the numerical label are input into the initial LSTM for iterative training. During the training process, the model is trained using the labeled sample data so that the model can learn the relationship between conversation intention, domain information, target slot values, charging influencing factors and charging sentiment scores, and obtain a trained decision model when the maximum number of iterations is reached or the model converges.

[0050] In this embodiment, after obtaining the decision model, the trained model can be evaluated using the test set data to calculate the model's accuracy, recall rate, F1 score and other indicators, and the model's prediction performance can be improved by adjusting the model's hyperparameters (such as learning rate, number of iterations, regularization coefficient, etc.).

[0051] The large-model-based intelligent charging decision method provided by the present invention maps sample conversation intentions, sample domain information, and sample slot values ​​to high-dimensional space vectors to obtain mapped samples; then a sample data set is constructed according to the mapped samples and charging influencing factors, and the sample data set is used as a training sample, and the long short-term memory network LSTM is iteratively trained with the numerical labels obtained by converting the sentiment scores as training labels to obtain a decision model. By performing sentiment analysis on multiple charging influencing factors of users and establishing a correlation between user requests and charging decisions, the charging needs of different users can be met, and the generalization ability of the decision model and the accuracy of the decision results are improved.

[0052] In some embodiments, after obtaining the decision model, the intelligent charging decision method based on the large model further includes: (1) Obtain real-time text or voice recording data, real-time charging decision data, and real-time charging call service data.

[0053] In this embodiment, advanced speech recognition technology (such as intelligent speech-to-text function) is used to convert real-time speech data into text records, and the text data is preprocessed such as word segmentation and stop word removal to obtain real-time data input by the user, that is, real-time text record data.

[0054] In this embodiment, charging decision data are acquired in real time from a charging management system or related data sources. These data may include charging strategies, charging time, charging power, etc. These data are cleaned and formatted to obtain real-time data of model decisions, i.e., real-time charging decision data.

[0055] In this embodiment, charging service call data, such as charging start time, charging end time, charging power, charging cost, etc., are acquired in real time from the charging service system or charging pile management platform; these data are verified and sorted to ensure the accuracy and completeness of the data, and the real-time data of the called charging service, i.e., the real-time charging service call data, is obtained.

[0056] In this embodiment, the real-time data input by the user and the real-time data of the model decision are used as the input of the model decision to retrain the model. During the training process, an online update method is adopted, that is, as new real-time data is added, the model is continuously fine-tuned to improve the accuracy and generalization ability of the model; finally, the model is evaluated through cross-validation, accuracy, recall rate and other indicators, and the model is optimized according to the evaluation results, such as adjusting model parameters, adding features, etc., thereby realizing online update of the decision model and improving the reliability of the decision results.

[0057] In this embodiment, a mapping relationship between charging decision results and charging services can be established based on historical charging call service data and historical charging decision results, and then the mapping relationship is continuously updated and optimized using the above-mentioned real-time charging call service data. For example, by analyzing abnormal values ​​or trend changes in the charging call service data, the mapping relationship is adjusted to obtain a new mapping relationship, which can improve the accuracy of calling charging services.

[0058] The large-model-based intelligent charging decision method provided by the present invention improves the dynamic decision-making ability of the decision model by acquiring real-time text or voice recording data and real-time charging decision data to update the decision model online, and updates the mapping relationship between the charging decision results and the charging service through real-time charging call service data, thereby realizing the dynamic call of the charging service.

[0059] In some embodiments, the charging service includes a backend call service and a hardware call service; determining the charging service based on the charging decision result includes: determining the target charging type based on the charging decision result, and determining the call instruction based on the target charging method to call the corresponding charging service; different target charging types are associated with different charging services through a mapping dictionary.

[0060] In this embodiment, the target charging type includes fast charging, slow charging, or timed charging.

[0061] In this embodiment, the backend calling service includes interaction with the charging management system or cloud platform for adjusting the charging strategy, recording charging data, etc.

[0062] In this embodiment, the hardware call service includes direct communication with a charging pile or a charging station to control the charging process.

