Property Q&A processing methods, devices, electronic equipment and storage media

Abstract

This application discloses a property management question-and-answer processing method, apparatus, electronic device, and storage medium, relating to the field of artificial intelligence technology. This application implements a method based on user input information, first determining the primary category corresponding to the user input information as the target first work order label, then using the target secondary work order label matched by the target first work order label as context, and further determining the secondary category corresponding to the user input information as the target second work order label through a property management question-and-answer model. Then, using the target tertiary work order label matched by the target second work order label, the target property knowledge statement, and multiple priority labels as context, the property management question-and-answer model generates the answer text, the tertiary category corresponding to the user input information, and the target work order priority label. Thus, by using context to guide the property management question-and-answer model, the accuracy of work order classification is improved, and the priority classification of work orders is more accurate, resulting in more accurate generated answer text.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a property question-and-answer processing method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the acceleration of urbanization, the property management industry is facing the dual challenges of expanding service scale and diversifying customer needs. Traditional Property Management Systems (PMS) have evolved from paper-based to electronic to rudimentary intelligent systems. In the electronic stage, work order processing mainly relied on manual input or simple rule-based systems. These systems typically predefine a series of keywords (such as "water leak" or "power outage"), and when the text submitted by the user contains these words, the system triggers the corresponding classification rules. However, this method can only perform simple word matching, resulting in a high error rate in work order classification and the inability to generate text responses to users. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a property-related question-and-answer processing method, apparatus, electronic device, and storage medium, capable of generating answer text replies to users and generating work order tags based on user input, with high accuracy in work order tag classification.

[0004] The property inquiry processing method according to the first aspect of this application includes: Obtain a property management request message, and obtain user input information based on the property management request message; Based on the user input information, an initial question statement is generated, and the initial question statement is converted into a question vector; Based on the question vector, at least one target knowledge vector matching the question vector is determined from the target vector library; the target vector library includes multiple knowledge vectors converted from property knowledge statements. Based on at least one of the target knowledge vectors, determine the target property knowledge statements corresponding to the target knowledge vectors; Based on the target property knowledge statement, the target first work order label is determined from multiple first-level work order labels; Obtain multiple target secondary work order tags that match the target first work order tag; The preset first prompt word, multiple target secondary work order tags, the initial question statement, and the target property knowledge statement are concatenated to obtain the first question text. The first question text is then input into the trained property question-and-answer model to obtain the target second work order tag. Obtain multiple target third-level work order tags that match the second work order tag; The second question text is obtained by concatenating the preset second prompt word, multiple target level 3 work order tags, multiple preset priority tags, the initial question statement, and the target property knowledge statement. The second question text is input into the property question-and-answer model to obtain the answer text, the target third work order label, and the target work order priority label.

[0005] The property question-and-answer processing method according to the embodiments of this application has at least the following beneficial effects: In the property question-and-answer processing method of this application, the user input information is first converted into a question vector, the matching target knowledge vector is determined, and then the matching target property knowledge statement is determined according to the target knowledge vector. Based on the target property knowledge statement, the target first work order label is determined to achieve the first classification of subsequently generated work orders. The target second-level work order label matching the target first work order label is obtained, and the second-level work order label is a subcategory of the first-level work order label. The preset first prompt word, multiple target second-level work order labels, the initial question statement, and the target property knowledge statement are concatenated to obtain the first question text, and the first question text is input into the trained property question-and-answer model to obtain the target second work order label; the first prompt word is used to instruct the property question-and-answer model to generate the target second work order label, achieving the second classification of the work order. The target third-level work order label is obtained, and the target third-level work order label is a subcategory of the second work order label. The second prompt word, multiple target level 3 work order tags, multiple preset priority tags, the initial question statement, and the target property knowledge statement are concatenated to obtain the second question text. This second question text is then input into the property question-and-answer model to obtain the answer text, the target level 3 work order tag, and the target work order priority tag. The second prompt word instructs the property question-and-answer model to use the target property knowledge statement, multiple target level 3 work order tags, and multiple priority tags as context to answer the initial question statement and generate the target level 3 work order tag and the target work order priority tag. Thus, this application achieves the following: based on user input information, first, the primary category corresponding to the user input information is determined as the target level 1 work order tag; then, the target level 2 work order tag matched by the target level 1 work order tag is used as context, and the property question-and-answer model further determines the secondary category corresponding to the user input information as the target level 2 work order tag. Finally, the target level 3 work order tag matched by the target level 2 work order tag, the target property knowledge statement, and multiple priority tags are used as context to enable the property question-and-answer model to generate the answer text, the level 3 category corresponding to the user input information, and the target work order priority tag. In this way, by guiding the property question-and-answer model through context, the robustness of the model is improved, the accuracy of work order classification based on user input information is achieved, the priority classification of work orders is more accurate, and the generated answer text is more accurate.

