Visual component interaction method and system based on retail industry template knowledge base

Abstract

The application relates to a visual component interaction method and system based on a retail industry template knowledge base, which uses a large language knowledge base model to connect user needs, automatically performs demand understanding and key indicator extraction matching, and analyzes charts through an industry knowledge base. Users only need to describe the analysis scene or target through natural language, and the system automatically generates chart results while also providing more analysis perspective suggestions and complete analysis report templates, which are more intelligent and automated, helping users quickly discover hidden information and rules in data. At the same time, users can also use interactive controls or natural language to change the corresponding charts to quickly explore data. In the process of generating new demands for business parties, data analysts can generate examples on the spot through system conversation to avoid subsequent ambiguity and rework rate below 10%, thereby improving the overall business analysis efficiency of enterprises.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of retail industry, and in particular to a retail industry template knowledge base construction method, a retail industry template knowledge base, a visual component interaction method and system based on the retail industry template knowledge base, and an electronic device. BACKGROUND

[0002] The data analysis exploration system can be used for business data analysis of a business, and provides an analysis report of business data for the business, so as to facilitate the business to analyze its business behavior and customer behavior and to adjust the business.

[0003] However, in the process of using the traditional commercial data analysis, there is still a lack of effective temporary demand response and exploration analysis function. This is because the traditional commercial data analysis tool only provides data processing and chart functions, and the specific analysis idea still depends on the manual experience of an analyst to design and build. Therefore, this process requires about 50%-60% of the working time of a data analyst, including demand communication, text confirmation, board making, conclusion synchronization, and additional demands or possible rework. Moreover, the rework caused by the demand may account for more than 80% of the time for processing the demand. Therefore, the analysis process cycle is relatively long, and the "ineffective work" generated during the cycle gradually reduces the overall project efficiency. SUMMARY

[0004] To solve the above problems, the present application provides a retail industry template knowledge base construction method, a retail industry template knowledge base, a visual component interaction method and system based on the retail industry template knowledge base, and an electronic device.

[0005] In one aspect of the present application, a retail industry template knowledge base construction method is provided, which includes the following steps:

[0006] Collecting original knowledge records of a business;

[0007] Text extraction is performed on the original knowledge records, and the extracted text is segmented and fragmented by a segmenter;

[0008] Word embedding is performed on the segmented content to obtain a vectorized representation of the vector;

[0009] The vector and the corresponding original knowledge record are stored in a vector database in the form of key-value to obtain a retail industry template knowledge base.

[0010] In another aspect of the present application, a retail industry template knowledge base is provided, which is obtained by using the retail industry template knowledge base construction method.

[0011] Another aspect of the present application also provides a visual component interaction method based on a retail industry template knowledge base, comprising the following steps:

[0012] Collecting a user input business consultation question and sending it to an agent module;

[0013] The agent module disassembles the question, extracts key words, and matches the extracted key words with fields in a preset data set;

[0014] After a successful match, a business index search is initiated to a business knowledge base, and the business knowledge base returns corresponding knowledge to the agent module after the search;

[0015] The agent module integrates the question and the corresponding knowledge, sends the integrated question and the corresponding knowledge to a generation module, and obtains parameters for generating a chart from the generation module;

[0016] The generation module generates the parameters according to the integrated question and the corresponding knowledge, returns the parameters to the agent module, and generates a chart corresponding to the question according to the parameters and sends the chart to the user for confirmation;

[0017] After the user confirms, the agent module stores the chart in the template knowledge base in a word segmentation vectorization manner for subsequent conversation query.

[0018] As an optional embodiment of the present application, when the extracted key words are matched with the fields in the preset data set, it also includes:

[0019] If the key words do not match the fields in the preset data set at all, the matching result is fed back to the user, and the user input business consultation question is reminded that it is not supported.

[0020] As an optional embodiment of the present application, before the agent module sends the chart to the user for confirmation, it also includes:

[0021] The agent module sends the integrated question and the corresponding knowledge to the template knowledge base;

[0022] In the template knowledge base, similar scene retrieval is performed to obtain corresponding similar scenes, and the obtained similar scenes are returned to the agent module;

[0023] The agent module supplements example information in the chart according to the similar scenes, and after the supplement, the chart is sent to the user for secondary confirmation.

