Problem processing method, device and equipment based on artificial intelligence and storage medium

Abstract

The invention relates to the technical field of information processing, in particular to a question processing method, device and equipment based on artificial intelligence and a storage medium. Question recognition is performed on text information, a business consultation intention is determined according to a recognition result, and a corresponding business query model is called according to the business consultation intention; and inputting the identification result into a business query model for business query to obtain a query result, which is suitable for business data consultation, and when a business consultation intention cannot be determined, calling a preset artificial intelligence large model, inputting the identification result into the artificial intelligence large model for question answering to obtain an answering result, so that the intelligent degree is high, and the efficiency is high. The method is suitable for solving the logistics problem, is wide in application range, generates feedback information according to the query result or the solution result, sends the feedback information to the client, timely processes the logistics problem of the customer, reduces the workload of customer service, improves the quality and efficiency of express service, improves the user experience of the customer, and improves the customer satisfaction.

Description

Technical Field

[0001] The present invention relates to the field of information processing technology, and in particular to an artificial intelligence-based problem processing method, device, equipment and storage medium. Background Art

[0002] With the popularity of online shopping, express delivery services have become an indispensable part of people's daily lives. Whether it is daily necessities, food, or electronic products, they are inseparable from express delivery. However, with the increasing demand for express delivery services, the requirements for its service quality are also constantly improving. Today's consumers not only pay attention to the price of express delivery, but also pay more attention to the speed, safety and convenience of express delivery. In the current logistics operation process, there are a large number of standardized and standardized customer service work. The manual processing method has limited processing efficiency. The service quality of manual customer service mainly depends on the experience and work attitude of customer service personnel, which is prone to unstable service quality. Long-term high-intensity work, especially when there are omissions in the reply or it cannot be followed up in time, is easy to cause customer dissatisfaction, unable to maintain a unified service quality, and unable to guarantee the stability of the service. Therefore, customer service intelligent robots can be applied to actual business, such as logistics problem solving and business data consulting. However, intelligent robots cannot perform intelligent conversion from these two business scenarios. The degree of intelligence is low. Only one model can be used for intent recognition. The scope of application is small, and user intent and needs cannot be recognized in time. The efficiency of business problem handling is low. Summary of the invention

[0003] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide an artificial intelligence-based problem-solving method, device, equipment and storage medium that is highly intelligent, suitable for business data consulting and logistics problem solving, has a wide range of applications, can promptly handle customers' logistics problems, reduce customer service workload, improve the quality and efficiency of express services, enhance customers' user experience, and improve customer satisfaction.

[0004] The first aspect of the present invention provides an artificial intelligence-based problem handling method, comprising: obtaining real-time logistics problem information sent by a client, parsing the real-time logistics problem information to obtain text information; performing problem recognition on the text information to obtain recognition results, and determining business consulting intentions based on the recognition results; calling a corresponding business query model based on the business consulting intentions, inputting the recognition results into the business query model to perform business queries, and obtaining query results; when the business consulting intentions cannot be determined, calling a preset artificial intelligence big model, inputting the recognition results into the artificial intelligence big model to perform problem answering, and obtaining answer results; generating feedback information based on the query results or the answer results, and sending the feedback information to the client.

[0005] Optionally, in a first implementation method of the first aspect of the present invention, the real-time logistics problem information sent by the client is obtained, and the real-time logistics problem information is parsed to obtain text information, including: analyzing historical voice data to obtain analysis results, and constructing a voice semantic library according to the analysis results; obtaining the real-time logistics problem information sent by the client; determining whether the real-time logistics problem information is voice information; if so, denoising the real-time logistics problem information to obtain voice denoising information; extracting voice features of the voice denoising information, and sorting the voice features to obtain a feature sequence; parsing the feature sequence through the voice semantic library to obtain semantic parsing information, and converting the semantic parsing information into text information.

[0006] Optionally, in a second implementation method of the first aspect of the present invention, the text information is subjected to question identification to obtain identification results, and the business consulting intent is determined based on the identification results, including: segmenting the text information based on natural language processing technology to obtain multiple phrases; identifying keywords in the multiple phrases, and matching corresponding business names based on the keywords; and determining the business consulting intent based on the business names.

[0007] Optionally, in a third implementation method of the first aspect of the present invention, calling a corresponding business query model according to the business consulting intent, inputting the recognition result into the business query model to perform a business query, and obtaining a query result, includes: classifying the business consulting intent to obtain a query business category; calling a business query model that has an association with the query business category; inputting the recognition result into the business query model to perform a business query, and obtaining a query result.

[0008] Optionally, in a fourth implementation method of the first aspect of the present invention, when the business consulting intention cannot be determined, calling a preset artificial intelligence big model, inputting the recognition result into the artificial intelligence big model for question answering, and obtaining an answer result, including: obtaining a logistics knowledge question and answer corpus graph, dividing the logistics knowledge question and answer corpus graph into a training set, a verification set, and a test set; establishing an artificial intelligence big model according to the training set, the verification set, and the test set; when the business consulting intention cannot be determined, calling the artificial intelligence big model, and inputting the recognition result into the artificial intelligence big model for question answering, and outputting a preliminary reply; calculating the confidence of the preliminary reply, and judging whether the confidence meets a preset first threshold; if so, determining the preliminary reply as the answer result.

[0009] Optionally, in a fifth implementation method of the first aspect of the present invention, generating feedback information based on the query result or the answer result and sending the feedback information to the client includes: generating feedback information based on the query result or the answer result; obtaining a preset reply template, filling the feedback information into the reply template to obtain reply information; sending the reply information to the client, so that the user end generates and displays a question reply page based on the reply information.

