Question intention recognition method, device and equipment under background of smart community

Abstract

The invention provides a question intention recognition method, device and equipment under a smart community background. According to the method, the BERT deep learning technology is introduced, a high-quality data set is constructed, and the structure and parameters of the model are optimized, so that the accuracy and efficiency of the social worker question-answering system in the aspect of intention recognition are remarkably improved. Therefore, community service quality and efficiency can be improved, innovative development of intelligent community construction is promoted, and happiness and satisfaction of community residents can be enhanced. Furthermore, through continuous innovation and application of the technology, a new technical path and a new practical scheme are provided for construction of the smart community.

Description

Technical Field

[0001] The present application relates to a method, device and equipment for identifying question intentions in the context of a smart community, and belongs to the technical field of big data and analysis engines. Background Art

[0002] In recent years, with the in-depth promotion of smart community construction, the application of digital technology in community governance has become more and more extensive. The construction of smart communities has not only improved management efficiency, but also improved the quality of life of residents. However, the construction of smart communities faces many challenges, such as the standardization of management, the accuracy and personalization of services, etc.

[0003] The application of high-tech in smart communities can significantly improve service quality and management efficiency. Through technologies such as the Internet of Things, big data and artificial intelligence, intelligent management of communities can be achieved to improve the quality of life of residents. However, although digital technology can improve efficiency in community governance, it also faces many practical difficulties and needs to take corresponding countermeasures. For example, data silos, technical barriers and low resident participation have seriously affected the construction effect of smart communities.

[0004] Human-computer dialogue systems provide a convenient way for humans to interact with computers. Using natural language as the interface for human-computer interaction is more in line with human operating habits. In a dialogue system, in order to respond correctly to user instructions, the user instructions must first be translated into machine instructions that the computer can recognize. Intent recognition is a key task in natural language processing (NLP), which aims to accurately understand the intention or purpose behind the natural language text input by the user by analyzing it. In a human-computer dialogue system, intent recognition is the first step in translating user instructions into machine instructions that the computer can recognize.

[0005] The methods of intent recognition have evolved from statistical methods and traditional machine learning to deep learning and transfer learning. Early methods mainly relied on traditional technologies such as word frequency statistics and TFIDF to infer user intent by capturing important features of the text. These methods played an important role in the early stages, but with the development of technology, deep learning methods have gradually become mainstream, using complex neural network models to better capture the semantic and contextual information in the text.

[0006] However, in practical applications, these deep learning methods still have problems such as insufficient accuracy in understanding user input questions. Summary of the invention

[0007] This application provides a method for identifying question intent in the context of a smart community to solve problems that still exist in current solutions, such as insufficient accuracy in understanding questions input by users.

[0008] In order to achieve the above objectives, this application provides the following technical solutions:

[0009] In a first aspect, an embodiment of the present application provides a method for identifying question intentions in a smart community context, which includes:

[0010] Acquire historical records of social worker questions and answers collected through various methods as a basic data set, and preprocess the data in the basic data set to obtain a high-quality data set;

[0011] Build an intent recognition model based on BERT and use the high-quality dataset to train the model;

[0012] During the model training process, the model effect is evaluated, and the structure and parameters of the intent recognition model are optimized based on the evaluation results so that the effect of the intent recognition model meets the requirements;

[0013] Perform question intent recognition based on the trained intent recognition model.

[0014] Based on the above method, optionally, preprocessing the data in the basic data set includes:

[0015] The data in the basic data set is cleaned, labeled, segmented and divided.

[0016] Based on the above method, optionally, constructing a BERT-based intent recognition model and using the high-quality dataset to perform model training includes:

[0017] Based on the structure and principles of the BERT model, a model framework suitable for intent recognition in social engineering question-answering systems is constructed;

[0018] Initialize using pre-trained BERT model parameters;

[0019] Format and encode the preprocessed data;

[0020] The BERT model is trained using the formatted and encoded data.