[0006] According to some embodiments of the first aspect of this application, before determining at least one target knowledge vector matching the question vector from the target vector library based on the question vector, the method includes: Obtain public knowledge files and construct a public vector library based on the public knowledge files; Obtain the property knowledge file for each community unit, and construct a community vector library based on the property knowledge file; and the community vector library corresponds one-to-one with the community unit.

[0007] According to some embodiments of the first aspect of this application, the target vector library is determined through the following steps: Obtain user identity information based on the property request message; The target cell unit is determined based on the user identity information; From the multiple cell vector libraries, determine the target cell vector library that matches the target cell cell; The target cell vector library and the public vector library are used as the target vector library.

[0008] According to some embodiments of the first aspect of this application, generating an initial question statement based on the user input information includes: If the user input information is detected to include both text and image components, the image component is input into a preset visual language model to obtain an image description. The text portion is concatenated with the image description to obtain the initial question statement.

[0009] According to some embodiments of the first aspect of this application, after inputting the second question text into the property question-and-answer model to obtain the answer text, the target third work order tag, and the target work order priority tag, the following steps are included: The answer text is sent to the user terminal; Based on the initial question, the answer text, the target third work order tag, and the target work order priority tag, a target work order is constructed. Send the target work order to the terminal associated with the target third work order tag.

[0010] According to some embodiments of the first aspect of this application, each of the first-level work order tags corresponds to a keyword corpus; The step of determining the target first work order tag from multiple first-level work order tags based on the target property knowledge statement includes: The target property knowledge statement is matched with the keyword corpus corresponding to the multiple first-level work order tags to obtain multiple matching degree values; The target word corpus with the highest matching degree value is determined from multiple keyword corpora, and the first-level work order tag corresponding to the target word corpus is used as the target first work order tag.

[0011] According to some embodiments of the first aspect of this application, the construction of a public vector library based on the public knowledge document includes: The public knowledge document is divided into multiple property knowledge statements; The property knowledge statements are converted into knowledge vectors, and the public vector library is constructed based on the knowledge vectors.

[0012] A second aspect of this application provides a property inquiry processing device, comprising: The first acquisition module is used to acquire a property request message and acquire user input information based on the property request message. The conversion module is used to generate an initial question statement based on the user input information and convert the initial question statement into a question vector; The first determining module is used to determine at least one target knowledge vector that matches the question vector from a target vector library based on the question vector; the target vector library includes multiple knowledge vectors converted from property knowledge statements. The second determining module is used to determine the target property knowledge statement corresponding to the target knowledge vector based on at least one of the target knowledge vectors; The third determination module is used to determine the target first work order label from multiple first-level work order labels based on the target property knowledge statement. The second acquisition module is used to acquire multiple target secondary work order tags that match the target first work order tag; The first splicing module is used to splice the preset first prompt word, multiple target secondary work order tags, the initial question statement and the target property knowledge statement to obtain the first question text, and input the first question text into the trained property question and answer model to obtain the target second work order tag; The third acquisition module is used to acquire multiple target third-level work order tags that match the second work order tag; The second splicing module is used to splice the preset second prompt word, multiple target level 3 work order tags, multiple preset priority tags, the initial question statement, and the target property knowledge statement to obtain the second question text; The input module is used to input the second question text into the property question-and-answer model to obtain the answer text, the target third work order label, and the target work order priority label.

[0013] A third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the property inquiry processing method described in any one of the first aspects of the embodiment.

[0014] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the property inquiry-and-answer processing method described in any one of the first aspects of the embodiment.