[0010] Optionally, in a sixth implementation method of the first aspect of the present invention, after generating feedback information based on the query result or the answer result and sending the feedback information to the client, it also includes: monitoring the evaluation operation information fed back by the client; when the evaluation operation information is monitored, receiving the evaluation operation information, and obtaining a timestamp from a timestamp service platform based on the receiving time; determining whether the evaluation operation information is a negative review information; if so, recording the negative review information and the timestamp to obtain recorded information, and counting the number of times the information is recorded; when the number reaches a preset second threshold, iteratively updating the artificial intelligence large model.

[0011] The second aspect of the present invention provides an artificial intelligence-based problem handling device, including: an acquisition and analysis module, used to acquire real-time logistics problem information sent by a client, parse the real-time logistics problem information, and obtain text information; an identification and determination module, used to identify problems on the text information, obtain identification results, and determine business consulting intentions based on the identification results; a call query module, used to call a corresponding business query model based on the business consulting intention, input the identification results into the business query model to perform business query, and obtain query results; a call answer module, used to call a preset artificial intelligence big model when the business consulting intention cannot be determined, input the identification results into the artificial intelligence big model to perform problem answering, and obtain answer results; a generation and sending module, used to generate feedback information based on the query results or the answer results, and send the feedback information to the client.

[0012] Optionally, in a first implementation method of the second aspect of the present invention, the acquisition and parsing module includes: an analysis and construction unit, which is used to analyze historical voice data, obtain analysis results, and construct a voice semantic library according to the analysis results; an acquisition unit, which is used to obtain real-time logistics problem information sent by the client; a judgment unit, which is used to judge whether the real-time logistics problem information is voice information; a denoising unit, which is used to denoise the real-time logistics problem information if so, to obtain voice denoising information; an extraction and sorting unit, which is used to extract voice features of the voice denoising information, and sort the voice features to obtain a feature sequence; and a parsing and conversion unit, which is used to parse the feature sequence through the voice semantic library to obtain semantic parsing information, and convert the semantic parsing information into text information.

[0013] Optionally, in a second implementation method of the second aspect of the present invention, the identification and determination module includes: a word segmentation unit, used to segment the text information based on natural language processing technology to obtain multiple phrases; an identification and matching unit, used to identify keywords in the multiple phrases and match corresponding business names according to the keywords; a first determination unit, used to determine the business consulting intention based on the business name.

[0014] Optionally, in a third implementation of the second aspect of the present invention, the calling query module includes: a classification unit, used to classify the business consulting intention to obtain a query business category; a calling unit, used to call a business query model that has an association with the query business category; and a query unit, used to input the identification result into the business query model to perform a business query and obtain a query result.

[0015] Optionally, in a fourth implementation method of the second aspect of the present invention, the calling answer module includes: an acquisition division unit, used to acquire a logistics knowledge question and answer corpus graph, and divide the logistics knowledge question and answer corpus graph into a training set, a verification set and a test set; an establishment unit, used to establish an artificial intelligence big model according to the training set, the verification set and the test set; a calling answer unit, used to call the artificial intelligence big model when the business consulting intention cannot be determined, and input the recognition result into the artificial intelligence big model for question answering, and output a preliminary reply; a calculation and judgment unit, used to calculate the confidence of the preliminary reply, and judge whether the confidence meets a preset first threshold; a second determination unit, used to, if so, determine the preliminary reply as the answer result.

[0016] Optionally, in a fifth implementation of the second aspect of the present invention, the generation and sending module includes: a generation unit, used to generate feedback information based on the query result or the answer result; an acquisition and filling unit, used to obtain a preset reply template, fill the feedback information into the reply template, and obtain reply information; a sending unit, used to send the reply information to the client, so that the user end generates and displays a question reply page based on the reply information.

[0017] Optionally, in a sixth implementation method of the second aspect of the present invention, it also includes: a monitoring module, which is used to monitor the evaluation operation information fed back by the client; a receiving and obtaining module, which is used to receive the evaluation operation information when the evaluation operation information is monitored, and obtain a timestamp from a timestamp service platform based on the receiving time; a judgment module, which is used to judge whether the evaluation operation information is negative review information; a recording and statistics module, which is used to record the negative review information and the timestamp if so, obtain recorded information, and count the number of times the information is recorded; an updating module, which is used to iteratively update the artificial intelligence large model when the number reaches a preset second threshold.

[0018] The third aspect of the present invention provides an artificial intelligence-based problem-solving device, which includes: a memory and at least one processor, wherein the memory stores instructions; at least one of the processors calls the instructions in the memory to enable the artificial intelligence-based problem-solving device to perform each step of any of the above-mentioned artificial intelligence-based problem-solving methods.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored thereon, which, when executed by a processor, implement the various steps of any of the above-mentioned artificial intelligence-based problem-solving methods.

[0020] In the technical solution of the present invention, problem identification is performed on text information, the business consulting intention is determined according to the identification result, the corresponding business query model is called according to the business consulting intention, and the identification result is input into the business query model for business query to obtain the query result, which is suitable for business data consulting. When the business consulting intention cannot be determined, the preset artificial intelligence big model is called, and the recognition result is input into the artificial intelligence big model for problem answering to obtain the answer result. The invention has a high degree of intelligence and is suitable for answering logistics problems. It has a wide range of applications. Feedback information is generated according to the query result or the answer result, and the feedback information is sent to the client to promptly handle the customer's logistics problems, reduce the customer service workload, improve the quality and efficiency of express services, improve the customer's user experience, and improve customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A first flow chart of a problem-solving method based on artificial intelligence provided by an embodiment of the present invention;