[0021] Based on the above method, optionally, the training of the BERT model using the formatted and encoded data includes:

[0022] Dividing the high-quality data set into a training set, a test set, and a validation set;

[0023] Use the training set to train the model and optimize the training process by adjusting the hyperparameters;

[0024] Use the validation set to validate the model and monitor the performance changes of the model in a timely manner to prevent overfitting or underfitting;

[0025] In addition, based on the performance on the validation set, the model is fine-tuned, including adjusting the model structure, increasing or decreasing the number of layers, and modifying the loss function to improve the accuracy and generalization ability of the model.

[0026] Based on the above method, optionally, the evaluation of the model effect includes:

[0027] The model effect is evaluated using a variety of evaluation indicators, including precision, recall and F1 value.

[0028] Based on the above method, optionally, the social work question and answer history records collected in the multiple ways include records of community service centers, user feedback on online service platforms, and interaction records on social media.

[0029] In a second aspect, the embodiment of the present application further provides a problem intention identification device in the context of a smart community, which includes:

[0030] A data acquisition module is used to acquire historical records of social worker questions and answers collected in various ways as a basic data set, and preprocess the data in the basic data set to obtain a high-quality data set;

[0031] A model training module, used to build a BERT-based intent recognition model and use the high-quality dataset to train the model;

[0032] The model evaluation and optimization module is used to evaluate the model effect during the model training process, and optimize the structure and parameters of the intent recognition model based on the evaluation results so that the effect of the intent recognition model meets the requirements;

[0033] Application module, used to identify question intent based on the trained intent recognition model.

[0034] In a third aspect, an embodiment of the present application further provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor calls and executes the computer program, the method for identifying question intention in the context of a smart community as described in any one of the first aspects is implemented.

[0035] In the question intent identification method, device and equipment in the context of smart communities provided by this application, by introducing BERT deep learning technology, building a high-quality data set, and optimizing the structure and parameters of the model, the accuracy and efficiency of the social worker question-and-answer system in intent identification are significantly improved. As a result, it can not only improve the quality and efficiency of community services, promote the innovative development of smart community construction, but also enhance the happiness and satisfaction of community residents. Furthermore, through continuous innovation and application of technology, the present invention provides a new technical path and practical solution for the construction of smart communities. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application. In addition, these drawings and text descriptions are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application for those skilled in the art by referring to specific embodiments.

[0037] Figure 1 A flowchart of a method for identifying question intentions in a smart community context provided by an embodiment of the present application;

[0038] Figure 2 A schematic diagram of the structure of a question intention identification device in the context of a smart community provided by an embodiment of the present application;

[0039] Figure 3 A schematic diagram of the structure of an electronic device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0041] Current deep learning methods still have problems such as insufficient accuracy in understanding user input questions.

[0042] After research, it was found that the large language model based on BERT (Bidirectional Encoder Representations from Transformers) can better understand contextual relationships through its bidirectional encoder architecture, thereby significantly improving the accuracy and effectiveness of intent recognition. The BERT model uses large-scale unlabeled data for pre-training in two stages, pre-training, and then fine-tuning on specific tasks, thereby achieving significant performance improvements in various natural language processing tasks. The application of deep learning models such as BERT has brought new breakthroughs in the field of intent recognition. The advantage of deep learning lies in its powerful feature learning ability, which enables the model to automatically learn useful features from large amounts of data without human intervention. Compared with traditional methods, deep learning models can better capture complex semantic relationships and contextual information, thereby significantly improving the performance of intent recognition.

[0043] Rule-based methods also perform well in specific fields. By formulating clear rules, these systems can operate effectively in specific environments. These methods are generally suitable for domain-specific tasks, and by manually defining rules, basic intent recognition functions can be quickly implemented. However, rule-based methods may be insufficient when faced with complex and changing language environments because they are difficult to adapt to new language phenomena and user needs. In order to solve the above problems, the present application provides a method for question intent recognition in the context of a smart community, which improves the accuracy and efficiency of social work question and answer systems in intent recognition by introducing BERT deep learning technology. The following is a non-restrictive description of the specific implementation scheme through several examples or embodiments.

[0044] Some embodiments of the present application provide a method for identifying question intentions in the context of a smart community. In specific implementation, the method may be executed by an electronic device, such as a computer or a server. More specifically, a processing software may be running in the electronic device, and the software implements the corresponding method when running.