[0015] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0016] The present application will be further described below with reference to the accompanying drawings and embodiments, wherein: Figure 1 This is a flowchart illustrating the steps of the property inquiry processing method according to an embodiment of this application; Figure 2 This is a schematic diagram of a process prior to step S130 in an embodiment of this application; Figure 3 This is a schematic diagram illustrating the process of determining the target vector library in an embodiment of this application; Figure 4 This is a schematic diagram of the property inquiry processing device according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0017] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0018] In the description of this application, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0019] In the description of this application, "several" means one or more, "multiple" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0020] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0021] In the description of this application, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] It should be noted that in all specific embodiments of this application, when processing data related to user identity or characteristics, such as user information, user behavior data, user historical data, and user location information, user permission or consent is obtained first. Furthermore, the collection, use, and processing of this data comply with relevant laws, regulations, and standards. In addition, when embodiments of this application require access to sensitive personal information of users, separate permission or consent from the user is obtained through pop-ups or redirection to confirmation pages. Only after obtaining the user's separate permission or consent is the necessary user-related data required for the proper functioning of these embodiments acquired.

[0023] The first aspect of this application provides a property management Q&A processing method. This method can be deployed and executed on a terminal, on a server, or as software running on either the terminal or the server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the property management Q&A processing method, but is not limited to the above forms.

[0024] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0025] Reference Figure 1 , Figure 1 This is a flowchart illustrating the steps of a property inquiry processing method according to an embodiment of this application. The property inquiry processing method of the first aspect of this application may include, but is not limited to, steps S110 to S200.

[0026] Step S110: Obtain the property management request message and obtain user input information based on the property management request message; In some embodiments, the property management question-and-answer processing method is deployed and executed on a server. Users communicate with the server through a user terminal, input question information through the user terminal, and send a property management request message to the server as user input information, so that the server can obtain the property management request message and extract user input information from the property management request message.

[0027] Step S120: Based on the user input information, generate an initial question statement and convert the initial question statement into a question vector; Step S130: Based on the question vector, determine at least one target knowledge vector that matches the question vector from the target vector library; the target vector library includes multiple knowledge vectors converted from property knowledge statements; It is worth noting that before step S130, multiple vector libraries are constructed in advance. Specifically, each property management document or legal document corresponds to one vector library. The property management document or legal document is cut into text blocks of fixed length, and then the text blocks are converted into vectors to obtain knowledge vectors, thus constructing multiple vector libraries, and each vector library includes multiple knowledge vectors. Furthermore, the corresponding text blocks can be determined through the knowledge vectors.

[0028] It is worth noting that in step S130, either a dense vector retrieval algorithm or a sparse vector retrieval algorithm can be used to perform the retrieval, thereby determining the target knowledge vector that matches the query vector.

[0029] Step S140: Based on at least one target knowledge vector, determine the target property knowledge statement corresponding to the target knowledge vector; Step S150: Based on the target property knowledge statement, determine the target first work order label from multiple first-level work order labels; In some embodiments, each primary work order label corresponds to a keyword corpus, which includes multiple keywords set by relevant personnel. Step S150 may include steps S151 and S152.

[0030] Step S151: Match the target property knowledge statement with the keyword corpus corresponding to multiple first-level work order tags to obtain multiple matching degree values; Specifically, the target property knowledge statement is broken down into multiple words to be matched. These multiple words are then matched against keyword corpora corresponding to multiple primary work order tags. If a keyword matching the word appears in the keyword corpus corresponding to a primary work order tag, it is considered a confirmed match keyword. The value of the confirmed match keyword is used as the match value between the target property knowledge statement and the keyword corpus corresponding to the primary work order tag. The larger the match value, the more confirmed match keywords are in the keyword corpus.

[0031] Step S152: Determine the target word corpus with the highest matching degree value from multiple keyword corpora, and use the first-level work order tag corresponding to the target word corpus as the target first work order tag.

[0032] Through steps S151 and S152, the target first work order label is determined from multiple first-level work order labels based on the target property knowledge statement. Since the classification of first-level work order labels is relatively simple, there is no need to use a property question-and-answer model for classification. Instead, classification is performed by word matching, which can improve efficiency while ensuring classification accuracy.

[0033] It is worth noting that before implementing the method of this application embodiment, the work order labels are set to three levels: Level 1 work order label, Level 2 work order label, and Level 3 work order label. Each Level 1 work order label corresponds to multiple Level 2 work order labels, and each Level 2 work order label corresponds to multiple Level 3 work order labels. For example, Level 1 work order labels may include safety management, customer service, and consultation. Level 2 work order labels corresponding to safety management may include equipment safety, circuit safety, and personnel safety. Level 2 work order labels corresponding to customer service may include customer complaints and work order processing feedback. Level 2 work order labels corresponding to consultation may include consultation on decoration regulations and consultation on vehicle management. Those skilled in the art can set work order labels at each level according to actual conditions.