[0022] Figure 2 A second flow chart of the problem-solving method based on artificial intelligence provided by an embodiment of the present invention;

[0023] Figure 3 A third flow chart of the problem-solving method based on artificial intelligence provided by an embodiment of the present invention;

[0024] Figure 4 A fourth flow chart of the problem-solving method based on artificial intelligence provided by an embodiment of the present invention;

[0025] Figure 5 A schematic diagram of a structure of an artificial intelligence-based problem-solving device provided in an embodiment of the present invention;

[0026] Figure 6 Another structural schematic diagram of the problem processing device based on artificial intelligence provided by an embodiment of the present invention;

[0027] Figure 7 A schematic diagram of the structure of an artificial intelligence-based problem-solving device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The present invention provides an artificial intelligence-based problem-solving method, device, equipment and storage medium, which have a high degree of intelligence and are suitable for business data consulting and logistics problem solving. They have a wide range of applications and can promptly handle customers' logistics problems, reduce customer service workload, improve the quality and efficiency of express delivery services, enhance customers' user experience, and improve customer satisfaction.

[0029] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0030] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1, an embodiment of the problem processing method based on artificial intelligence in the embodiment of the present invention includes:

[0031] 101. Obtain real-time logistics problem information sent by the client, parse the real-time logistics problem information, and obtain text information;

[0032] In this embodiment, by establishing a communication channel with the client, the system can receive various logistics issues raised by users in real time, providing a data source for subsequent processing. After obtaining the real-time logistics problem information sent by the user, the system needs to perform parsing and convert the text information into a format that can be understood by the computer. By parsing the real-time logistics problem information, the system finally obtains the form of text information, which contains the specific questions and related content raised by the user.

[0033] 102. Identify the problem in the text information, obtain the identification result, and determine the business consultation intention according to the identification result;

[0034] In this embodiment, the system needs to use natural language processing technology to analyze and process text information in order to identify the specific content of the question raised by the user. By identifying the question in the text information, the system will obtain an identification result, which indicates the type, topic or key content of the question raised by the user. Based on the previous identification results, the system will determine the specific business consulting intent to which the user's question belongs according to a pre-set business consulting intention classification system.

[0035] 103. Call the corresponding business query model according to the business consultation intention, input the recognition result into the business query model to perform business query, and obtain the query result;

[0036] In this embodiment, the corresponding business query model will be selected to process the questions raised by the user based on the previously determined business consulting intention. Each business consulting intention corresponds to a specific business query model. These models are pre-trained machine learning models, rule-based query engines or other forms of information retrieval systems. The system uses the identified user question text as input and passes it to the selected business query model for specific query operations. After calling the business query model for query operations, the system will obtain the query results.

[0037] 104. When the intention of business consultation cannot be determined, the preset artificial intelligence big model is called, and the recognition result is input into the artificial intelligence big model to answer the question and obtain the answer;

[0038] In this embodiment, the system will enter this branch when the system cannot accurately determine the business consultation intention raised by the user. This happens when the user raises a question that the system cannot recognize, or the question expressed by the user involves multiple possible business areas, which makes the system unable to clearly classify it. In the case where the business consultation intention cannot be determined, the system will call the preset artificial intelligence big model. The system uses the question text raised by the user as input and passes it to the preset artificial intelligence big model to understand and answer the question. This type of model can handle various types of natural language understanding and generation tasks, thereby providing users with more intelligent and flexible answers. After calling the artificial intelligence big model to answer the question, the system will obtain the answer result.

[0039] 105. Generate feedback information according to the query result or the answer result, and send the feedback information to the client;

[0040] In this embodiment, the system will generate corresponding feedback information based on the results of previous query operations or question answering, which involves converting the query results into easy-to-understand text form, or organizing the answer results into user-friendly reply content. The generated feedback information will be sent to the client, which is the interface for the user to interact with the system. This client can be a web interface, mobile application, chat robot, etc. On this interface, the user can see the system's replies and information feedback. Once the system sends the feedback information to the client, the user can receive this information on his or her end.

[0041] In an embodiment of the present invention, problem identification is performed on text information, business consulting intent is determined based on the identification result, and the corresponding business query model is called according to the business consulting intent. The identification result is input into the business query model to perform business query and obtain the query result, which is suitable for business data consulting. When the business consulting intent cannot be determined, the preset artificial intelligence big model is called, and the identification result is input into the artificial intelligence big model to answer the question and obtain the answer result. The method has a high degree of intelligence and is suitable for answering logistics problems. It has a wide range of applications. Feedback information is generated based on the query result or the answer result, and the feedback information is sent to the client, so as to promptly handle the customer's logistics problems, reduce the customer service workload, improve the quality and efficiency of express delivery services, improve the customer's user experience, and improve customer satisfaction.

[0042] See also Figure 2 , a second embodiment of the problem processing method based on artificial intelligence in the embodiment of the present invention includes:

[0043] 201. Analyze the historical speech data to obtain analysis results, and build a speech semantic library according to the analysis results;

[0044] In this embodiment, the system will collect and process historical voice data, and will perform analysis operations such as transcription, voice recognition, and voice feature extraction on the historical voice data. After analyzing the historical voice data, the system will obtain a series of analysis results, including voice-to-text transcription, speaker recognition, voice sentiment analysis, keyword extraction and other information, which will help the system to better understand the voice content and background information. Based on the analysis results obtained previously, the system will build a voice semantic library. This semantic library contains the mapping relationship between voice content and semantics, the correspondence between voice commands and operations, and other related voice data and semantic information. By building such a voice semantic library, the system can more accurately understand and respond to user voice input.

[0045] 202. Obtain real-time logistics problem information sent by the client;

[0046] In this embodiment, real-time logistics problem information is sent to the system via a mobile application, a web page, or a client via other channels.

