Data screening method and device, electronic equipment and computer readable storage medium

Abstract

The invention provides a data screening method and device based on a large model, electronic equipment and a computer readable storage medium, and relates to the technical field of data processing, in particular to the technical fields of artificial intelligence, commodity recommendation and the like. According to the specific implementation scheme, the method comprises the steps of obtaining user interaction input, analyzing the user interaction input based on a pre-trained large language model, and generating a commodity screening condition; according to the commodity screening condition, obtaining commodity information conforming to the commodity screening condition from an e-commerce platform through an application programming interface of the e-commerce platform; and feeding back the commodity information so that a user can obtain the commodity information through interaction.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technologies, and particularly to technologies such as artificial intelligence and product recommendation. Specifically, this disclosure relates to a data screening method and apparatus, an electronic device, and a computer-readable storage medium based on a large model. Background Art

[0002] With the development of Internet technologies, e-commerce has developed rapidly due to its advantages.

[0003] When a user searches for a certain product on an e-commerce platform, the e-commerce platform can match the keywords of the product description entered by the user with the database inside the e-commerce platform.

[0004] On the basis of keyword search, the user can also further filter and refine by manually selecting conditions such as brand, price range, color, etc. to obtain products that meet their requirements on the e-commerce platform. Summary of the Invention

[0005] This disclosure provides a data screening method and apparatus, an electronic device, and a computer-readable storage medium based on a large model.

[0006] According to the first aspect of this disclosure, a data screening method based on a large model is provided. The method includes:

[0007] Obtain user interaction input, and parse the user interaction input based on a pre-trained large language model to generate product screening conditions;

[0008] According to the product screening conditions, obtain product information that meets the product screening conditions from the e-commerce platform through the application programming interface of the e-commerce platform;

[0009] Feedback the product information for the user to obtain the product information through interaction.

[0010] According to the second aspect of this disclosure, a data screening apparatus based on a large model is provided. The apparatus includes:

[0011] A condition module, configured to obtain user interaction input, and parse the user interaction input based on a pre-trained large language model to generate product screening conditions;

[0012] A screening module, configured to obtain product information that meets the product screening conditions from the e-commerce platform through the application programming interface of the e-commerce platform according to the product screening conditions;

[0013] A feedback module, configured to feedback the product information for the user to obtain the product information through interaction.

[0014] According to a third aspect of the present disclosure, there is provided an electronic device, which includes:

[0015] at least one processor; and

[0016] a memory communicatively connected to the at least one processor; wherein,

[0017] the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the above-mentioned data screening method based on a large model.

[0018] According to a fourth aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to execute the above-mentioned data screening method based on a large model.

[0019] According to a fifth aspect of the present disclosure, there is provided a computer program product including a computer program, which implements the above-mentioned data screening method based on a large model when executed by a processor.

[0020] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings are used to better understand the solution and do not constitute a limitation to the present disclosure. Among them:

[0022] Figure 1 is a schematic flowchart of a data screening method based on a large model provided by an embodiment of the present disclosure;

[0023] Figure 2 is a schematic flowchart of some steps of a data screening method based on a large model provided by an embodiment of the present disclosure;

[0024] Figure 3 is a schematic flowchart of some steps of a data screening method based on a large model provided by an embodiment of the present disclosure;

[0025] Figure 4 is a schematic flowchart of some steps of a data screening method based on a large model provided by an embodiment of the present disclosure;

[0026] Figure 5 is a schematic flowchart of some steps of a data screening method based on a large model provided by an embodiment of the present disclosure;

[0027] Figure 6It is a schematic flowchart of some steps of a data screening method based on a large model provided by an embodiment of the present disclosure;

[0028] Figure 7 It is a schematic flowchart of some steps of a data screening method based on a large model provided by an embodiment of the present disclosure;

[0029] Figure 8 It is a schematic structural diagram of a data screening device based on a large model provided by an embodiment of the present disclosure;

[0030] Figure 9 It is a block diagram of an electronic device for implementing the data screening method based on a large model of an embodiment of the present disclosure. Detailed implementation manners

[0031] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to assist understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted below.

[0032] In some related technologies, the most basic shopping search function of an e-commerce platform relies on accurate keywords input by users to match with the database inside the e-commerce platform. Based on keyword search, users can further filter and refine by manually selecting conditions such as brand, price range, color, etc.