[0034] Step S160: Obtain multiple target secondary work order tags that match the target first work order tag; Step S170: The preset first prompt word, multiple target secondary work order labels, initial question statement and target property knowledge statement are concatenated to obtain the first question text, and the first question text is input into the trained property question answering model to obtain the target second work order label; It is worth noting that the first prompt word is used to instruct the property management question-and-answer model to generate a target second work order label, thereby achieving a second classification of the work order. For example, the first concatenated text is: "You are a professional, patient, and efficient property manager. Please classify the user input information into a work order based on the reference materials. Please select one of the following categories as the target category: fire hazard, facility and equipment safety hazard, personnel safety hazard. Reference material: Exposed wires may cause electrical leakage. User input information: Exposed wires in the corridor, and the lights are not working." In this first concatenated text, "You are a professional, patient, and efficient property manager. Please classify the user input information into a work order based on the reference materials. Please select one of the following categories as the target category:", "Reference material:", and "User input information:" serve as the first prompt word; fire hazard, facility and equipment safety hazard, and personnel safety hazard are the target second-level work order labels; "Exposed wires may cause electrical leakage" is the target property knowledge statement; and "Exposed wires in the corridor, and the lights are not working" is the initial question statement. The above is merely an example and should not be construed as limiting this application. Those skilled in the art are free to set the first prompt word, as well as the splicing order between the first prompt word, multiple target secondary work order labels, the initial question statement, and the target property knowledge statement.

[0035] Step S180: Obtain multiple target third-level work order tags that match the second work order tag; Step S190: The preset second prompt word, multiple target level 3 work order tags, multiple preset priority tags, initial question statement and target property knowledge statement are concatenated to obtain the second question text; It is worth noting that the second prompt word instructs the property management question-answering model to use the target property knowledge statement, multiple target level 3 work order tags, and multiple priority tags as context to answer the initial question and generate the target level 3 work order tag and the target work order priority tag. This allows the target property knowledge statement, multiple target level 3 work order tags, and multiple priority tags to guide the property management question-answering model, improving the model's accuracy. For example, the second concatenated text is: "You are a professional, patient, and efficient property manager. Please answer the user's work order questions based on the following references and provide preliminary processing opinions. Also, categorize the user's work order questions; please select one of the following categories as the target category:" Exposed circuits, tripped circuit breakers, blackened circuit breaker exterior; Please prioritize user work order issues and select one of the following priority tags as the target priority tag: Very Urgent; Moderately Urgent; Not Urgent. The reference material states: Exposed electrical wires may cause electrical leakage; The user's work order problem is: "Exposed wires in the hallway, and the lights are not working." In the second concatenated text, the following are the prompts: "You are a professional, patient, and efficient property manager. Please answer the user's work order problem based on the following references and provide preliminary solutions. The work order problem is categorized; please select one of the following categories as the target category:", "The user's work order problem is prioritized; please select one of the following priority tags as the target priority tag", "References are:", and "User's work order problem is:" are the second prompts. Exposed wiring, tripped circuit breaker, and blackened circuit breaker are the target level 3 work order tags. Very urgent, moderately urgent, and not urgent are the priority tags. "Exposed wires may cause electrical leakage" is the target property knowledge statement. "Exposed wires in the hallway, and the lights are not working" is the initial question statement. This is merely an example and should not be construed as a limitation of this application. Those skilled in the art can freely set the second prompts, and freely set the concatenation order between the second prompts, multiple target level 3 work order tags, multiple priority tags, and the initial question statement and the target property knowledge statement.

[0036] Step S200: Input the second question text into the property question-and-answer model to obtain the answer text, the target third work order label, and the target work order priority label.