[0024] As an optional embodiment of the present application, after the user performs secondary confirmation, it also includes:

[0025] The agent module integrates a graph example of the secondary confirmation of the user, and sends the integrated graph example to the generation module to obtain parameters for regenerating a graph from the generation module;

[0026] The generation module regenerates the parameters according to the integrated graph example and returns to the agent module, and the agent module regenerates a graph corresponding to the problem according to the parameters and sends the graph to the user for secondary confirmation.

[0027] As an optional embodiment of the application, after the secondary confirmation of the user, the method further includes:

[0028] The agent module stores the graph confirmed by the user in the template knowledge base through word segmentation and vectorization, and returns the graph to the front end for display.

[0029] Another aspect of the application also provides a system for implementing the visual component interaction method based on the retail industry template knowledge base, which includes:

[0030] The agent module is configured to collect a business consultation question input by a user, disassemble the question, extract key words, match the extracted key words with fields in a preset data set, and initiate a business index search to a business knowledge base after a successful match.

[0031] The business knowledge base is configured to search for knowledge corresponding to the question after the business index search, and return the knowledge to the agent module.

[0032] The agent module is further configured to integrate the question and the corresponding knowledge, and send the integrated question and the corresponding knowledge to the generation module to obtain parameters for generating a graph.

[0033] The generation module is configured to generate the parameters according to the integrated question and the corresponding knowledge, and return the parameters to the agent module.

[0034] The agent module is further configured to generate a graph corresponding to the question according to the parameters, and send the graph to the user for confirmation. After the user confirms, the agent module stores the graph in the template knowledge base through word segmentation and vectorization, for subsequent session query.

[0035] Another aspect of the application also provides an electronic device, which includes:

[0036] A processor;

[0037] A memory for storing processor-executable instructions;

[0038] When the processor is configured to execute the executable instructions, the method for visual component interaction based on the retail industry template knowledge base is implemented.

[0039] Technical effects of the present application:

[0040] The present application uses a large language knowledge base model to match user needs by combining retail industry knowledge base corpus, automatically performs demand understanding and key indicator extraction, and generates analysis chart results and related data insight suggestions through industry knowledge base.

[0041] Users can only describe the analysis scene or target through natural language, and the system will also give more analysis perspective suggestions and complete analysis report templates while automatically generating chart results. Compared with traditional business data analysis tools (which only provide data processing and chart functions, and specific analysis ideas still rely on analysts' manual experience to design and build), the system has more intelligent and automated features, helping users quickly discover hidden information and rules in data. At the same time, users can also use interactive controls or natural language to change the corresponding charts to quickly explore data.

[0042] Using the present solution, data analysts can generate examples on the spot through system conversation during the process of generating new needs from business parties, avoiding subsequent ambiguity and reducing rework rate to less than 10%. Even business personnel can quickly generate the required analysis charts through conversation, speeding up the entire analysis link. Trivial one-time needs can also be answered through conversation during meetings or conversations, without the need for additional time to explore. In addition, if there are already existing analysis scenes, they can be quickly retrieved, so there is no need to reinvent the wheel.

[0043] Other features and aspects of the present application will become apparent from the following detailed description of exemplary embodiments with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate exemplary embodiments, features, and aspects of the present application, and serve to explain the principles of the present application.

[0045] Figure 1 An implementation flowchart schematic diagram showing the construction method of the retail industry template knowledge base of the present application;

[0046] Figure 2 A timing diagram showing the visual component interaction method based on the retail industry template knowledge base of the present application;

[0047] Figure 3 An application schematic diagram of the electronic device of the present application. DETAILED DESCRIPTION

[0048] Various exemplary embodiments, features, and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numbers in different drawings represent the same or similar elements. Although various aspects of embodiments are illustrated in the drawings, the drawings are not necessarily drawn to scale unless specifically noted.

[0049] The term "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any implementation described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other implementations.