[0033] Keyword search and manual screening have relatively high requirements for user input. If the user's expression is vague or not completely accurate, it is difficult for the system to provide ideal search results. Fuzzy queries, ambiguous statements, or scenario requirements cannot be well supported.

[0034] In some related technologies, an e-commerce platform can provide personalized recommendations for users based on the user's historical browsing and purchase records, combined with collaborative filtering or content-based recommendation algorithms.

[0035] Although the recommendation algorithm can recommend products based on the user's historical behavior, it is often generated offline or in batches, and the real-time response ability is weak.

[0036] In some related technologies, with the development of artificial intelligence technology, some e-commerce platforms have begun to use artificial intelligence assistants to help users perform simple natural language product searches. However, these artificial intelligence assistants mostly adopt fixed conversation scripts and predefined rules, with low flexibility and limited processing of the complexity of user semantics.

[0037] The data screening method and device, electronic device, and computer-readable storage medium based on a large model provided by the embodiments of the present disclosure aim to solve at least one of the above technical problems in the prior art.

[0038] The data screening method based on a large model provided by the embodiments of the present disclosure can be executed by an electronic device such as a terminal device or a server. The terminal device can be an in-vehicle device, a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, an in-vehicle device, a wearable device, etc. The method can be implemented by a processor calling computer-readable program instructions stored in a memory. Alternatively, the method can be executed by the server.

[0039] Figure 1 A flowchart showing the data screening method based on a large model provided by the embodiments of the present disclosure is shown. As Figure 1 shown, the data screening method based on a large model provided by the embodiments of the present disclosure can include step S110, step S120, and step S130.

[0040] In step S110, user interaction input is obtained, and based on a pre-trained large language model, the user interaction input is parsed to generate product screening conditions.

[0041] In step S120, according to the product screening conditions, product information meeting the product screening conditions is obtained from an e-commerce platform through an application programming interface of the e-commerce platform.

[0042] In step S130, the product information is fed back for the user to obtain the product information through interaction.

[0043] For example, the data screening method based on a large model provided by the embodiments of the present disclosure can be used in intelligent dialogue products. In some possible implementation manners, in step S110, the user interaction input can be the content input by the user to the intelligent dialogue product through an interaction device, which is the user's description of the product in natural language. The user interaction input can be used as the Prompt of a pre-trained LLM (Large Language Model), belonging to the pre-trained LLM.

[0044] The pre-trained LLM is used to parse the user interaction input, determine the user's requirements from the user interaction input, and generate product screening conditions.

[0045] Among them, determining the user requirements from the user interaction input can be to extract keywords from the user interaction input using natural language processing methods, determine the user requirements by combining the keywords, and generate product screening conditions.

[0046] Before using the LLM to process the user interaction input, the user interaction input can also be preprocessed first to convert the user interaction input into a form that is easier for the LLM to understand.

[0047] The generated product screening conditions can include product categories, as well as product attributes such as color, price, and usage.

[0048] In some possible implementation manners, in step S120, the e-commerce platform is a platform that provides product purchase services. The API (Application Programming Interface) of the e-commerce platform is provided by the e-commerce platform and is an interface for accessing the database of the e-commerce platform. By calling the API of the e-commerce platform, the latest inventory, price, activities, and other product information of the e-commerce platform can be retrieved.

[0049] By calling the API of the e-commerce platform, according to the product screening conditions, a search can be performed in the database of the e-commerce platform to obtain products that meet the product screening conditions and the corresponding product information.

[0050] The product information includes product attribute information such as the color, price, and usage of the product, and can also include subjective information such as the description of the product by the merchant and user comments.

[0051] In the case where the data screening method provided in the embodiments of the present disclosure is used for an intelligent dialogue product, and the intelligent dialogue product is an intelligent dialogue product provided by a certain e-commerce platform, the API of the e-commerce platform called in this step can be the API of the e-commerce platform, and the product information of the e-commerce platform obtained.

[0052] In some specific implementation manners, the APIs of multiple e-commerce platforms can also be called to obtain product information that meets the product screening conditions from multiple e-commerce platforms, so as to avoid the situation where the product categories of a single e-commerce platform are not rich enough to obtain products that meet the user's screening conditions.

