Property customer service question and answer model training method, question and answer method, equipment and medium

Abstract

The application discloses a training method of a property customer service question and answer model, and comprises the following steps: inputting a multi-stage knowledge mask strategy into a BERT model for pre-training to obtain different language granularity information; and performing domain recognition training based on the language granularity information to obtain the property customer service question and answer model. According to the application, the multi-stage knowledge mask strategy and the domain recognition are trained in the BERT model, so that the user's question can be accurately understood, and the user experience is improved.

Description

Technical Field

[0001] This invention relates to the field of smart property management, and in particular to a training method, question-and-answer method, equipment, and medium for a property customer service question-and-answer model. Background Technology

[0002] In the construction and real estate property services sector, there are issues such as poor quality and low accuracy of human-computer interaction, limited application of such technology, a high proportion of manual customer service, and high customer service turnover. Industry research shows that among online service personnel in the construction industry, traditional customer service practitioners have a job dissatisfaction rate as high as 51%. The main reasons for dissatisfaction include high workload, monotonous and repetitive work content. In the work of customer service personnel, simple inquiries account for 39%, and are highly concentrated in a few categories. Repetitive and standardized questions consume a large proportion of the working time of customer service personnel.

[0003] In comparison, intelligent customer service provides accurate answers and can resolve inquiries. It can make semantic or intent judgments based on context. It improves the efficiency of property and real estate customer service in solving problems, supports one-to-one, one-to-many, and many-to-one communication, and can answer homeowners' questions in real time without queuing during peak hours. It can achieve 24 / 7 uninterrupted response, full-time coverage, and can resolve urgent issues in a timely manner, and can connect to human customer service when necessary.

[0004] Natural Language Processing (NLP) is a field that integrates linguistics, computer science, and artificial intelligence. A branch of data science, NLP is a systematic process for analyzing, understanding, and extracting information from textual data in an intelligent and efficient manner. By using NLP and its components, a wide variety of text-related problems can be solved, such as text similarity analysis, automatic summarization, machine translation, named entity recognition, relation extraction, sentiment analysis, and topic segmentation.

[0005] Currently, the leading multi-turn dialogue technology is Google's Chinese BERT model, which is trained on Chinese Wikipedia data and belongs to the category of general-domain pre-trained language models. However, it lacks business corpora in the property management field, resulting in low accuracy in understanding business common sense and semantics, and a poor user experience. Summary of the Invention

[0006] This invention provides a training method, question-and-answer method, equipment, and medium for a property customer service question-and-answer model, aiming to solve the problems of low accuracy in understanding business common sense and semantics and poor user experience mentioned above.

[0007] The technical solution is as follows:

[0008] On the one hand, a training method for a property customer service question-and-answer model is provided, including the following steps:

[0009] A multi-stage knowledge masking strategy is input into the BERT model for pre-training to obtain information at different language granularities.

[0010] Domain recognition training based on language granularity information is used to obtain a property customer service question-and-answer model.

[0011] Preferably, the multi-stage knowledge masking strategy includes a single-word masking stage, a phrase masking stage, an entity masking stage, and a high-frequency speech masking stage.

[0012] Preferably, the entity masking stage includes technical terms, personal names, geographical names, organization names, and product names.

[0013] Preferably, the data types used for domain recognition training include labeled data from knowledge bases of various domains, unlabeled data under weak supervision in various domains, and personalized data.

[0014] Preferably, personalized data includes the user's order status.

[0015] On the other hand, a customer service Q&A method is provided, including the following steps:

[0016] Obtain the question information input by the user;

[0017] Based on the question information, the answer information corresponding to the question information is obtained from the property customer service question and answer model, wherein the property customer service question and answer model is the property customer service question and answer model obtained by the aforementioned training method.

[0018] Provide the answer information to the user.

[0019] Preferably, the step of obtaining the answer information corresponding to the question information from the property customer service question-and-answer model specifically includes:

[0020] Determine the business type of the question information using a property customer service Q&A model;

[0021] Retrieve the data source type for the business type;

[0022] Retrieve answer information based on data source type.

[0023] Preferably, the data source types include basic property information, community data, and unstructured data.

[0024] On the other hand, a computer device is provided, comprising:

[0025] The processor, memory, and communication circuitry are connected to the memory and communication circuitry, respectively.

[0026] The communication circuit is used for communication connection, the memory is used to store computer programs, and the processor is used to execute the computer programs to implement any of the above methods.

[0027] On the other hand, a computer-readable storage medium is provided, which stores a computer program that can be executed by a processor to implement any of the above methods.