[0050] In addition, for the purpose of convenience and brevity, detailed descriptions of well-known functions and structures incorporated in the application are omitted. It will be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described herein, embody the principles of the application and are included within its spirit and scope.

[0051] Embodiment 1

[0052] As shown in the following, in one aspect of the present application, a method for constructing a retail industry template knowledge base is provided, comprising the following steps: Figure 1

[0053] Collecting original knowledge records of the business;

[0054] Text extraction is performed on the original knowledge records, and the extracted text is segmented and sliced by a word segmenter;

[0055] Word embedding is performed on the segmented content to obtain a vectorized representation of the segmented words;

[0056] The vector and the corresponding original knowledge record are stored in a vector database in the form of key-value, and a retail industry template knowledge base is obtained.

[0057] The present solution first generates a large language knowledge base model in combination with the knowledge base corpus of the retail industry.

[0058] Subsequently, the large language knowledge base model is used to interface user needs, automatically perform demand understanding and key indicator extraction matching, and generate analysis chart results and related data insight suggestions through the industry knowledge base.

[0059] Industry general knowledge is extracted to form a business template knowledge base (see the template knowledge base in Figure 2 for details, which only stores business data sets and business indicator retrieval in the business knowledge base.

[0060] ​The retail industry analysis template is summarized by precipitation, and various analysis charts and reports are created by business personnel in the BI platform. These charts, reports and templates reflect the understanding and analysis ideas of business personnel on enterprise data, and are stored in the system in the form of metadata. Metadata refers to some basic information of the chart, such as: chart name, chart involved field, business field meaning and calculation method, etc.

[0061] The construction of the business template knowledge base is to store the metadata in a vector form.

[0062] Specifically includes the following steps:

[0063] Step 1: Collect platform metadata and general metadata of Guanyuan solution (original business knowledge record, including the questions and answers between customers and merchants), which mainly includes business index name, index calculation logic description and index calculation example;

[0064] Step 2: Extract metadata in text, and perform tokenization and fragmentation by tokenizer. This can split the original text knowledge into multiple relatively independent knowledge points with certain length and relationship;

[0065] Step 3: Perform embedding processing on the fragmented content to obtain a vectorized representation. In this way, when the user performs similar scene analysis, similarity search can be performed, and BM25 algorithm is also used in the search. This is an algorithm for evaluating the relevance of search words and documents, which can be understood as keyword search, because in addition to similarity matching, some core business indicators also need to be accurately matched;

[0066] Step 4: Store the vectors after embedding and the original knowledge in the vector database in the form of key-value.

[0067] The business template knowledge base formed above, when business personnel make analysis and decision, through the demand description system, automatically matches the corresponding business background knowledge from the vector database, generates the corresponding chart report, and matches the similar analysis chart from the template knowledge base as search recommendation, to speed up the decision landing.

[0068] Embodiment 2

[0069] In another aspect of the present application, a retail industry template knowledge base is proposed, which is obtained by using the construction method of the retail industry template knowledge base in embodiment 1.

[0070] For details, see the description of embodiment 1.

[0071] Embodiment 3

[0072] Based on the principle of embodiment 1, such as Figure 2As shown, this application also proposes a visual component interaction method based on a retail industry template knowledge base, including the following steps:

[0073] Collect user-inputted business inquiries and send them to the agent module;

[0074] The proxy module breaks down the problem, extracts key words, and matches the extracted key words with fields in a preset dataset;

[0075] After a successful match, a business metric search is initiated in the business knowledge base. After the business knowledge base is searched, the knowledge corresponding to the question is returned to the agent module.

[0076] The proxy module integrates the issues and corresponding knowledge, and then sends the integrated information to the generation module to obtain the parameters for generating the chart.

[0077] The generation module generates the parameters based on the integrated question and corresponding knowledge and returns them to the proxy module. The proxy module then generates a chart corresponding to the question based on the parameters and sends the chart to the user for confirmation.

[0078] After the user confirms, the proxy module stores the chart in the template knowledge base by word segmentation and vectorization for subsequent session queries.

[0079] like Figure 2 As shown, the system mainly consists of a preprocessing layer, a chart generation layer, and a chart retrieval layer.