[0053] In some possible implementation manners, in step S130, the product information is fed back through an interaction device for the user to obtain the product information through interaction. For example, the product information can be fed back to the screen for the user to obtain the product information through interaction using interaction devices such as the screen, mouse, and keyboard.

[0054] Since the types of product information obtained are rich, it may include data with different data structures, different data formats, different data sources, and different data standards. That is, the essence of the obtained product information is large-scale heterogeneous data. Therefore, it is necessary to process the product information, unify it into available information, and then provide feedback.

[0055] When providing feedback on product information, the product information can be sorted based on an algorithm, such as in ascending or descending order of product evaluation ratings, prices, etc. Based on the sorting results, the product information is presented as feedback.

[0056] In the data screening method based on a large model provided in the embodiments of the present disclosure, a large language model is used to semantically analyze user interaction inputs, generate product screening conditions, and by calling an API, products that meet user needs are obtained from the real-time data of an e-commerce platform according to the product screening conditions, providing users with personalized and intelligent product screening and recommendation services.

[0057] The following provides a specific introduction to the data screening method based on a large model provided in the embodiments of the present disclosure.

[0058] As described above, in some possible implementation manners, a pre-trained LLM is used to parse user interaction inputs, determine user requirements from the user interaction inputs, and generate product screening conditions.

[0059] Among them, determining user requirements from user interaction inputs can be to extract keywords from user interaction inputs using natural language processing methods, and determine user requirements and generate product screening conditions by combining the keywords.

[0060] Figure 2 The flowchart shows an implementation manner of using natural language processing methods to extract keywords from user interaction inputs, determine user requirements by combining the keywords, and generate product screening conditions, as Figure 2 shown, which may include step S210 and step S220.

[0061] In step S210, the user interaction input is tokenized, and the tokenization result is input into a pre-trained large language model for parsing to obtain at least one keyword and the key semantic components corresponding to the keyword from the parsing result of the tokens;

[0062] In step S220, product screening conditions are determined based on the keywords and key semantic components. The product screening conditions include the category to which the product belongs and at least one product attribute condition.

[0063] In some possible implementation manners, in step S210, the large language model may be GPT (Generative Pre-trained Transformer, a generative pre-trained language model based on the Transformer architecture) or BERT (Bidirectional Encoder Representations from Transformers).

[0064] In some possible implementation manners, the tokenizer is used to perform tokenization processing on the user interaction input, and the tokenization processing result output by the tokenizer is input into GPT or BERT to analyze the semantic structure of the user interaction input. According to the semantic structure of the user interaction input, the tokenization processing result is parsed based on the attention mechanism, and at least one keyword is determined from the parsed tokenization processing result (i.e., the token parsing result), and the semantic component corresponding to the keyword is determined through the semantic feature corresponding to the keyword, that is, the key semantic component.

[0065] Among them, the semantic component refers to the basic semantic features that constitute a word meaning. That is to say, the key semantic component is the word meaning corresponding to the keyword, which is an explanation and supplement to the keyword.

[0066] In some possible implementation manners, in step S220, the commodity category information that the user wants, that is, the category to which the commodity belongs, and other constraint conditions of the commodity for the user, such as commodity attributes conditions such as commodity color, commodity price, commodity quantity, commodity origin, etc., are determined according to the keyword and the key semantic component, and the commodity category and the commodity attribute conditions are combined into a commodity screening condition.

[0067] In some specific implementation manners, the commodity attribute conditions may include commodity attribute conditions corresponding to the inherent attributes of the commodity, may also include commodity attribute conditions corresponding to the sales attributes of the commodity, and may further include commodity attribute conditions composed of commodity evaluation attributes.

[0068] Among them, the inherent attribute of the commodity refers to the nature inherent in the commodity itself, which is a set of differences (properties different from other products) of the commodity in different fields.

[0069] The inherent attributes of the commodity may include:

[0070] Commodity physical attributes, such as including commodity size, commodity weight, commodity material, commodity color, etc.

[0071] Commodity functional attributes: that is, the functions and performances that the commodity can provide, such as the battery life of electronic products, the processing speed of the processor, the pixel of the camera, etc.

[0072] Commodity usage attributes: including durability, ease of operation, stability, safety, reliability, etc.

[0073] Commodity design attributes: including the aesthetic feeling of the commodity appearance, the fashion degree of the commodity, etc.

[0074] Commodity environmental attributes: the impact of the commodity on the environment and its sustainability characteristics, such as environmentally friendly materials, low energy consumption, waste reduction, etc.