[0037] The property question-and-answer processing method of this application embodiment, through the above steps S110 to S200, first converts the user input information into a question vector, determines the matching target knowledge vector, then determines the matching target property knowledge statement based on the target knowledge vector, and determines the target first work order label based on the target property knowledge statement to achieve the first classification of subsequently generated work orders. It then obtains the target second-level work order label matching the target first work order label; the second-level work order label is a subcategory of the first-level work order label. The preset first prompt word, multiple target second-level work order labels, the initial question statement, and the target property knowledge statement are concatenated to obtain the first question text, which is then input into the trained property question-and-answer model to obtain the target second work order label. The first prompt word is used to instruct the property question-and-answer model to generate the target second work order label, achieving the second classification of the work order. Finally, it obtains the target third-level work order label; the target third-level work order label is a subcategory of the second work order label. The second prompt word, multiple target level 3 work order tags, multiple preset priority tags, the initial question statement, and the target property knowledge statement are concatenated to obtain the second question text. This second question text is then input into the property question-and-answer model to obtain the answer text, the target level 3 work order tag, and the target work order priority tag. The second prompt word instructs the property question-and-answer model to use the target property knowledge statement, multiple target level 3 work order tags, and multiple priority tags as context to answer the initial question statement and generate the target level 3 work order tag and the target work order priority tag. Thus, this application achieves the following: based on user input information, first, the primary category corresponding to the user input information is determined as the target level 1 work order tag; then, the target level 2 work order tag matched by the target level 1 work order tag is used as context, and the property question-and-answer model further determines the secondary category corresponding to the user input information as the target level 2 work order tag. Finally, the target level 3 work order tag matched by the target level 2 work order tag, the target property knowledge statement, and multiple priority tags are used as context to enable the property question-and-answer model to generate the answer text, the level 3 category corresponding to the user input information as the target level 3 work order tag, and the target work order priority tag. In this way, by guiding the property question-and-answer model through context, the robustness of the model is improved, the accuracy of work order classification based on user input information is achieved, the priority classification of work orders is more accurate, and the generated answer text is more accurate.

[0038] In some embodiments, the property question-answering model employs a large language model, which uses the existing Qwen-7B. While Qwen-7B can be directly used in this application, it needs to be trained to improve its applicability in the property question-answering domain. Specifically, the training steps for the property question-answering model include steps S210 to S280.

[0039] Step S210: Construct a quantitative low-rank adapter based on the property question-and-answer model; It is worth noting that the quantized low-rank adapter is specifically a weight matrix that runs parallel to the parameter matrix of the property question-answering model. This application does not elaborate on the specific construction of the quantized low-rank adapter; existing methods can be directly used for its construction.

[0040] Step S220: Obtain the training question statement and the corresponding real label question and answer results, as well as the knowledge training statement associated with the training question statement; It is worth noting that the training questions, real-world tag-based question-and-answer results, and related knowledge training statements are all obtained from historical work order processing records. Specifically, these records contain user questions, work order handlers' replies, property management knowledge statements cited in the replies, work order third-level classification tags identified by the handlers, and work order priority tags identified by the handlers. The user questions and work order handlers' replies serve as training questions. The property management knowledge statements cited in the work order handlers' replies are used as related knowledge training statements. The work order handlers' replies, their identified work order third-level classification tags, and their identified priority tags serve as real-world tag-based question-and-answer results.

[0041] Step S230: Based on the training question statement, the second prompt word, the knowledge training statement, multiple third-level work order labels, and multiple priority labels, the training input statement is constructed. Step S240: Input the training input statement into the pre-trained property question-answering model to obtain the first training question-answering result; input the training input statement into the quantization low-rank adapter to obtain the second training question-answering result; Step S250: The first training question-and-answer result and the second training question-and-answer result are concatenated to obtain the third training question-and-answer result; Step S260: Input the third training question-answering result and the real label question-answering result into the preset loss function to obtain the loss value; this application does not limit the specific loss function, and those skilled in the art can set the loss function according to the actual situation.

[0042] Step S270: Iteratively update the quantized low-rank adapter based on the loss value to obtain the trained quantized low-rank adapter; Step S280: The parameter set of the trained quantized low adapter is fused with the parameter set of the pre-trained property question answering model to obtain the trained property question answering model.

[0043] It is worth noting that, through the above steps S210 to S280, the property question-answering model is trained and its parameters are fine-tuned, making the property question-answering model suitable for the property management field. This makes the answer text, target third work order label, and target work order priority label output by the property question model more accurate, thereby improving the accuracy of the model. Furthermore, by constructing a quantized low-rank adapter, the training efficiency can be improved.

[0044] It should be noted that LoRA (Low-Rank Adaptation) is a lightweight and efficient parameter fine-tuning technique for large language models (LLMs). Proposed by the Microsoft team in 2021, it addresses the pain points of full-scale fine-tuning of large models, such as high memory consumption, high computational cost, easy overfitting, and catastrophic forgetting. By injecting an "adaptor" with a very small number of parameters into the core layer of the model through low-rank decomposition, all parameters of the main model are frozen and only the adapter is trained. Finally, the adapter parameters are fused back into the main model to complete the fine-tuning, which can improve training efficiency.