[0080] When a user inputs their requirements via natural language, the first step involves a "preprocessing layer" that breaks down the requirement and extracts and matches key metrics. Then, relevant business knowledge is retrieved from the platform's business template knowledge base, and the question and knowledge are integrated. The second step moves to the "chart generation layer," which uses the platform's private domain computing engine to quickly generate the data charts the user needs. This step provides an immediate and accurate response to the user's interactive needs. Simultaneously, the third step, through a "chart retrieval layer," uses concurrent tasks to perform similarity matching of related analysis scenarios within the template knowledge base, thereby recommending more analytical perspective charts and reports to achieve proactive exploration.

[0081] In the preprocessing layer, assuming a user inputs the question, "What are the monthly numbers of new and returning customers for each channel this year?", the question enters the agent module, where the agent extracts key words: "channel," "new customer," and "returning customer." The agent matches the extracted words with fields in the dataset. If the data fields do not contain "new customer" or "returning customer," the agent initiates a query to the business template knowledge base module to obtain their definitions, and then proceeds to the "chart generation layer."

[0082] In the chart generation layer, when the business template knowledge base is matched, the agent module integrates the question and knowledge, and sends the integrated content to the generation module for chart parameter generation. The generated parameters are confirmed again by the agent, the JSON format is uniformly modified, and the result is returned to the front end. At this time, the related chart for the user question is displayed. If the user confirms the feedback in the front end, the generated chart is vectorized through the "knowledge base generation path" and stored in the template knowledge base for subsequent session query.

[0083] As an optional embodiment of the present application, when the extracted key words are matched with the fields in the preset data set, the following steps are further included:

[0084] If the key words do not match the fields in the preset data set at all, the matching result is fed back to the user, and it is reminded that the user's input business consultation question is not supported.

[0085] If all the above key words do not match, the front end is directly fed back that the question is not supported, which means that the user's question is not related to the data set of the current work area. This step is mainly to quickly connect the user's cognition in the early stage of interaction and establish a connection.

[0086] Enter the chart retrieval layer.

[0087] As an optional embodiment of the present application, before the agent module sends the chart to the user for confirmation, the following steps are further included:

[0088] The agent module sends the integrated question and corresponding knowledge to the template knowledge base;

[0089] In the template knowledge base, similar scene retrieval is performed to obtain a corresponding similar scene, and the obtained similar scene is returned to the agent module;

[0090] The agent module supplements the example information in the chart according to the similar scene, and then sends the chart to the user for secondary confirmation.

[0091] After the agent module integrates the question and knowledge, a concurrent task will deliver the integrated content to the template knowledge base for retrieval to obtain a similar analysis scene, and then return to the agent module for chart example information supplement, and then transmit to the front end for user selection.

[0092] As an optional embodiment of the present application, after the user performs secondary confirmation, the following steps are further included:

[0093] The agent module integrates the graphical examples of the secondary confirmation of the user, and sends the integrated graphical examples to the generation module to obtain the parameters for regenerating the chart from the generation module;

[0094] The generation module regenerates the parameters according to the integrated graphical examples and returns to the agent module, and the agent module regenerates the chart corresponding to the problem according to the parameters and sends the chart to the user for secondary confirmation.

[0095] The above-mentioned way of reprocessing parameters and regenerating charts can refer to the process of chart generation.

[0096] As an optional embodiment of the present application, after the secondary confirmation of the user, it further includes:

[0097] The agent module stores the chart of the secondary confirmation of the user in the template knowledge base through the word segmentation vectorization method, and returns the chart to the front end for display.

[0098] If the user selects some of the graphical examples, it will enter the generation module to generate related parameters, similar to the process of "chart generation", and finally return the generated content to the front end for result display.

[0099] The system interaction is carried out in the form of natural language dialogue, so the system will store the dialogue state, and there is no need to worry about the disappearance of historical dialogue records due to exit or page refresh. The data exploration can continue at any time following the last analysis idea. In addition, during the subsequent interactive dialogue between the customer and the merchant, if it is desired to reproduce the analysis results of the previous stage during the analysis process, a one-key backtracking function can be used to dispatch the previous analysis chart again.