[0075] Commodity sales attributes refer to the sales situation of the commodity, such as the sales quantity of the commodity, the inventory quantity of the commodity, the distribution of the sales regions of the commodity, the distribution of the characteristics of the sales population of the commodity, etc.

[0076] Commodity evaluation attributes refer to the evaluations of users on the commodity, which can include the content of the commodity evaluation (the evaluation of the commodity by mobile phone users can be obtained, and the commodity evaluation keywords can be obtained by processing the commodity evaluation as described in step S210), the number of evaluations of the commodity, and the corresponding favorable rate, etc.

[0077] In some possible implementation manners, commodity screening conditions can be generated by combining keywords and key semantic components, but there may be some fuzzy keywords in the keywords, and these fuzzy keywords cannot be directly corresponding to commodity attributes. For example, the keyword is "practical", and it cannot be directly corresponding to commodity attributes such as commodity color and commodity price.

[0078] In some possible implementation manners, a knowledge graph can be used to process the fuzzy keywords to expand the fuzzy keywords into commodity attribute conditions.

[0079] In some possible implementation manners, user intention reasoning can also be used to process the fuzzy keywords to expand the fuzzy keywords into commodity attribute conditions.

[0080] Figure 3 The flowchart shows an implementation manner of using a knowledge graph to process fuzzy keywords. As Figure 3 described, it may include step S310 and step S320.

[0081] In step S310, semantic features of the keyword are extracted;

[0082] In step S320, based on a pre-trained scenario knowledge graph, according to the semantic features of the keyword, the commodity usage scenarios associated with the keyword and the commodity screening conditions associated with the commodity usage scenarios are obtained.

[0083] In some possible implementation manners, in step S310, the keyword can be the above-mentioned fuzzy keyword that cannot be directly corresponding to commodity attributes.

[0084] Extracting the semantic features of the keyword can be directly obtaining the semantic features of the keyword obtained in the above step S210, or can be re-extracting the semantic features of the keyword using a new semantic feature extractor.

[0085] In some possible implementation manners, in step S320, the pre-trained scenario knowledge graph may be a semantic network that provides structured information for natural language content by defining the relationships between mistakes, behaviors, and scenarios, so as to help deduce the product screening conditions.

[0086] In some possible implementation manners, information can be extracted from different data sources (user interaction data, product data of e-commerce platforms, external knowledge base data) to provide materials for constructing the knowledge graph. Nodes are extracted from the collected data, such as products, product attributes, product usage scenarios (such as product usage seasons, types of activities participated by products, etc.) and edges (the relationships between them). A preliminary knowledge graph is established through association, and the preliminary knowledge graph is extended through graph neural networks and machine learning to generate a scenario knowledge graph.

[0087] According to the semantic features of the keyword, based on the knowledge graph, determine the product usage scenario associated with the keyword in the knowledge graph, obtain the product attributes associated with the product usage scenario, and use the product attributes associated with the product usage scenario as product attribute conditions to construct product screening conditions.

[0088] In some possible implementation manners, write the correspondence between the keyword and the obtained product attribute conditions associated with the product usage scenario into the scenario knowledge graph to extend the scenario knowledge graph.

[0089] In this way, as more user interaction inputs processed by the data screening method based on a large model provided by the embodiments of the present disclosure, the scenario knowledge graph can be continuously updated and optimized, the "understanding" of the keyword by the scenario knowledge graph is more accurate, and the obtained product screening conditions are also more accurate.

[0090] Figure 4 The flowchart shows an implementation manner of processing fuzzy keywords using user intent reasoning, as Figure 4 shown, and may include step S410 and step S420.

[0091] In step S410, extract the semantic features of the keyword;

[0092] In step S420, according to the semantic features of the keyword and historical user interaction inputs, obtain the user intent related to the keyword and the product screening conditions associated with the user intent through semantic reasoning.

[0093] In some possible implementation manners, in step S410, the keyword may be the above-mentioned fuzzy keyword that cannot be directly corresponded to product attributes.

[0094] The semantic features of the extracted keywords can be directly obtained from the semantic features of the keywords obtained in the previous step S210, or the semantic features of the keywords can be re-extracted using a new semantic feature extractor.