[0047] 203. Determine whether the real-time logistics problem information is voice information;

[0048] In this embodiment, the real-time logistics problem information may be transmitted in the form of text or voice, and when the real-time logistics problem information is received, it is determined whether the real-time logistics problem information is voice information.

[0049] 204. If yes, denoising is performed on the real-time logistics problem information to obtain voice denoising information;

[0050] In this embodiment, when the real-time logistics problem information is voice information, the real-time logistics problem information is denoised, specifically eliminating the noise in the voice signal to improve the voice quality and accuracy. This can include using digital signal processing technology, such as filtering, noise reduction algorithms, etc., to reduce background noise, extract effective signals, etc. After denoising, the system will obtain processed voice denoising information.

[0051] 205. Extracting speech features of speech denoising information, and sorting the speech features to obtain a feature sequence;

[0052] In this embodiment, speech features are extracted from speech denoising information through methods such as Mel-frequency cepstral coefficients (MFCC), linear predictive coding (LPC), short-time energy, and short-time zero-crossing rate, and the speech signal is converted into a recognizable feature vector. Next, the system can sort the extracted speech features to determine which features are most critical for speech recognition and understanding. Finally, the system can obtain the feature sequence.

[0053] 206. Parse the feature sequence through the speech semantic library to obtain semantic parsing information, and convert the semantic parsing information into text information;

[0054] In this embodiment, the feature sequence is parsed using a pre-constructed speech semantic library, which contains a large number of mapping relationships between speech features and corresponding semantic information. The system performs semantic analysis based on the speech features extracted from the feature sequence by matching the corresponding information in the semantic library. By performing semantic library parsing on the feature sequence, the system can obtain corresponding semantic analysis information, which covers semantic content such as user intentions, operating instructions, and question content. The semantic analysis information helps the system understand the speech information input by the user and perform subsequent processing. Next, the system needs to convert the acquired semantic analysis information into text information.

[0055] In the embodiment of the present invention, by analyzing historical voice data, building a voice semantic library, and performing voice analysis and semantic conversion on real-time logistics problem information, the system can understand the user's needs more accurately, improve the user's interactive experience with the system, and build a voice semantic library based on the analysis results of historical data, so that the system has certain intelligent recognition and processing capabilities, can respond to user problems more quickly, and improve the intelligence level of the system. It is determined whether the real-time logistics problem information is voice information, and denoising, feature extraction and analysis are performed, which helps to improve the stability and accuracy of the system, reduce the error rate, and ensure the normal operation of the system. Through the sorting of feature sequences and the conversion of semantic analysis information, the system can quickly and accurately handle user problems, speed up problem solving, and improve work efficiency.

[0056] See also Figure 3 , a third embodiment of the problem processing method based on artificial intelligence in the embodiment of the present invention includes:

[0057] 301. Segment the text information based on natural language processing technology to obtain multiple phrases;

[0058] In this embodiment, the system will receive text information that needs to be processed, which may be text information converted from voice or real-time logistics problem information sent by the client. The word segmentation module in the natural language processing technology is used to perform word segmentation on the input text information. Through word segmentation, the system will obtain multiple phrases in the text information. Each phrase represents a word or phrase with an independent meaning. These phrases can be single words or phrases or terms composed of multiple words. The multiple phrases obtained after word segmentation can be used for subsequent text processing tasks.

[0059] 302. Identify keywords in multiple phrases, and match corresponding business names according to the keywords;

[0060] In this embodiment, multiple phrases obtained by previous word segmentation are processed to identify keywords therein. Keywords generally refer to words or phrases that are of great significance or represent the core of a topic in a specific context. Various natural language processing technologies, such as keyword extraction algorithms, TF-IDF (term frequency-inverse document frequency) algorithms, text summarization algorithms, etc., are used to identify keywords in multiple phrases. These algorithms can determine keywords based on factors such as the importance and frequency of the words in the text. Once the keywords are identified, the system can match them based on these keywords to find the corresponding business names, which involves a predefined list or database of business names. The system will compare the keywords with these business names to find the best matching business name. Different strategies can be used to match the correspondence between keywords and business names, such as similarity calculation, semantic relevance, rule matching, etc.

[0061] 303. Determine the business consultation intention based on the business name;

[0062] In this embodiment, by identifying keywords in multiple phrases and matching corresponding business names, the specific business field or topic that the user wants to consult is determined. Now the system needs to further analyze based on the determined business name to determine the user's specific business consultation intention. The business consultation intention may include the user wanting to know the specific content of a certain service, seek help to solve a problem, consult price information, etc.

[0063] 304. Classify the business consultation intention to obtain the query business category;

[0064] In this embodiment, the user's business consultation intention is input into a preset intention classification model for prediction. The model determines which specific business category the user's intention belongs to based on the learned features and patterns, and obtains the query business category.

[0065] 305. Calling a business query model associated with the query business category;

[0066] In this embodiment, for each business category, the system needs to prepare a corresponding business query model, which includes a freight query model and an express query model, etc., depending on specific business needs.

[0067] 306. Input the recognition result into the business query model to perform business query and obtain the query result;

[0068] In this embodiment, once the system determines the user's query business category, it will pass this recognition result as input to the corresponding business query model. The system will select an appropriate business query model to process the user's query based on the recognition result. For example, if the user's query belongs to the freight category, the system can call the freight query model. If the query belongs to the express status category, the system can call the corresponding express query model. The system will pass the user's query and recognition result as input to the business query model. The business query model will process the user's query and output the corresponding query result based on its internal algorithm and logic.