[0063] In this embodiment, a mapping dictionary is constructed to associate the target charging type with the specific charging service. The construction of the mapping dictionary is specifically implemented by the following steps: (1) List all possible charging types (such as fast charging, slow charging, timed charging, etc.).

[0064] (2) For each charging type, determine the corresponding backend call service and hardware call service; store this information in the form of key-value pairs in a mapping dictionary, where the key is the target charging type and the value is the corresponding charging service.

[0065] In this embodiment, according to the target charging type, after the corresponding charging service is found from the mapping dictionary, an instruction for calling the corresponding charging service is generated; according to the calling instruction, a request is sent to the charging management system, such as setting charging parameters, starting a charging task, etc.; or according to the calling instruction, a control signal is sent to the charging pile, such as starting charging, adjusting the charging current, etc.

[0066] In some embodiments, in order to ensure charging safety, the vehicle camera can be controlled to monitor the charging process in real time to ensure that the charging service is performed as expected. If an abnormality is found (such as charging interruption, battery overheating, etc.), measures are taken immediately (such as stopping charging, alarming, etc.), and data from the charging process (such as charging time, charging amount, charging cost, etc.) are collected to optimize the decision model and mapping dictionary. Through this process, it can be ensured that the charging service system can accurately call the corresponding back-end and hardware services according to the charging decision results, thereby achieving an efficient and safe charging process.

[0067] The large-model-based intelligent charging decision method provided by the present invention determines the target charging type through the charging decision result, determines the calling instruction according to the target charging mode, and calls the corresponding back-end calling service and hardware calling service, thereby improving the service calling efficiency in the intelligent charging process and improving the user experience.

[0068] In some embodiments, after determining the value of the target slot according to the text data, the conversation intention and the domain information, the large model-based intelligent charging decision method further includes: (1) Calculate the TF-IDF value of the text data.

[0069] Specifically, the text data corresponding to the user input is first preprocessed by word segmentation, stop word removal, etc., and then the frequency of each word in the document (Term Frequency, TF) is counted, and the inverse document frequency (Inverse Document Frequency, IDF) of each word is calculated to measure the prevalence of the word in the document set; the TF and IDF of each word are multiplied to obtain the TF-IDF value of the word (used to indicate the importance of the word in the document).

[0070] (2) Perform keyword matching on the dialogue intent, domain information, and target slot values ​​with the TF-IDF values ​​to obtain keyword matching results.

[0071] In this embodiment, keyword matching is performed between the conversation intention and the TF-IDF value, including checking whether the conversation intention is associated with certain keywords in the text data; keyword matching is performed between the domain information and the TF-IDF value, including verifying whether the text data belongs to the specified domain; keyword matching is performed between the value of the target slot and the TF-IDF value, including checking whether the value of the target slot matches certain keywords in the text data, thereby further verifying or refining the value of the target slot.

[0072] In this embodiment, the value of the target slot can be corrected, confirmed or further processed based on the keyword matching result; for example, if the matching result is good, it can be considered that the text data, conversation intent and domain information are consistent with the value of the target slot, which helps with subsequent processing or decision-making; if the matching result is not good, it may be necessary to re-analyze the text data, conversation intent and domain information, or adjust the method for determining the value of the target slot.

[0073] This embodiment can also be implemented by comparing with the information in the domain knowledge base, using the rule engine for verification, etc., and based on the verification results, the slot filling strategy is fed back and adjusted to continuously optimize the performance of the system; for example, if it is found that the filling accuracy of some slots is low, the filling accuracy can be improved by adjusting the model parameters, increasing the training data or improving the domain knowledge base.

[0074] The large-model-based intelligent charging decision method provided by the present invention can improve the ability to understand and process text data by introducing TF-IDF calculation and keyword matching, thereby more accurately determining the value of the target slot and providing more reliable information for subsequent charging decisions.