[0095] In some possible implementation manners, in step S420, the historical user interaction input may be the user interaction input entered by the user through interaction before the obtained user interaction input. In the case where the data screening method based on the large model provided in the embodiments of the present disclosure is used for an intelligent dialogue product, the historical user interaction input may be the dialogue content of the user's previous conversation with the intelligent dialogue product.

[0096] The context learning ability of the large model can be utilized to learn the semantic features of the keywords and the context relationship of the historical user interaction input, and semantic reasoning can be performed according to the context relationship to infer the intention of the keywords input by the user, that is, the user intention. After obtaining the user intention, the commodity attributes that meet the user intention can be inferred according to the user intention as the commodity screening conditions.

[0097] For example, the user describes "I want to find a winter coat that is relatively warm, preferably with a color that is dirt-resistant". Through semantic reasoning, commodity attributes such as "warm = high-loft down material" and "color dirt-resistant = dark color series" will be deduced, and these commodity attributes are added to the commodity screening conditions.

[0098] After obtaining the commodity screening conditions in the above-described manner, global search can be performed first according to the commodity screening conditions, and then refined search can be performed to improve the search speed and efficiency.

[0099] Figure 5 The flowchart shows an implementation manner of first performing a global search and then a refined search according to the commodity screening conditions, as Figure 5 shown, which may include step S510, step S520, and step S530.

[0100] In step S510, through the application programming interface of the e-commerce platform, a global fuzzy search is performed on the e-commerce platform according to the category to which the commodity belongs to obtain commodities that meet the category to which the commodity belongs;

[0101] In step S520, according to the commodity attribute conditions, the commodities that meet the category to which the commodity belongs are screened to obtain commodities that meet the commodity attribute conditions;

[0102] In step S530, the commodity information that meets the commodity attribute conditions is obtained.

[0103] In some possible implementation manners, in step S510, performing a global fuzzy search on the e-commerce platform means performing a fuzzy search on the entire database of the e-commerce platform.

[0104] Among them, fuzzy search allows for a certain degree of difference between the information to be searched and the search query. Therefore, performing a global fuzzy search on an e-commerce platform according to the category to which the product belongs means searching the entire database of the e-commerce platform for products whose category is the same as or similar to the category to which the product belongs.

[0105] In some possible implementation manners, in step S520, when there are multiple product attribute conditions, the products that meet the category to which the product belongs are gradually screened according to each product attribute condition. Specifically, according to the effect of the product attribute condition on product screening, first use the product attribute condition with a greater effect on product screening for screening. Based on the screening result, then use the product attribute condition with the second greatest effect on product screening for screening until all product attribute conditions are used for screening.

[0106] First, performing a global fuzzy search on an e-commerce platform according to the category to which the product belongs is to quickly screen out most products from a database with a large number of products, reduce the number of objects for fine screening, and improve the query speed. And first using the product attribute condition with a greater effect on product screening is also to reduce the number of screening objects for subsequent screening and improve the query speed.

[0107] In some possible implementation manners, in step S530, the method for obtaining product information that meets the product attribute conditions is as described in step S120, and will not be elaborated here.

[0108] In some possible implementation manners, after the above steps, the number of products that meet the product attribute conditions is still very large. Obviously, this is not what the user wants. Therefore, further screening can be performed.

[0109] In some possible implementation manners, when the number of products that meet the product attribute conditions exceeds a preset number, based on a pre-trained user shopping preference model, the products that meet the product attribute conditions are screened.

[0110] Among them, the user shopping preference model is a model established by collecting behavioral characteristics such as the user's purchase history, browsing history, and liked and collected products on the e-commerce platform, and using a machine learning model, such as a collaborative filtering algorithm, based on the above data.

[0111] Screening the products that meet the product attribute conditions based on a pre-trained user shopping preference model means outputting the products that best match the user's habits and the current conditions based on the user's personal preferences.

[0112] In some possible implementations, products can be divided into multiple relevance levels. When there is no historical behavior, the user will be guided to clarify their preferences and learn from the behavior feedback to update the user shopping preference model.

[0113] In some possible implementations, when the user is not satisfied with the products obtained according to the product filtering conditions, the user can dynamically modify and optimize the filtering to improve the flexibility of the query and the user experience.

[0114] Figure 6 The flowchart shows an implementation of dynamically modifying and optimizing filtering, as Figure 6 shown, which may include step S610, step S620, and step S630.