[0069] 307. Obtain a logistics knowledge question and answer corpus graph, and divide the logistics knowledge question and answer corpus graph into a training set, a validation set, and a test set;

[0070] In this embodiment, question and answer data in the field of logistics are collected. These data may include common logistics questions and corresponding answers. Data sources may include public data sets, documents on the Internet, professional knowledge bases, etc. These data should cover various aspects of the logistics industry, such as transportation, warehousing, supply chain management, etc. The acquired question and answer data need to be cleaned and organized, including removing duplicate data, normalizing data formats, processing missing values, etc. At the same time, the question and answer pairs need to be labeled for subsequent model training and evaluation. After preparing the cleaned and organized question and answer data, the data needs to be divided into a training set, a validation set, and a test set. Usually, it can be divided according to a ratio, for example, 70% of the data is used as a training set, 15% as a validation set, and 15% as a test set. In order to avoid uneven data distribution, random processing can be performed when dividing the data set to ensure that the samples in each data set represent the overall distribution of questions and answers in the logistics field. The divided training set, validation set, and test set are saved as independent data files for subsequent model training and evaluation.

[0071] 308. Establish an artificial intelligence model based on the training set, validation set and test set;

[0072] In this embodiment, according to the task requirements and data characteristics, a suitable artificial intelligence model framework is selected, such as Transformer, BERT, etc. These model frameworks can be used to build a powerful natural language processing model suitable for processing question-answering tasks, and the architecture of the model is designed, including the number of layers of the neural network, the number of hidden units, the attention mechanism, etc. The depth and width of the model are determined according to the complexity of the problem and the amount of data. The model is trained using the training set data, and the parameters of the model are updated through the back propagation algorithm and the optimizer so that it gradually converges to the optimal solution. During the training process, the performance of the model on the validation set can be monitored so that the model structure and hyperparameters can be adjusted in time, and the model is tuned on the validation set, including Adjust hyperparameters such as learning rate, regularization term, batch size, etc. to improve the generalization ability and performance of the model. Use the test set data to evaluate the trained model and calculate the performance indicators of the model on the test set, such as accuracy, recall rate, F1 value, etc. The evaluation results can help judge the actual effect and generalization ability of the model. Deploy the trained artificial intelligence large model to the production environment for practical application. It can be deployed through API interface, Web service, etc., so that it can accept user input and output corresponding question and answer results. Monitor and feedback the deployed model to continuously optimize the performance and effect of the model. The model can be retrained regularly and new data can be introduced for updating to adapt to the ever-changing needs and environment.

[0073] 309. When the intention of the business consultation cannot be determined, the artificial intelligence big model is called, and the recognition results are input into the artificial intelligence big model to answer the questions and output a preliminary response;

[0074] In this embodiment, when the system receives a user's consultation, it first needs to try to identify the user's consultation intention through certain rules or simple classifiers. If the user's specific consultation intention cannot be determined or belongs to an unknown field, it is necessary to enter this step. When the business consultation intention cannot be determined, the system can call a pre-trained artificial intelligence big model and use the user's question whose intention cannot be determined as input and pass it to the artificial intelligence big model for processing. The model will infer and analyze the content of the input question, and then generate a corresponding preliminary reply.

[0075] 310. Calculate the confidence of the preliminary reply, and determine whether the confidence meets a preset first threshold;

[0076] In this embodiment, for the preliminary reply, some evaluation indicators are designed or the information inside the model is used to calculate the confidence of the preliminary reply. These indicators may include the model's predicted probability of the reply, the degree of match between the reply and the question, the degree of coverage of key information in the reply, etc. A preset first threshold is set in the system. This first threshold can be set based on actual needs and business scenarios. Generally speaking, the selection of the first threshold needs to take into account the balance between accuracy and recall rate to ensure better performance at a certain confidence level. The system compares the calculated confidence of the preliminary reply with the preset first threshold. If the confidence of the preliminary reply is higher than or equal to the preset first threshold, the system can consider that the preliminary reply meets the requirements.

[0077] 311. If so, the preliminary reply shall be determined as the answer result;

[0078] In this embodiment, if the confidence level of the preliminary answer reaches a preset first threshold, the preliminary answer is determined as the answer result.

[0079] In the embodiment of the present invention, the text information is segmented and keywords are identified through natural language processing technology, and the consultation intention is determined according to the business name, so that the questions raised by the user can be understood more accurately, thereby improving the accuracy of the system's understanding of the user's intention. The consultation intention is determined according to the business name and the business category is queried by classification. The corresponding business query model is called for query, which can realize intelligent business processing and improve the processing efficiency and accuracy of the system. The logistics knowledge question and answer corpus is obtained, and an artificial intelligence large model is established based on the training set, the verification set and the test set, which is helpful for the system to build a complete knowledge base, and the knowledge base can be continuously updated and optimized according to actual needs, and the intelligence level of the system is improved. When the business consultation intention cannot be determined, the artificial intelligence large model is called to answer the question, output a preliminary reply, and judge whether the reply meets the preset first threshold according to the confidence level, which can intelligently answer the user's question and improve the system's automatic reply capability. By accurately understanding the user's intention, intelligent business processing and intelligent answering of questions, the system can provide users with the required information more quickly and accurately, improve user satisfaction, and enhance users' trust in the system.

[0080] See also Figure 4 The fourth embodiment of the problem-solving method based on artificial intelligence in the embodiment of the present invention includes:

[0081] 401. Generate feedback information according to the query result or answer result;

[0082] In this embodiment, based on the content and analysis of the query results or answer results, the system can generate corresponding feedback information.