[0045] Reference Figure 1 , Figure 1 A flowchart of a method for identifying question intentions in a smart community context provided by an embodiment of the present application. Figure 1 As shown, the problem intention identification method in the context of a smart community in this embodiment includes the following steps:

[0046] Step S101: Acquire historical records of social worker questions and answers collected in various ways as a basic data set, and preprocess the data in the basic data set to obtain a high-quality data set.

[0047] Specifically, in order to make the model cover as many different application scenarios as possible, we first collect historical records of social workers’ questions and answers as basic data through various means, including field surveys, face-to-face conversations with community workers, and software systems. Data sources include records from community service centers, user feedback from online service platforms, interaction records on social media, and so on. In addition, it is necessary to ensure the diversity of data and ensure that the data set covers a variety of scenarios and question types, including but not limited to community service consultation, complaint handling, event scheduling, emergency assistance, etc.

[0048] After that, data preprocessing is performed to improve the validity of the data. Preprocessing the data in the basic data set may specifically include: data cleaning, data labeling, data segmentation and data division of the data in the basic data set. Among them:

[0049] Data cleaning: remove irrelevant information, correct errors, fill in missing values, and ensure data accuracy and consistency.

[0050] Data annotation: Each conversation record in the dataset is annotated in detail, including question type, user intent, background information, etc. The annotation work is completed by an experienced annotation team.

[0051] Data segmentation: Use natural language processing tools to perform preprocessing operations such as word segmentation and part-of-speech tagging to prepare for subsequent model training.

[0052] Data partitioning: Divide the preprocessed data set into training set, validation set, and test set to ensure the scientificity and reliability of model training and evaluation.

[0053] Through the above process, a high-quality data set is obtained for subsequent model training. By building a high-quality data set that covers a variety of scenarios and problem types and performing preprocessing, we ensure that the model can cope with various complex situations. This enables community workers to understand residents' needs more accurately and provide personalized services and support.

[0054] Step S102: Build a BERT-based intent recognition model and use a high-quality dataset to train the model.

[0055] Specifically, in this embodiment, BERT is selected as the core model. BERT can better understand the contextual relationship through its bidirectional encoder architecture, thereby significantly improving the accuracy and robustness of intent recognition. The model structure includes an input layer, a BERT encoding layer, a fully connected layer, and an output layer.

[0056] In some embodiments, a BERT-based intent recognition model is constructed and a high-quality dataset is used for model training, which may specifically include:

[0057] Based on the structure and principles of the BERT model, a model framework suitable for intent recognition in social engineering question-answering systems is constructed;

[0058] Initialize using pre-trained BERT model parameters to fully leverage its prior knowledge in natural language processing tasks;

[0059] Format and encode the preprocessed data to meet the requirements of BERT model input;

[0060] The BERT model is trained using the formatted and encoded data.

[0061] Further, the BERT model is trained using the formatted and encoded data, which may include:

[0062] Dividing the high-quality data set into a training set, a test set, and a validation set; wherein the training set is used for model training, the validation set is used to adjust model hyperparameters and evaluate model performance, and the test set is used to evaluate the performance of the trained model;

[0063] Use the training set to train the model and optimize the training process by adjusting hyperparameters, including learning rate, batch size, etc.

[0064] Use the validation set to validate the model and monitor the performance changes of the model in a timely manner to prevent overfitting or underfitting;

[0065] In addition, based on the performance on the validation set, the model is fine-tuned, including adjusting the model structure, increasing or decreasing the number of layers, and modifying the loss function to improve the accuracy and generalization ability of the model.

[0066] In this way, by optimizing the generalization ability and practical application effect of the model, it can ensure that the model's performance in different communities and different scenarios can achieve the expected results. This will help community managers deal with various problems of residents more efficiently and improve management efficiency.

[0067] Step S103: During the model training process, the model effect is evaluated, and the structure and parameters of the intent recognition model are optimized based on the evaluation results so that the effect of the intent recognition model meets the requirements.