[0075] Figure 2 This is the second flow chart of the intelligent charging decision method based on the large model provided by the present invention. Figure 2 In the illustrated embodiment, the user request is first obtained, and context management is performed to obtain text information, which is then determined based on the text information using the intent model, and finally a charging decision is made based on the text information using the decision model. The decision model makes a charging decision based on the conversation intent, domain information, and intent slot results to obtain a charging decision result, and finally a charging service call instruction is generated based on the charging decision result, and the charging service information and charging decision result corresponding to the instruction are fed back to the user end through the API interface for the user to query and select.

[0076] The following is a description of the intelligent charging decision-making device based on a large model provided by the present invention. The intelligent charging decision-making device based on a large model described below and the intelligent charging decision-making method based on a large model described above can refer to each other.

[0077] Figure 3 is a schematic diagram of the structure of the intelligent charging decision-making device based on the large model provided by the present invention, such as Figure 3 As shown, the intelligent charging decision device based on the large model includes: a data acquisition module 310, an intention recognition module 320 and a charging decision module 330.

[0078] The data acquisition module 310 is used to convert the user's target input into text data; wherein the target input includes at least one of touch input and voice input; The intention recognition module 320 is used to recognize the text data based on a preset intention model, obtain the dialogue intention and domain information, and determine the value of the target slot according to the text data, the dialogue intention and the domain information; The charging decision module 330 is used to make a charging decision based on the decision model according to the conversation intention, domain information and the value of the target slot, obtain a charging decision result, and determine the charging service based on the charging decision result; wherein the decision model is based on mathematical modeling with sample conversation intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

[0079] The large-model-based intelligent charging decision device provided by the present invention recognizes text data corresponding to user input information through an intention model, and determines the value of a target slot according to the text data, conversation intention and domain information, and finally uses the decision model to make a charging decision according to the conversation intention, domain information and the value of the target slot to obtain a charging decision result, and determines the charging service according to the charging decision result, thereby improving the charging decision efficiency and charging efficiency, simplifying user operations and improving user experience.

[0080] In some embodiments, the large model-based intelligent charging decision device also includes: an intention verification module, which is used to perform TF-IDF calculation on the text data to obtain the TF-IDF value after determining the value of the target slot based on the text data, dialogue intent and domain information; and perform keyword matching on the dialogue intent, domain information and target slot value with the TF-IDF value to obtain a keyword matching result.

[0081] The large-model-based intelligent charging decision device provided by the present invention can improve the ability to understand and process text data by introducing TF-IDF calculation and keyword matching, thereby more accurately determining the value of the target slot and providing more reliable information for subsequent charging decisions.

[0082] Figure 4 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 4As shown, the electronic device may include: a processor (processor) 410 , a communication interface (Communications Interface) 420 , a memory (memory) 430 and a communication bus 440 , wherein the processor 410 , the communication interface 420 , and the memory 430 communicate with each other through the communication bus 440 . The processor 410 can call the logic instructions in the memory 430 to execute the intelligent charging decision method based on the large model, which includes: converting the user's target input into text data; wherein the target input includes at least one of touch input and voice input; identifying the text data based on a preset intention model to obtain dialogue intention and domain information, and determining the value of the target slot based on the text data, dialogue intention and domain information; making a charging decision based on the dialogue intention, domain information and the value of the target slot based on the decision model to obtain a charging decision result, and determining the charging service based on the charging decision result; wherein the decision model is based on mathematical modeling with sample dialogue intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

[0083] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform 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, etc. Various media that can store program codes.

[0084] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the large model-based intelligent charging decision method provided by the above methods, the method including: converting the user's target input into text data; wherein the target input includes at least one of touch input and voice input; identifying the text data based on a preset intention model to obtain dialogue intention and domain information, and determining the value of the target slot based on the text data, dialogue intention and domain information; making a charging decision based on the dialogue intention, domain information and the value of the target slot based on the decision model to obtain a charging decision result, and determining the charging service based on the charging decision result; wherein the decision model is based on mathematical modeling with sample dialogue intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