[0028] The beneficial effects of this invention are: by using the solution of this invention to train the multi-stage knowledge masking strategy and domain recognition in the BERT model, it is possible to accurately understand the user's questions and improve the user experience. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating a training method for a property customer service question-and-answer model according to an embodiment of the present invention.

[0030] Figure 2 This is a flowchart illustrating a customer service question-and-answer method according to an embodiment of the present invention;

[0031] Figure 3 This is a schematic block diagram illustrating the structure of the computer device embodiment of this application;

[0032] Figure 4 This is a schematic block diagram of the structure of an embodiment of a computer-readable storage medium according to this application. Detailed Implementation

[0033] To facilitate understanding of the present invention, a more detailed description is provided below with reference to the accompanying drawings and specific embodiments. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0034] It should be noted that, unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention.

[0035] Please see Figure 1 , Figure 1 This is a flowchart illustrating a training method for a property customer service question-and-answer model according to an embodiment of the present invention. This application provides a training method for a property customer service question-and-answer model, comprising the following steps:

[0036] S101: Input the multi-stage knowledge masking strategy into the BERT model for pre-training to obtain different language granularity information.

[0037] The property customer service question-answering model in this application is based on BERT, which stands for Bidirectional Encoder Representations from Transformer. As the name suggests, the goal of the BERT model is to train on a large-scale unlabeled corpus to obtain a representation of the text containing rich semantic information, i.e., the semantic representation of the text. Then, the semantic representation of the text is fine-tuned for a specific NLP task and finally applied to that NLP task.

[0038] Among them, the multi-stage knowledge masking strategy includes single-word masking stage, phrase masking stage, entity masking stage, and high-frequency speech masking stage.

[0039] Phase 1: Single-word masking. This is a method inherent to BERT, as masking is performed at the "word" level in English and at the "character" level in Chinese. After this phase, the model learns a basic character-based language representation.

[0040] The second stage: Phrase masking. For Chinese, language-dependent word segmentation tools are used to obtain phrase boundaries. In this stage, the basic unit of masking is the phrase, and phrases are treated as a whole. Through this process, phrase information is added to the pre-selected words.

[0041] Phase Three: Entity Masking. Unlike phrases, entities tend to be more technical terms, personal names, geographical names, organization names, and product names. This material comes from the property industry's own knowledge base. The basic unit of masking in this phase is the entity, and entities are treated as a whole for operation.

[0042] Phase Four: High-Frequency Phrases Masking. Because there are many common high-frequency phrases in the property management industry, such as "power outage in the community," these phrases will be treated as a whole and managed accordingly.

[0043] The specific results are shown in the table below:

[0044]

[0045] This training method can yield rich and varied linguistic granular information, enabling accurate understanding of user questions during customer service conversations.

[0046] S102: Domain recognition training based on language granularity information to obtain a property customer service question-and-answer model.

[0047] In this application, the domain recognition training mainly involves modeling three types of data: labeled data from knowledge bases in various domains, a large amount of weakly supervised unlabeled data from various domains, and personalized data.

[0048] (1) Based on the problem understanding model signals learned from the labeled data of the knowledge base of each domain, it is possible to determine the probability that the user input belongs to each business and each intention.

[0049] (2) Different application integrated service entry points involve multiple businesses, and the dialogue data entering through these entry points has clear business label information. Therefore, a large amount of weakly supervised data from various business domains can be obtained, and based on this data, we can train a first-level classification model.

[0050] (3) Some issues require further clarification by combining personalized data such as user order status. For example, "I want to pay" appears in multiple business transactions. Therefore, it is necessary to train a secondary model by combining user status features to ultimately determine which business transaction the user's input belongs to.

[0051] After being trained using a multi-stage knowledge masking strategy and domain recognition, the property customer service question-answering model can accurately understand users' question information and classify it in preparation for answering questions.

[0052] Please see Figure 2 , Figure 2 This is a flowchart illustrating a customer service question-and-answer method according to an embodiment of the present invention. This application provides a training method for a property customer service question-and-answer model, comprising the following steps:

[0053] S201: Obtain the question information input by the user.

[0054] In a specific application scenario, the user's question originates from a terminal application (such as the Youjia app of China Overseas Property). This application contains multiple business functions, and the question information entered by the user may involve any of these functions. These functions include, but are not limited to, reporting issues and repairs, paying property fees, calling the property manager, visitor appointments, convenience services, release slips, shopping on premium products, and community forums.

[0055] S202: Obtain the answer information corresponding to the question information from the property customer service question and answer model based on the question information, wherein the property customer service question and answer model is the property customer service question and answer model obtained by the aforementioned training method of the property customer service question and answer model.