[0083] 402. Obtain a preset reply template, fill the feedback information into the reply template, and obtain reply information;

[0084] In this embodiment, a preset reply template is obtained, and the key content of the feedback information is filled into the reply template. The key content includes the questions raised by the user, the answers provided by the system, the summary of the query results, etc. The filled reply template is integrated and formatted, including operations such as variable replacement, grammar correction, and sentence structure adjustment of the template to generate the final reply information.

[0085] 403. Send the reply information to the client, so that the user terminal generates and displays a question reply page based on the reply information;

[0086] In this embodiment, a communication channel is established with the client, and the prepared reply information is sent to the client through the established communication channel. This can package the information into a specific format (such as JSON format) for transmission, or directly send a text message, etc. After the client receives the reply information sent by the system, it needs to perform corresponding processing, including parsing the information, generating the structure and layout of the question reply page according to the information content, etc. Based on the received reply information, the client will use relevant technologies and interface designs to generate the content of the question reply page, including presenting the reply text, possible pictures, links, etc., and display it to the user. The client displays the generated question reply page to the user, and the user can view the reply information generated by the system and conduct further interaction, such as asking more questions, requesting detailed explanations, and evaluating the replies.

[0087] 404. Monitor the evaluation operation information fed back by the client;

[0088] In this embodiment, a corresponding monitoring function is set to receive evaluation operation information from the client in real time and provide the client with a corresponding evaluation operation interface, such as "good review", "bad review" and other operations. When the user clicks on the relevant evaluation operation interface in the question reply page generated by the system, the client will send the corresponding operation information to the system. The system needs to monitor the evaluation operation information sent by the client and perform corresponding processing. According to the received evaluation operation information, the quality score of the reply is updated. For example, if the user clicks the "good review" button, the system can increase the quality score of the reply based on this operation, otherwise it will be reduced.

[0089] 405. When the evaluation operation information is monitored, the evaluation operation information is received, and a timestamp is obtained from the timestamp service platform based on the receiving time;

[0090] In this embodiment, when the system monitors evaluation operation information, it needs to promptly receive and parse the information, including the evaluation type (good review, bad review, etc.) and related data identifiers. While receiving the evaluation operation information, the system needs to record the time of receiving the information. This time will be used as the basis for subsequent queries to the timestamp service platform. The system calls the interface provided by the timestamp service platform according to the time of the received evaluation operation information and requests it to generate a timestamp. After receiving the system's request, the timestamp service platform will generate a corresponding timestamp based on the time data in the request and return it to the system.

[0091] 406. Determine whether the evaluation operation information is negative evaluation information;

[0092] In this embodiment, the criteria for negative reviews are defined in advance, which include indicators such as the user's dissatisfaction with the reply, the quality of the question, the accuracy of the reply, etc., and evaluation operation information is received from the client, which generally includes the user's evaluation type of the reply (good review, bad review, etc.) and related data identifiers. According to the pre-defined negative review criteria, the received evaluation operation information is evaluated, and based on the evaluation results, the system can determine whether the evaluation operation information is negative review information.

[0093] 407. If yes, record the negative review information and timestamp, obtain record information, and count the number of times the record information is obtained;

[0094] In this embodiment, if the evaluation operation information is judged to be a negative review, the system needs to record the relevant information, including the evaluation content, timestamp, evaluator, etc. These records can be stored in the database for subsequent statistics and analysis. By querying the record information in the database, the statistical function (such as COUNT) is used to calculate the number of negative review information. This can be for a specific time period, a specific user, a specific problem, etc. After the statistics are completed, the system can analyze the number of recorded information, such as generating reports, charts, calculating the negative review rate, etc., in order to gain a deeper understanding of the negative review situation.

[0095] 408. When the number of times reaches a preset second threshold, the artificial intelligence model is iteratively updated;

[0096] In this embodiment, a second threshold is set in advance to determine whether the number of times the information is recorded has reached the condition for triggering an iterative update. The number of times the information is recorded is monitored periodically or in real time so that subsequent iterative update operations are triggered when the number reaches the preset second threshold. When the number of times the recorded information is monitored to have reached the preset second threshold, the system will make a judgment to confirm whether the condition for triggering an iterative update is met. If the number of times the information is recorded has reached the preset second threshold, the system will trigger an iterative update operation, i.e., update the artificial intelligence big model, which may include retraining the model, adding new data sets, adjusting parameters, etc. After the update is completed, the system needs to be tested and verified to ensure that the updated model has improved quality and performance and will not introduce new problems. After verification, the system can deploy the updated artificial intelligence big model to the production environment.

[0097] In an embodiment of the present invention, by generating feedback information according to the query results or the answer results and sending the reply information to the client, the system can realize real-time feedback, respond to the user's questions and needs in time, improve the user experience, obtain a preset reply template, fill it in according to the feedback information, generate reply information, and realize personalized replies to user questions, increase the quality of user interaction with the system, monitor the evaluation operation information fed back by the client, determine whether the evaluation is negative review information, and record the negative review information and timestamp, count the number of times, and iteratively update the artificial intelligence big model when the number reaches a preset threshold, which is helpful to monitor user feedback in real time, improve the system in time, and improve system performance and user satisfaction. By recording and counting the negative review information, the iterative update of the artificial intelligence big model is triggered, and the system can continuously optimize and improve its own intelligence level and service quality, so that the system can better adapt to user needs and provide more accurate replies. The real-time feedback, personalized replies, and continuous optimization of models work together to improve user experience and increase user trust in the system, thereby improving user satisfaction and loyalty.