[0068] Specifically, in order to ensure that the model achieves the expected effect, the model needs to be continuously evaluated and optimized during the model training process.

[0069] The optimization of the model includes:

[0070] Hyperparameter tuning: Use grid search, random search and other methods to find the best hyperparameter combination to improve model performance.

[0071] Model pruning: Through model pruning technology, the computational complexity of the model is reduced and the reasoning speed of the model is improved.

[0072] Knowledge distillation: Through knowledge distillation technology, the knowledge of large models is transferred to small models to improve the computational efficiency and generalization ability of the model.

[0073] In addition, in some embodiments, the model effect is evaluated, including:

[0074] Use a variety of evaluation indicators to evaluate the model effect to ensure that the model's performance in different scenarios meets the expected results. The various evaluation indicators include precision, recall, and F1 value. The calculation method is as follows:

[0075]

[0076] In the formula, TP means that the prediction is a positive example and the actual one is a positive example, FN means that the prediction is a negative example but the actual one is a positive example, and PR means precision*recall.

[0077] Through the above evaluation indicators, the performance of the model can be evaluated so that the model can be optimized according to the evaluation results.

[0078] Step S104: performing question intent recognition based on the trained intent recognition model.

[0079] Specifically, after the model training is completed, it can be used to identify the intention of the question. For example, the model is configured in a social engineering question-answering system, which can obtain the questions input by the user, process them, and input them into the intention recognition model, which then outputs the corresponding answers.

[0080] In addition, a user feedback mechanism can be established to continuously collect user opinions and suggestions, optimize the model in a timely manner, and improve service quality and user experience.

[0081] Based on the above scheme, by introducing the BERT deep learning technology, the present invention can more accurately understand the natural language text input by the user, thereby improving the accuracy of intent recognition. The bidirectional encoder architecture of the BERT model can better capture contextual information, allowing the model to perform better when processing complex and changeable user input. The introduction of a multi-task learning strategy, while training the model to perform tasks such as intent recognition and slot filling, further enhances the comprehensive capabilities of the model and improves the accuracy and robustness of recognition. By optimizing the structure and parameters of the model, the present invention significantly improves the reasoning speed of the model, ensuring that the model can quickly respond to user requests in practical applications and provide instant service support. Specifically:

[0082] This invention builds a high-quality social worker question-answering system dataset that covers a variety of scenarios and question types, ensuring that the model can cope with a variety of complex situations. This enables community workers to understand residents' needs more accurately and provide personalized services and support. Through real-time performance optimization, the model can quickly give accurate answers after users ask questions, improving the timeliness and satisfaction of services.

[0083] The present invention ensures that the performance of the model in different communities and different scenarios achieves the expected effect by optimizing the generalization ability and practical application effect of the model. This helps community managers to deal with various problems of residents more efficiently and improve management efficiency. By establishing a user feedback mechanism, the present invention can continuously collect users' opinions and suggestions, optimize the model in a timely manner, and improve service quality and user experience.

[0084] This invention provides solid technical support for the intelligent improvement of social work question-answering systems by introducing the latest deep learning technology, especially the BERT model. This not only enriches the research in the field of intent recognition, but also provides new ideas and methods for technological innovation in the construction of smart communities. Through technologies such as multi-task learning and transfer learning, this invention effectively solves the problem of small sample learning, improves the performance and robustness of the system, and provides a new technical path for the construction of smart communities.

[0085] The present invention can solve various problems and needs of residents in a timely manner by providing efficient and accurate social work question-and-answer services, thereby improving residents' satisfaction and sense of happiness. Through the user feedback mechanism, the present invention can continuously optimize services, ensure that residents' needs are better met, and improve the overall service level of the community.

[0086] In summary, the present invention significantly improves the accuracy and efficiency of the social worker question-and-answer system in intent recognition by introducing BERT deep learning technology, constructing a high-quality data set, and optimizing the structure and parameters of the model. The present invention not only improves the quality and efficiency of community services, promotes the innovative development of smart community construction, but also enhances the happiness and satisfaction of community residents. Through continuous innovation and application of technology, the present invention provides a new technical path and practical solution for the construction of smart communities.