[0100] Therefore, using the present solution, the data analyst can generate examples on the spot through system conversation during the process of generating new requirements by the business party, avoiding subsequent ambiguity and reducing the rework rate to less than 10%. Even the business personnel can directly generate the required analysis chart through conversation, speeding up the entire analysis link. Trivial and one-time requirements can also be answered through conversation during meetings or conversations, without the need for additional time to explore. At the same time, if there are already existing analysis scenarios, they can be quickly retrieved, so there is no need to repeatedly reinvent the wheel.

[0101] Obviously, those skilled in the art should understand that all or part of the processes in the above embodiments can be instructed by a computer program to relevant hardware, and the program can be stored in a computer readable storage medium, and the program can include the processes of the above controlled embodiments when executed. Those skilled in the art can understand that all or part of the processes in the above embodiments can be instructed by a computer program to relevant hardware, and the program can be stored in a computer readable storage medium, and the program can include the processes of the above controlled embodiments when executed. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid state drive (SSD), etc. The storage medium can also include a combination of the above types of memories.

[0102] Embodiment 4

[0103] Based on the implementation principle of embodiment 1, another aspect of the present application also provides a system for implementing the visual component interaction method based on the retail industry template knowledge base, comprising:

[0104] The agent module is configured to collect a user input business consultation question, disassemble the question, extract key words, match the extracted key words with fields in a preset data set, and initiate a business index search to the business knowledge base after a successful match.

[0105] The business knowledge base is configured to search for business indexes and return corresponding knowledge to the agent module after the search.

[0106] The agent module is further configured to integrate the question and the corresponding knowledge, send the integrated question and the corresponding knowledge to the generation module, and obtain parameters for generating a chart from the generation module.

[0107] The generation module is configured to generate the parameters based on the integrated question and the corresponding knowledge and return the parameters to the agent module.

[0108] The agent module is further configured to generate a chart corresponding to the question based on the parameters, send the chart to the user for confirmation, and store the chart in the template knowledge base by word segmentation vectorization after the user confirms the chart, for subsequent session query.

[0109] The functions and interactions of the above modules are described in detail in embodiment 3.

[0110] The modules or steps of the application described above can be implemented by a general computing system, which can be centralized on a single computing system or distributed on a network composed of multiple computing systems. Alternatively, the modules or steps can be implemented by program codes executable by a computing system, so that they can be stored in a storage system and executed by a computing system, or they can be respectively manufactured into individual integrated circuit modules, or multiple modules or steps among them can be manufactured into a single integrated circuit module. Thus, the application is not limited to any specific combination of hardware and software.

[0111] Embodiment 5

[0112] As Figure 3 Further, another aspect of the present application also provides an electronic device, comprising:

[0113] a processor;

[0114] a memory for storing processor-executable instructions;

[0115] wherein the processor is configured to implement the method for visual component interaction based on a retail industry template knowledge base when executing the executable instructions.

[0116] The electronic device according to the embodiment of the present application comprises a processor and a memory for storing processor-executable instructions. Wherein the processor is configured to implement the method for visual component interaction based on a retail industry template knowledge base when executing the executable instructions.

[0117] It should be pointed out here that the number of processors can be one or more. Meanwhile, the electronic device according to the embodiment of the present application can also comprise an input system and an output system. Wherein the processor, the memory, the input system and the output system can be connected through a bus or other means, which is not limited here.

[0118] The memory as a computer readable storage medium can be used to store software programs, computer executable programs and various modules, such as the programs or modules corresponding to the method for visual component interaction based on a retail industry template knowledge base according to the embodiment of the present application. The processor executes the software programs or modules stored in the memory, thereby performing various functional applications and data processing of the electronic device.

[0119] The input system can be used to receive input numbers or signals. Wherein the signals can be key signals related to the user settings and function control of the device / terminal / server. The output system can include display devices such as display screens.

[0120] Having described various embodiments of the application, it is to be understood that the above description is meant not to limit and not to encompass all of the possible embodiments. Many modifications and variations of this application can be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. It is intended that the scope of the application be defined by the scope of the patent and by the claims as allowed by the patent office, which can include adaptations, modifications, and variations that fall within the scope of the patent. The terms used herein are to be given their ordinary and customary meaning to a person of ordinary skill in the art and are not to be limited to a specialized meaning that can be intended in some cases by the original inventors or any successors or assignees.