[0085] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the large-model-based intelligent charging decision method provided by the above-mentioned methods, the method comprising: converting the user's target input into text data; wherein the target input includes at least one of touch input and voice input; recognizing the text data based on a preset intention model to obtain dialogue intention and domain information, and determining the value of the target slot based on the text data, dialogue intention and domain information; making a charging decision based on the dialogue intention, domain information and the value of the target slot based on the decision model to obtain a charging decision result, and determining the charging service based on the charging decision result; wherein the decision model is based on mathematical modeling with sample dialogue intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

[0086] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0087] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart charging decision method based on a large model, characterized in that: include: Converting a user's target input into text data; wherein the target input includes at least one of a touch input and a voice input; Recognize the text data based on a preset intention model to obtain a dialogue intention and domain information, and determine a value of a target slot according to the text data, the dialogue intention and the domain information; Based on the decision model, a charging decision is made according to the conversation intention, the domain information and the value of the target slot to obtain a charging decision result, and a charging service is determined according to the charging decision result; wherein the decision model is based on mathematical modeling with sample conversation intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

2. The large model-based intelligent charging decision method according to claim 1, characterized in that: The decision model is obtained through the following steps: Mapping the sample conversation intention, the sample domain information, and the sample slot value to a high-dimensional space vector to obtain a mapped sample; Constructing a sample data set according to the mapped samples and the charging influencing factors, taking the sample data set as training samples, and iteratively training a long short-term memory network LSTM with numerical labels obtained by converting the sentiment scores as training labels, and obtaining the decision model when a preset number of training times is reached; The emotion score is obtained by weighted calculation based on at least two of the charging influencing factors.

3. The large model-based intelligent charging decision method according to claim 2, characterized in that: After obtaining the decision model, the method further includes: Obtain real-time text or voice recording data, real-time charging decision data, and real-time charging call service data; The decision model is updated online based on the real-time text or voice recording data and the real-time charging decision data to obtain a new decision model; and the mapping relationship between the charging decision result and the charging service is updated according to the real-time charging calling service data.

4. The intelligent charging decision method based on a large model according to claim 1, characterized in that: The charging service includes a backend calling service and a hardware calling service; Determining the charging service according to the charging decision result includes: Determine the target charging type according to the charging decision result, Determine a calling instruction according to the target charging mode to call a corresponding charging service; Different target charging types are associated with different charging services through a mapping dictionary.

5. The large model-based intelligent charging decision method according to claim 1, characterized in that: After determining the value of the target slot according to the text data, the dialogue intention and the domain information, the method further includes: Performing TF-IDF calculation on the text data to obtain a TF-IDF value; The conversation intention, the domain information and the target slot value are respectively matched with the TF-IDF value to obtain a keyword matching result.

6. An intelligent charging decision-making device based on a large model, characterized in that: include: A data acquisition module, used to convert a user's target input into text data; wherein the target input includes at least one of a touch input and a voice input; An intention recognition module, used to recognize the text data based on a preset intention model, obtain the dialogue intention and domain information, and determine the value of the target slot according to the text data, the dialogue intention and the domain information; A charging decision module is used to make a charging decision based on a decision model according to the conversation intention, the domain information and the value of the target slot, obtain a charging decision result, and determine the charging service according to the charging decision result; wherein the decision model is based on mathematical modeling with sample conversation intention, sample domain information, sample slot value and charging influence factor as input and charging emotion score as output; the charging influence factor includes at least one of the distance between the user and the charging pile, charging time, charging price, charging amount, charging pile type and user identity information.

7. The large model-based intelligent charging decision device according to claim 6, characterized in that: The device also includes: The intention verification module is used to perform TF-IDF calculation on the text data to obtain a TF-IDF value after determining the value of the target slot according to the text data, the dialogue intention and the domain information; and perform keyword matching on the dialogue intention, the domain information and the value of the target slot with the TF-IDF value respectively to obtain a keyword matching result.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the large model-based intelligent charging decision method as described in any one of claims 1 to 5 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the large model-based intelligent charging decision method as described in any one of claims 1 to 5 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the large model-based intelligent charging decision method as described in any one of claims 1 to 5 is implemented.

Images