[0087] In addition, the embodiment of the present application provides a problem intention recognition device in the context of a smart community, referring to Figure 2 , the device comprises:

[0088] The data acquisition module 21 is used to acquire the social worker question and answer history records collected in various ways as a basic data set, and pre-process the data in the basic data set to obtain a high-quality data set;

[0089] A model training module 22 is used to build a BERT-based intent recognition model and use high-quality data sets for model training;

[0090] The model evaluation and optimization module 23 is used to evaluate the model effect during the model training process, and optimize the structure and parameters of the intent recognition model based on the evaluation results so that the effect of the intent recognition model meets the requirements;

[0091] The application module 24 is used to perform question intent recognition based on the trained intent recognition model.

[0092] Among them, regarding the specific implementation methods of each module of the question intention identification device in the above-mentioned smart community background, reference can be made to the corresponding contents in the aforementioned method embodiment, which will not be repeated here.

[0093] In addition, an embodiment of the present application provides an electronic device, such as Figure 3 As shown, the electronic device includes a memory 31 and a processor 32; wherein the memory 31 stores a computer program, and when the processor 32 calls and executes the computer program, the problem intention identification method in the context of a smart community in any of the above embodiments is implemented.

[0094] The electronic device may be a computer or a server.

[0095] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0096] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0097] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0098] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0099] A person skilled in the art may understand that all or part of the steps in the above-mentioned embodiment method may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0100] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0101] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0102] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for identifying question intentions in the context of smart communities, characterized in that: include: Acquire historical records of social worker questions and answers collected through various methods as a basic data set, and preprocess the data in the basic data set to obtain a high-quality data set; Build an intent recognition model based on BERT and use the high-quality dataset to train the model; During the model training process, the model effect is evaluated, and the structure and parameters of the intent recognition model are optimized based on the evaluation results so that the effect of the intent recognition model meets the requirements; Perform question intent recognition based on the trained intent recognition model.

2. The method according to claim 1, characterized in that The preprocessing of the data in the basic data set includes: The data in the basic data set is cleaned, labeled, segmented and divided.

3. The method according to claim 1, characterized in that: The construction of the BERT-based intent recognition model and the use of the high-quality dataset for model training include: Based on the structure and principles of the BERT model, a model framework suitable for intent recognition in social engineering question-answering systems is constructed; Initialize using pre-trained BERT model parameters; Format and encode the preprocessed data; The BERT model is trained using the formatted and encoded data.

4. The method according to claim 3, characterized in that The BERT model is trained using the formatted and encoded data, including: Dividing the high-quality data set into a training set, a test set, and a validation set; Use the training set to train the model and optimize the training process by adjusting the hyperparameters; Use the validation set to validate the model and monitor the performance changes of the model in a timely manner to prevent overfitting or underfitting; In addition, based on the performance on the validation set, the model is fine-tuned, including adjusting the model structure, increasing or decreasing the number of layers, and modifying the loss function to improve the accuracy and generalization ability of the model.

5. The method according to claim 1, characterized in that The evaluation of the model effect includes: The model effect is evaluated using a variety of evaluation indicators, including precision, recall and F1 value.

6. The method according to claim 1, characterized in that The social work question and answer history records collected through the various methods include records from community service centers, user feedback from online service platforms, and interaction records on social media.

7. A device for identifying question intentions in the context of a smart community, characterized in that: include: A data acquisition module is used to acquire historical records of social worker questions and answers collected in various ways as a basic data set, and preprocess the data in the basic data set to obtain a high-quality data set; A model training module, used to build a BERT-based intent recognition model and use the high-quality dataset to train the model; The model evaluation and optimization module is used to evaluate the model effect during the model training process, and optimize the structure and parameters of the intent recognition model based on the evaluation results so that the effect of the intent recognition model meets the requirements; Application module, used to identify question intent based on the trained intent recognition model.

8. An electronic device, characterized in that: It includes a memory and a processor, the memory stores a computer program, and when the processor calls and executes the computer program, it implements the problem intention identification method in the context of a smart community as described in any one of claims 1 to 6.

Images