Claims

1. A method for visual component interaction based on a retail industry template knowledge base, characterized by, The retail industry template knowledge base is constructed by the following method steps: Collecting original knowledge records of the business; Text extraction is performed on the original knowledge records, and the extracted text is segmented by a segmenter; The segmented text is subjected to word embedding to obtain a vectorized representation of the segmented text; The vector and the corresponding original knowledge record are stored in a vector database in the form of key-value, obtaining a retail industry template knowledge base; The visual component interaction method includes the following steps: Collecting user input business consultation questions and sending them to the agent module; The agent module disassembles the question and extracts the key segments, matches the extracted key segments with the fields in the preset data set; After successful matching, the agent module initiates a business index search to the business knowledge base, and the business knowledge base returns the corresponding knowledge to the agent module after searching; The agent module integrates the question and the corresponding knowledge, and sends the integrated question and the corresponding knowledge to the generation module to obtain parameters for generating a chart; The generation module generates the parameters according to the integrated question and the corresponding knowledge and returns them to the agent module, and the agent module generates a chart corresponding to the question according to the parameters and sends the chart to the user for confirmation; After the user confirms, the agent module stores the chart in the template knowledge base through segmentation vectorization for subsequent session query.

2. The method of claim 1, wherein, When matching the extracted key segments with the fields in the preset data set, it also includes: If the key segments do not match the fields in the preset data set at all, the matching result is fed back to the user, and the user input business consultation question is reminded that it is not supported.

3. The method of claim 1, wherein the method further comprises: Before the agent module sends the chart to the user for confirmation, it also includes: The agent module sends the integrated question and the corresponding knowledge to the template knowledge base; In the template knowledge base, similar scenarios are searched to obtain corresponding similar scenarios, and the obtained similar scenarios are returned to the agent module; The agent module supplements the example information in the chart according to the similar scenarios, and then sends the chart to the user for secondary confirmation.

4. The method of claim 3, wherein the method further comprises: After the user performs secondary confirmation, it also includes: The agent module integrates the chart examples confirmed by the user, and sends the integrated chart examples to the generation module to obtain parameters for regenerating the chart; The generation module regenerates the parameters according to the integrated chart examples and returns them to the agent module, and the agent module regenerates the chart corresponding to the question according to the parameters and sends the chart to the user for secondary confirmation.

5. The method of claim 4, wherein, After the user performs secondary confirmation, it also includes: The agent module stores the chart confirmed by the user in the template knowledge base through segmentation vectorization, and returns the chart to the front end for display.

6. An apparatus for implementing the method of any one of claims 1-5 for visual component interaction based on a retail industry template knowledge base, characterized by ​ An agent module is configured to collect a user-input business consultation question, disassemble the question, extract key words, and match the extracted key words with fields in a preset data set; after a successful match, the agent module initiates a business index search in a business knowledge base; The business knowledge base is configured to search for knowledge corresponding to the question and return the knowledge to the agent module; The agent module is further configured to integrate the question and the corresponding knowledge, send the integrated question and the corresponding knowledge to a generation module, and obtain parameters for generating a chart from the generation module; The generation module is configured to generate the parameters based on the integrated question and the corresponding knowledge and return the parameters to the agent module; The agent module is further configured to generate a chart corresponding to the question based on the parameters, send the chart to a user for confirmation, and store the chart in the template knowledge base in a word vectorization manner after the user confirms the chart, so as to be used for subsequent conversation queries.

7. An electronic device, comprising: Comprise: A processor; A memory for storing processor-executable instructions; When the processor is configured to execute the executable instructions, the method for visual component interaction based on the retail industry template knowledge base of any one of claims 1-5 is implemented.

8. A storage medium, characterized by The storage medium stores a computer executable program, and the computer executable program implements the process of the method for visual component interaction based on the retail industry template knowledge base of any one of claims 1-5 when executed.

Images