[0098] The problem-solving method based on artificial intelligence in the embodiment of the present invention is described above. The problem-solving device based on artificial intelligence in the embodiment of the present invention is described below. Figure 5 , an embodiment of the problem processing device based on artificial intelligence in the embodiment of the present invention includes:

[0099] The acquisition and analysis module 501 is used to acquire the real-time logistics problem information sent by the client, and analyze the real-time logistics problem information to obtain text information;

[0100] The identification and determination module 502 is used to identify the problem of the text information, obtain the identification result, and determine the business consultation intention according to the identification result;

[0101] The query calling module 503 is used to call the corresponding business query model according to the business consultation intention, input the recognition result into the business query model to perform business query, and obtain the query result;

[0102] The calling answering module 504 is used to call the preset artificial intelligence big model when the business consultation intention cannot be determined, input the recognition result into the artificial intelligence big model to answer the question, and obtain the answer result;

[0103] The generating and sending module 505 is used to generate feedback information according to the query result or the answer result, and send the feedback information to the client.

[0104] In this embodiment, problem identification is performed on text information, the business consultation intention is determined according to the identification result, the corresponding business query model is called according to the business consultation intention, and the identification result is input into the business query model for business query to obtain the query result, which is suitable for business data consultation. When the business consultation intention cannot be determined, the preset artificial intelligence big model is called, and the identification result is input into the artificial intelligence big model for problem answering to obtain the answer result. It has a high degree of intelligence and is suitable for answering logistics problems. It has a wide range of applications. Feedback information is generated according to the query result or the answer result, and the feedback information is sent to the client, so as to promptly handle the customer's logistics problems, reduce the workload of customer service, improve the quality and efficiency of express delivery services, improve the customer's user experience, and improve customer satisfaction.

[0105] See also Figure 6 Another embodiment of the problem processing device based on artificial intelligence in the embodiment of the present invention includes:

[0106] The acquisition and analysis module 501 is used to acquire the real-time logistics problem information sent by the client, and analyze the real-time logistics problem information to obtain text information;

[0107] The identification and determination module 502 is used to identify the problem of the text information, obtain the identification result, and determine the business consultation intention according to the identification result;

[0108] The query calling module 503 is used to call the corresponding business query model according to the business consultation intention, input the recognition result into the business query model to perform business query, and obtain the query result;

[0109] The calling answering module 504 is used to call the preset artificial intelligence big model when the business consultation intention cannot be determined, input the recognition result into the artificial intelligence big model to answer the question, and obtain the answer result;

[0110] A generating and sending module 505 is used to generate feedback information according to the query result or the answer result, and send the feedback information to the client;

[0111] In this embodiment, the acquisition and analysis module 501 includes: an analysis and construction unit 5011, which is used to analyze the historical voice data, obtain the analysis results, and construct a voice semantic library according to the analysis results; an acquisition unit 5012, which is used to obtain the real-time logistics problem information sent by the client; a judgment unit 5013, which is used to judge whether the real-time logistics problem information is voice information; a denoising unit 5014, which is used to denoise the real-time logistics problem information if so, to obtain voice denoising information; an extraction and sorting unit 5015, which is used to extract the voice features of the voice denoising information, and sort the voice features to obtain a feature sequence; a parsing and conversion unit 5016, which is used to parse the feature sequence through the voice semantic library, obtain semantic parsing information, and convert the semantic parsing information into text information.

[0112] In this embodiment, the identification and determination module 502 includes: a word segmentation unit 5021, which is used to segment text information based on natural language processing technology to obtain multiple phrases; an identification and matching unit 5022, which is used to identify keywords in multiple phrases and match corresponding business names according to the keywords; a first determination unit 5023, which is used to determine the business consultation intention according to the business name.

[0113] In this embodiment, the calling query module 503 includes: a classification unit 5031, which is used to classify the business consultation intention and obtain the query business category; a calling unit 5032, which is used to call a business query model that has an association relationship with the query business category; a query unit 5033, which is used to input the recognition result into the business query model to perform business query and obtain the query result.

[0114] In this embodiment, calling the answer module 504 includes: an acquisition division unit 5041, which is used to obtain a logistics knowledge question and answer corpus graph, and divide the logistics knowledge question and answer corpus graph into a training set, a verification set and a test set; an establishment unit 5042, which is used to establish an artificial intelligence big model according to the training set, the verification set and the test set; a calling answer unit 5043, which is used to call the artificial intelligence big model when the business consultation intention cannot be determined, and input the recognition result into the artificial intelligence big model for question answering, and output a preliminary reply; a calculation and judgment unit 5044, which is used to calculate the confidence of the preliminary reply, and judge whether the confidence meets a preset first threshold; a second determination unit 5045, which is used to determine the preliminary reply as the answer result if so.

[0115] In this embodiment, the generation and sending module 505 includes: a generation unit 5051, which is used to generate feedback information based on the query results or the answer results; an acquisition and filling unit 5052, which is used to obtain a preset reply template, fill the feedback information into the reply template, and obtain reply information; a sending unit 5053, which is used to send the reply information to the client, so that the user end generates and displays a question reply page based on the reply information.

[0116] In this embodiment, it also includes: a monitoring module 506, which is used to monitor the evaluation operation information fed back by the client; a receiving and obtaining module 507, which is used to receive the evaluation operation information when the evaluation operation information is monitored, and obtain a timestamp from the timestamp service platform based on the receiving time; a judgment module 508, which is used to judge whether the evaluation operation information is bad review information; a recording and statistics module 509, which is used to record the bad review information and the timestamp if so, obtain the record information, and count the number of times the information is recorded; an updating module 510, which is used to iteratively update the artificial intelligence large model when the number reaches a preset second threshold.