[0045] In some embodiments, refer to Figure 2 , Figure 2 This is a schematic diagram of a process before step S130 in an embodiment of this application. Figure 2 The illustrated steps include, but are not limited to, steps S310 and S320.

[0046] Step S310: Obtain public knowledge files and construct a public vector library based on the public knowledge files; Step S320: Obtain the property knowledge file for each community unit, and construct a community vector library based on the property knowledge file; and the community vector library corresponds one-to-one with the community unit.

[0047] It is worth noting that public knowledge documents are usually legal documents applicable to all community units. Property knowledge documents, on the other hand, can be property management regulations. These regulations may differ between different community units. Therefore, it is necessary to construct a separate community vector library for each community's property knowledge document, ensuring a one-to-one correspondence between the library and the community unit. This allows the property question-and-answer model to select the community vector library corresponding to the user's community unit as the target vector library when answering questions, further improving the accuracy of the model.

[0048] In some embodiments, step S310 specifically includes steps S311 and S312.

[0049] Step S311: Divide the public knowledge document into multiple property knowledge statements; Step S312: Convert property knowledge statements into knowledge vectors, and build a public vector library based on the knowledge vectors.

[0050] It is worth noting that by inputting property knowledge statements into an existing text vector conversion model, the corresponding knowledge vectors can be obtained. For example, the text vector conversion model is a trained BGE-M3 model.

[0051] In some embodiments, refer to Figure 3 , Figure 3 This is a schematic diagram illustrating the process of determining the target vector library in an embodiment of this application. Figure 3 The illustrated process includes, but is not limited to, steps S410 to S440.

[0052] Step S410: Obtain user identity information based on the property request message; Step S420: Determine the target cell unit based on the user's identity information; Step S430: Determine the target cell vector library that matches the target cell cell from multiple cell vector libraries; Step S440: Use the target cell vector library and the public vector library as the target vector library.

[0053] It is worth noting that public knowledge documents are usually legal documents that apply to all community units. Property knowledge documents can be property management regulations, which may differ between different community units. Therefore, by determining the target community vector library and public vector library corresponding to the user's identity information as the target vector library through steps S410 to S440, the accuracy of the property question-answering model can be improved.

[0054] In some embodiments, steps S510 to S530 are included after step S200.

[0055] Step S510: Send the answer text to the user terminal; Step S520: Based on the initial question statement, answer text, target third work order label, and target work order priority label, construct the target work order; Step S530: Send the target work order to the terminal associated with the target third work order tag.

[0056] It is worth noting that after step S200, steps S510 to S530 send the answer text to the user terminal to reply to the customer. Simultaneously, a target work order is constructed and sent to the terminal associated with the target third work order tag, thus notifying the holder of the terminal associated with the target third work order tag to process the work order. This achieves the simultaneous answering of customer questions and direct notification to the holder of the terminal associated with the target third work order tag, improving work order response speed, and reminding the holder to process according to priority through the target work order priority tag.

[0057] In some embodiments, an initial question statement is generated based on user input information, including: If the user input information includes both text and image, the image part is input into a preset visual language model to obtain an image description; The text portion is combined with the image description to obtain the initial question statement.

[0058] It is worth noting that the visual language model can employ the existing trained Qwen-VL model. This application obtains image descriptions through a visual language model, then concatenates the text portion with the image descriptions to obtain the initial question statement, thus realizing multimodal input for the property management question-and-answer processing method.

[0059] A second aspect of this application provides a property inquiry processing device. (See also...) Figure 4 , Figure 4 This is a schematic diagram of the property inquiry processing device according to an embodiment of this application. The property inquiry processing device includes: The first acquisition module 410 is used to acquire property request messages and acquire user input information based on the property request messages. The conversion module 420 is used to generate an initial question statement based on user input information and convert the initial question statement into a question vector; The first determining module 430 is used to determine at least one target knowledge vector that matches the question vector from the target vector library based on the question vector; the target vector library includes multiple knowledge vectors converted from property knowledge statements; The second determining module 440 is used to determine the target property knowledge statement corresponding to the target knowledge vector based on at least one target knowledge vector. The third determination module 450 is used to determine the target first work order label from multiple first-level work order labels based on the target property knowledge statement; The second acquisition module 460 is used to acquire multiple target secondary work order tags that match the target first work order tag; The first splicing module 470 is used to splice the preset first prompt word, multiple target secondary work order labels, initial question statement and target property knowledge statement to obtain the first question text, and input the first question text into the trained property question answering model to obtain the target second work order label; The third acquisition module 480 is used to acquire multiple target third-level work order tags that match the second work order tag; The second splicing module 490 is used to splice the preset second prompt word, multiple target level 3 work order tags, multiple preset priority tags, initial question statement and target property knowledge statement to obtain the second question text; Input module 500 is used to input the second question text into the property question-and-answer model to obtain the answer text, the target third work order label, and the target work order priority label.