[0056] The property customer service Q&A model described here is the same as the one in the previous embodiment, and will not be elaborated upon further. Once the property customer service Q&A model accurately understands the user's question, it can obtain the answer information through the following methods:

[0057] Determine the business type of the question information using a property customer service Q&A model;

[0058] For example, the business type of the problem information can be identified through the classification model formed during domain recognition training.

[0059] Retrieve the data source type for the business type;

[0060] The data source types include, but are not limited to, three types: basic property information, community data, and unstructured data.

[0061] (1) For basic property information, such as business hours, address, telephone number, etc., a graph is constructed using the basic information of the business. The problem is understood through the problem understanding model, and then the graph is queried to obtain the accurate answer.

[0062] (2) For community data, namely the community data of user questions and answers in the "Ask Everyone" module in the community details page, construct question and answer capabilities. By modeling the similarity between user questions and "question and answer pairs" in "Ask Everyone", select the one with the highest similarity as the answer to answer some open questions of users.

[0063] (3) For unstructured data, build document question answering capabilities. For user questions, use machine reading comprehension technology to extract answers from documents, similar to reading comprehension questions, to further answer some open questions from users.

[0064] Retrieve answer information based on data source type.

[0065] S203: Provide the answer information to the user.

[0066] The customer service Q&A method in this application can provide single-round or multi-round dialogue.

[0067] The beneficial effects of this invention are: by using the solution of this invention to train the multi-stage knowledge masking strategy and domain recognition in the BERT model, it is possible to accurately understand the user's questions and improve the user experience.

[0068] Please see Figure 3 The computer device 20 described in the embodiments of the computer device in this application may specifically include a processor 210 and a memory 220. The memory 220 is coupled to the processor 210.

[0069] Processor 210 is used to control the operation of computer device 20. Processor 210 may also be referred to as CPU (Central Processing Unit). Processor 210 may be an integrated circuit chip with signal processing capabilities. Processor 210 may also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. The general-purpose processor may be a microprocessor, or processor 210 may be any conventional processor.

[0070] Memory 220 is used to store computer programs and may be RAM, ROM, or other types of storage devices. Specifically, the memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory is used to store at least one line of program code.

[0071] The processor 210 is used to execute computer programs stored in the memory 220 to implement the methods described in the various method embodiments of this application.

[0072] In some embodiments, the computer device 20 may further include a peripheral device interface 230 and at least one peripheral device. The processor 210, memory 220, and peripheral device interface 230 may be connected via a bus or signal line. Each peripheral device may be connected to the peripheral device interface 230 via a bus, signal line, or circuit board. Specifically, the peripheral device includes at least one of a radio frequency circuit 240, a display screen 250, an audio circuit 260, and a power supply 270.

[0073] Peripheral device interface 230 can be used to connect at least one I / O (Input / output) related peripheral device to processor 210 and memory 220. In some embodiments, processor 210, memory 220 and peripheral device interface 230 are integrated on the same chip or circuit board; in some other embodiments, any one or two of processor 210, memory 220 and peripheral device interface 230 can be implemented on separate chips or circuit boards, which is not limited in this embodiment.

[0074] The radio frequency (RF) circuit 240 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The RF circuit 240 communicates with communication networks and other communication devices via electromagnetic signals; it is the communication circuit of the computer device 20. The RF circuit 240 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals back into electrical signals. Optionally, the RF circuit 240 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, etc. The RF circuit 240 can communicate with other terminals through at least one wireless communication protocol. This wireless communication protocol includes, but is not limited to: the World Wide Web, metropolitan area networks, intranets, various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks, and / or WiFi (Wireless Fidelity) networks. In some embodiments, the RF circuit 240 may also include circuitry related to NFC (Near Field Communication), which is not limited in this application.

[0075] Display screen 250 is used to display a UI (User Interface). This UI may include graphics, text, icons, videos, and any combination thereof. When display screen 250 is a touch display screen, it also has the ability to collect touch signals on or above its surface. These touch signals can be input as control signals to processor 210 for processing. In this case, display screen 250 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there may be one display screen 250, located on the front panel of the computer device 20; in other embodiments, there may be at least two display screens, respectively located on different surfaces of the computer device 20 or in a folded design; in still other embodiments, display screen 250 may be a flexible display screen, located on a curved or folded surface of the computer device 20. Furthermore, display screen 250 may be configured as a non-rectangular, irregular shape, i.e., a non-rectangular screen. Display screen 250 may be made of materials such as LCD (Liquid Crystal Display) or OLED (Organic Light-Emitting Diode).