[0117] above Figure 5 and Figure 6 The problem-solving apparatus based on artificial intelligence in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The problem-solving apparatus based on artificial intelligence in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0118] Figure 7 6 is a schematic diagram of the structure of an artificial intelligence-based problem processing device provided by an embodiment of the present invention. The artificial intelligence-based problem processing device 600 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 610 (for example, one or more processors) and a memory 620, and one or more storage media 630 (for example, one or more mass storage devices) storing application programs 633 or data 632. Among them, the memory 620 and the storage medium 630 can be short-term storage or permanent storage. The program stored in the storage medium 630 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the artificial intelligence-based problem processing device 600. Furthermore, the processor 610 can be configured to communicate with the storage medium 630, and execute a series of instruction operations in the storage medium 630 on the artificial intelligence-based problem processing device 600 to implement the steps of the artificial intelligence-based problem processing method provided by the above-mentioned method embodiments.

[0119] The artificial intelligence-based problem handling device 600 may also include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input and output interfaces 660, and / or one or more operating systems 631, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 7The structure of the artificial intelligence-based problem-solving device shown does not constitute a limitation on the artificial intelligence-based problem-solving device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0120] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the problem-solving method based on artificial intelligence.

[0121] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0122] If the integrated unit 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 is essentially or the part that contributes to the prior art or all or 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, including a number of instructions to enable 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), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0123] Finally, it should be noted that the above description is only a preferred example of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A problem-solving method based on artificial intelligence, characterized in that: include: Acquire real-time logistics problem information sent by the client, parse the real-time logistics problem information, and obtain text information; Performing problem identification on the text information to obtain an identification result, and determining the business consultation intention according to the identification result; Calling a corresponding business query model according to the business consultation intention, inputting the recognition result into the business query model to perform business query, and obtaining a query result; When the business consultation intention cannot be determined, a preset artificial intelligence big model is called, and the recognition result is input into the artificial intelligence big model to answer the question and obtain the answer result; Feedback information is generated according to the query result or the answer result, and the feedback information is sent to the client.

2. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: The acquiring of the real-time logistics problem information sent by the client, parsing the real-time logistics problem information, and obtaining text information includes: Analyze historical speech data to obtain analysis results, and build a speech semantic library based on the analysis results; Get real-time logistics problem information sent by the client; Determining whether the real-time logistics problem information is voice information; If yes, denoising is performed on the real-time logistics problem information to obtain speech denoising information; Extracting speech features of the speech denoising information, and sorting the speech features to obtain a feature sequence; The feature sequence is parsed through the speech semantic library to obtain semantic parsing information, and the semantic parsing information is converted into text information.

3. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: The step of performing problem identification on the text information to obtain an identification result, and determining the business consultation intention according to the identification result includes: Segmenting the text information based on natural language processing technology to obtain multiple phrases; Identify keywords in the plurality of phrases, and match corresponding business names according to the keywords; Determine the business consultation intention based on the business name.

4. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: The calling of the corresponding business query model according to the business consultation intention, inputting the recognition result into the business query model to perform business query, and obtaining the query result includes: Classifying the business consultation intention to obtain a query business category; Invoking a business query model associated with the query business category; The recognition result is input into the business query model to perform business query and obtain the query result.

5. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: When the business consultation intention cannot be determined, the preset artificial intelligence big model is called, and the recognition result is input into the artificial intelligence big model to answer the question, and the answer result is obtained, including: Obtaining a logistics knowledge question and answer corpus graph, and dividing the logistics knowledge question and answer corpus graph into a training set, a validation set, and a test set; Establishing an artificial intelligence large model according to the training set, the validation set and the test set; When the business consultation intention cannot be determined, the artificial intelligence big model is called, and the recognition result is input into the artificial intelligence big model to answer the question and output a preliminary answer; Calculating the confidence of the preliminary reply, and determining whether the confidence satisfies a preset first threshold; If so, the preliminary answer is determined as the answer result.

6. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: The step of generating feedback information according to the query result or the answer result and sending the feedback information to the client comprises: generating feedback information according to the query result or the answer result; Obtaining a preset reply template, filling the feedback information into the reply template, and obtaining reply information; The reply information is sent to the client, so that the user end generates and displays a question reply page based on the reply information.

7. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: After generating feedback information according to the query result or the answer result and sending the feedback information to the client, the method further includes: Monitoring the evaluation operation information fed back by the client; When the evaluation operation information is monitored, the evaluation operation information is received, and a timestamp is obtained from a timestamp service platform based on the receiving time; Determining whether the evaluation operation information is negative evaluation information; If yes, the negative review information and the timestamp are recorded to obtain record information, and the number of times the record information is recorded is counted; When the number reaches a preset second threshold, the artificial intelligence large model is iteratively updated.

8. A problem-solving device based on artificial intelligence, characterized in that: include: An acquisition and analysis module is used to acquire real-time logistics problem information sent by the client, and analyze the real-time logistics problem information to obtain text information; An identification and determination module is used to identify the problem of the text information, obtain an identification result, and determine the business consultation intention according to the identification result; A calling query module is used to call a corresponding business query model according to the business consultation intention, input the recognition result into the business query model to perform business query, and obtain a query result; A calling answering module is used to call a preset artificial intelligence big model when the business consultation intention cannot be determined, input the recognition result into the artificial intelligence big model to answer the question, and obtain the answer result; A generating and sending module is used to generate feedback information according to the query result or the answer result, and send the feedback information to the client.

9. An artificial intelligence-based problem-solving device, characterized in that: The artificial intelligence-based problem processing device comprises: a memory and at least one processor, wherein the memory stores instructions; At least one of the processors calls the instructions in the memory to enable the artificial intelligence-based problem-solving device to perform each step of the artificial intelligence-based problem-solving method as described in any one of claims 1-7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the problem-solving method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

Images