[0060] The property question-and-answer processing device of this application embodiment is used to execute the property question-and-answer processing method of the first aspect embodiment. When executing the method, the user input information is first converted into a question vector, a matching target knowledge vector is determined, and then a matching target property knowledge statement is determined based on the target knowledge vector. A target first work order label is determined based on the target property knowledge statement to achieve the first classification of subsequently generated work orders. A target second-level work order label matching the target first work order label is obtained; the second-level work order label is a subcategory of the first-level work order label. A preset first prompt word, multiple target second-level work order labels, the initial question statement, and the target property knowledge statement are concatenated to obtain a first question text. This first question text is then input into a trained property question-and-answer model to obtain a target second work order label. The first prompt word is used to instruct the property question-and-answer model to generate the target second work order label, achieving the second classification of the work order. A target third-level work order label is obtained; the target third-level work order label is a subcategory of the second work order label. The second prompt word, multiple target level 3 work order tags, multiple preset priority tags, the initial question statement, and the target property knowledge statement are concatenated to obtain the second question text. This second question text is then input into the property question-and-answer model to obtain the answer text, the target level 3 work order tag, and the target work order priority tag. The second prompt word instructs the property question-and-answer model to use the target property knowledge statement, multiple target level 3 work order tags, and multiple priority tags as context to answer the initial question statement and generate the target level 3 work order tag and the target work order priority tag. Thus, this application achieves the following: based on user input information, first, the primary category corresponding to the user input information is determined as the target level 1 work order tag; then, the target level 2 work order tag matched by the target level 1 work order tag is used as context, and the property question-and-answer model further determines the secondary category corresponding to the user input information as the target level 2 work order tag. Finally, the target level 3 work order tag matched by the target level 2 work order tag, the target property knowledge statement, and multiple priority tags are used as context to enable the property question-and-answer model to generate the answer text, the level 3 category corresponding to the user input information as the target level 3 work order tag, and the target work order priority tag. In this way, by guiding the property question-and-answer model through context, the robustness of the model is improved, the accuracy of work order classification based on user input information is achieved, the priority classification of work orders is more accurate, and the generated answer text is more accurate.

[0061] It should be noted that the specific implementation of this property management question and answer processing device is basically the same as the specific embodiment of the property management question and answer processing method described above, and will not be repeated here. Subject to meeting the requirements of the embodiments of this application, the property management question and answer processing device may also be equipped with other functional units to implement the property management question and answer processing method in the above embodiments.

[0062] A third aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the property inquiry-and-answer processing method of any of the first aspects of the embodiment. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0063] Reference Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device according to one embodiment. The electronic device includes: The processor 501 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 502 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 502 and is called and executed by the processor 501 using the property inquiry processing method of the embodiments of this application. The input / output interface 503 is used to implement information input and output; The communication interface 504 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 505 transmits information between various components of the device (e.g., processor 501, memory 502, input / output interface 503, and communication interface 504); The processor 501, memory 502, input / output interface 503, and communication interface 504 are connected to each other within the device via bus 505.

[0064] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the property inquiry and answering method of any one of the first aspects of this application.

[0065] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0066] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0067] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0068] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0069] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0070] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0071] It should be understood that in this application, "at least one (item)" means one or more, and "more than one" means two or more. "And / or" is used to describe the mapping relationship between the mapped objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following mapped objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0072] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of the units described above 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.

[0073] The units described above 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 according to actual needs.

[0074] 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.