[0076] The audio circuit 260 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, converting them into electrical signals that are input to the processor 210 for processing, or to the radio frequency circuit 240 for voice communication. For stereo sound acquisition or noise reduction purposes, multiple microphones may be used, each positioned at a different location on the computer device 20. The microphone may also be an array microphone or an omnidirectional microphone. The speaker is used to convert electrical signals from the processor 210 or the radio frequency circuit 240 into sound waves. The speaker may be a conventional diaphragm speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can convert electrical signals not only into audible sound waves but also into inaudible sound waves for purposes such as distance measurement. In some embodiments, the audio circuit 260 may also include a headphone jack.

[0077] Power supply 270 is used to supply power to the various components in computer device 20. Power supply 270 can be alternating current, direct current, a disposable battery, or a rechargeable battery. When power supply 270 includes a rechargeable battery, the rechargeable battery can be a wired rechargeable battery or a wireless rechargeable battery. A wired rechargeable battery is a battery that is charged via a wired line, while a wireless rechargeable battery is a battery that is charged via a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0078] For a detailed description of the functions and execution processes of each functional module or component in the computer device embodiments of this application, please refer to the descriptions in the above-described method embodiments of this application, which will not be repeated here.

[0079] In the embodiments provided in this application, it should be understood that the disclosed computer devices and methods can be implemented in other ways. For example, the computer device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0080] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0081] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] refer to Figure 4 If the integrated units described above are implemented as software functional units and sold or used as independent products, they can be stored in computer-readable storage medium 300. Based on this understanding, the technical solution of this application, in essence, 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. This computer software product is stored in a storage medium and includes several instructions / computer programs to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this application. The aforementioned storage medium includes various media such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, as well as electronic devices such as computers, mobile phones, laptops, tablets, and cameras that have the aforementioned storage media.

[0083] The execution process of program data in a computer-readable storage medium can be described with reference to the above-described method embodiments of this application, and will not be repeated here.

[0084] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A training method for a property customer service question-and-answer model, characterized in that, Includes the following steps: A multi-stage knowledge masking strategy is input into the BERT model for pre-training to obtain different language granularity information. The multi-stage knowledge masking strategy includes: single-word masking stage, phrase masking stage, entity masking stage, and high-frequency slang masking stage. A property customer service question-and-answer model is obtained by training domain recognition based on the aforementioned language granularity information. This model takes the user's question information as input and the business function category of the question as output. The business functions include reporting issues / repairs, property fee payment, calling the property manager, visitor appointments, convenience services, shopping from premium products, and community forums. The domain recognition training includes modeling labeled data from various domain knowledge bases, weakly supervised unlabeled data from various domains, and personalized data to obtain a property customer service question-and-answer model. Based on the problem understanding model signals learned from the labeled data of the knowledge bases of the various domains, the property customer service question-and-answer model is used to determine the likelihood that the user input belongs to each business intent. Train a first-level classification model based on the weakly supervised, unlabeled data from each of the aforementioned domains; A secondary model is trained based on the personalized data to determine which service the user's input belongs to.

2. The training method according to claim 1, characterized in that, The entity masking stage includes technical terms, personal names, geographical names, organization names, and product names.

3. The training method according to claim 2, characterized in that, The personalized data includes the user's order status.

4. A customer service question-and-answer method, characterized in that, Includes the following steps: Obtain the question information input by the user; Based on the question information, the answer information corresponding to the question information is obtained from the property customer service question and answer model, wherein the property customer service question and answer model is the property customer service question and answer model obtained by the training method of the property customer service question and answer model according to any one of claims 1-3; The answer information will be provided to the user.

5. The customer service question-and-answer method according to claim 4, characterized in that, The step of obtaining the answer information corresponding to the question information from the property customer service question-and-answer model based on the question information specifically includes: The business type of the question information is determined by the property customer service question-and-answer model. Obtain the data source type for the business type; The answer information is obtained based on the data source type.

6. The customer service question-and-answer method according to claim 5, characterized in that, The data source types include basic property information, community data, and unstructured data.

7. A computer device, characterized in that, include: The processor, memory, and communication circuitry are respectively connected to the memory and the communication circuitry. The communication circuit is used for communication connection, the memory is used to store computer programs, and the processor is used to execute the computer programs to implement the training method as described in any one of claims 1-3 and the customer service Q&A method as described in any one of claims 4-6.

8. A computer-readable storage medium, characterized in that, The system contains a computer program that can be executed by a processor to implement the training method as described in any one of claims 1-3 and the customer service question-and-answer method as described in any one of claims 4-6.

Images