[0075] If the integrated unit is implemented as 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 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 multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0076] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A property management Q&A processing method, characterized in that, include: Obtain a property management request message, and obtain user input information based on the property management request message; Based on the user input information, an initial question statement is generated, and the initial question statement is converted into a question vector; Based on the question vector, at least one target knowledge vector matching the question vector is determined from the target vector library; the target vector library includes multiple knowledge vectors converted from property knowledge statements. Based on at least one of the target knowledge vectors, determine the target property knowledge statements corresponding to the target knowledge vectors; Based on the target property knowledge statement, the target first work order label is determined from multiple first-level work order labels; Obtain multiple target secondary work order tags that match the target first work order tag; The preset first prompt word, multiple target secondary work order tags, the initial question statement, and the target property knowledge statement are concatenated to obtain the first question text. The first question text is then input into the trained property question-and-answer model to obtain the target second work order tag. Obtain multiple target third-level work order tags that match the second work order tag; The second question text is obtained by concatenating the preset second prompt word, multiple target level 3 work order tags, multiple preset priority tags, the initial question statement, and the target property knowledge statement. The second question text is input into the property question-and-answer model to obtain the answer text, the target third work order label, and the target work order priority label.

2. The property inquiry processing method according to claim 1, characterized in that, Before determining at least one target knowledge vector matching the question vector from the target vector library based on the question vector, the process includes: Obtain public knowledge files and construct a public vector library based on the public knowledge files; Obtain the property knowledge file for each community unit, and construct a community vector library based on the property knowledge file; and the community vector library corresponds one-to-one with the community unit.

3. The property inquiry processing method according to claim 2, characterized in that, The target vector library is determined through the following steps: Obtain user identity information based on the property request message; The target cell unit is determined based on the user identity information; From the multiple cell vector libraries, determine the target cell vector library that matches the target cell cell; The target cell vector library and the public vector library are used as the target vector library.

4. The property inquiry processing method according to claim 1, characterized in that, The step of generating an initial question statement based on the user input information includes: If the user input information is detected to include both text and image components, the image component is input into a preset visual language model to obtain an image description. The text portion is concatenated with the image description to obtain the initial question statement.

5. The property inquiry processing method according to claim 1, characterized in that, After inputting the second question text into the property question-and-answer model to obtain the answer text, the target third work order label, and the target work order priority label, the process includes: The answer text is sent to the user terminal; Based on the initial question, the answer text, the target third work order tag, and the target work order priority tag, a target work order is constructed. Send the target work order to the terminal associated with the target third work order tag.

6. The property inquiry processing method according to claim 1, characterized in that, Each of the aforementioned primary work order tags corresponds to a keyword corpus; The step of determining the target first work order tag from multiple first-level work order tags based on the target property knowledge statement includes: The target property knowledge statement is matched with the keyword corpus corresponding to the multiple first-level work order tags to obtain multiple matching degree values; The target word corpus with the highest matching degree value is determined from multiple keyword corpora, and the first-level work order tag corresponding to the target word corpus is used as the target first work order tag.

7. The property inquiry processing method according to claim 1, characterized in that, The construction of a public vector library based on the public knowledge files includes: The public knowledge document is divided into multiple property knowledge statements; The property knowledge statements are converted into knowledge vectors, and the public vector library is constructed based on the knowledge vectors.

8. A property inquiry processing device, characterized in that, include: The first acquisition module is used to acquire a property request message and acquire user input information based on the property request message. The conversion module is used to generate an initial question statement based on the user input information and convert the initial question statement into a question vector; The first determining module is used to determine at least one target knowledge vector that matches the question vector from a target vector library based on the question vector; the target vector library includes multiple knowledge vectors converted from property knowledge statements. The second determining module is used to determine the target property knowledge statement corresponding to the target knowledge vector based on at least one of the target knowledge vectors; The third determination module is used to determine the target first work order label from multiple first-level work order labels based on the target property knowledge statement. The second acquisition module is used to acquire multiple target secondary work order tags that match the target first work order tag; The first splicing module is used to splice the preset first prompt word, multiple target secondary work order tags, the initial question statement and the target property knowledge statement to obtain the first question text, and input the first question text into the trained property question and answer model to obtain the target second work order tag; The third acquisition module is used to acquire multiple target third-level work order tags that match the second work order tag; The second splicing module is used to splice the preset second prompt word, multiple target level 3 work order tags, multiple preset priority tags, the initial question statement, and the target property knowledge statement to obtain the second question text; The input module is used to input the second question text into the property question-and-answer model to obtain the answer text, the target third work order label, and the target work order priority label.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the property question-and-answer processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the property question-and-answer processing method according to any one of claims 1 to 7.

Images