[0115] In step S610, at least one product detail input by the user is obtained, and the product detail is parsed based on a pre-trained large language model to generate the current filtering conditions;

[0116] In step S620, the product filtering conditions are updated according to the current filtering conditions;

[0117] In step S630, according to the updated product filtering conditions, product information that meets the updated product filtering conditions is obtained from the e-commerce platform through the application programming interface of the e-commerce platform.

[0118] In some possible implementations, in step S610, the product detail can be a newly added condition (such as increasing the budget), a reduced condition (removing a specific product type), or a fuzzy requirement description. The above method can be used to parse the product detail to generate the current filtering conditions.

[0119] In some possible implementations, in step S620, if there is a conflict between the current filtering conditions and the product filtering conditions, then the current filtering conditions are used to replace the conditions in the product filtering conditions that conflict with the current filtering conditions. If the product attributes corresponding to the product attribute conditions in the current filtering conditions and the product filtering conditions are all different, the current filtering conditions and the product filtering conditions are combined into new product filtering conditions.

[0120] In some possible implementations, in step S630, the above method is used to obtain product information that meets the updated product filtering conditions according to the updated product filtering conditions.

[0121] In some possible implementations, the product details input by the user may be too general, resulting in too much product information that meets the updated product filtering conditions. At this time, the user can be intelligently queried and prompted to refine the product details.

[0122] Figure 7A flowchart showing an implementation method of intelligently asking and prompting users to refine product details is shown. Figure 7 As shown, it may include step S710, step S720, and step S730.

[0123] In step S710, when the number of products that meet the updated product screening conditions is greater than the preset number, questions about product details are generated and fed back;

[0124] In step S720, the user's answer input to the question about the product details is obtained, and the product details are updated according to the answer input;

[0125] In step S730, the commodities meeting the updated commodity screening conditions are screened according to the updated commodity details.

[0126] In some possible implementations, in step S710, questions about product details may be generated based on other product attributes of the product, and fed back on the interactive device for the user to obtain and input answers to the questions about product details, i.e., answer input.

[0127] In some possible implementations, in step S720, the answer input may be used as a user interaction input, and the above-mentioned method may be used to update the product details.

[0128] In some possible implementations, in step S730, the updated product details are used as the product details input by the user. Figure 6 The method is used to filter products that meet the updated product filtering conditions.

[0129] In some possible implementations, multiple rounds of questioning can be used to further refine product details and obtain products that meet user needs. In other words, multiple rounds of questioning can be used to further refine product details and obtain products that meet user needs. Figure 7 The corresponding method is used to further filter the products that meet the product filtering conditions.

[0130] In summary, in the data screening method based on a large model provided in the embodiment of the present disclosure, it is possible to process uncertain information such as fuzzy words and scenario-based demand expressions, and to mine the user needs behind them. Moreover, by mining the user's historical shopping behavior, the personalized recommendation algorithm is used to further optimize the accuracy of product recommendations, so that users can efficiently obtain products that meet their personal preferences, needs, and restrictions. Furthermore, the user can also dynamically adjust the product screening conditions or the products that meet the product screening conditions through multiple rounds of interactive dialogues to provide users with a priority product recommendation experience.

[0131] Based on Figure 1 The same principle as shown in the method, Figure 8The figure shows a schematic structural diagram of a data screening device based on a large model provided by an embodiment of the present disclosure. As Figure 8 shown, the data screening device 80 based on the large model may include:

[0132] A condition module 810, configured to obtain user interaction input, parse the user interaction input based on a pre-trained large language model, and generate product screening conditions;

[0133] A screening module 820, configured to obtain product information that meets the product screening conditions from an e-commerce platform through an application programming interface of the e-commerce platform according to the product screening conditions;

[0134] A feedback module 830, configured to feedback product information for the user to obtain product information through interaction.

[0135] In the data screening device based on the large model provided by the embodiment of the present disclosure, a large language model is used to perform semantic parsing on user interaction input to generate product screening conditions, and by calling the API, products that meet the user's needs are obtained from the real-time data of the e-commerce platform according to the product screening conditions, providing personalized and intelligent product screening and recommendation services for the user.

[0136] In some possible implementation manners, the condition module includes: a keyword parsing unit, configured to perform word segmentation processing on user interaction input, input the word segmentation processing result into a pre-trained large language model for parsing, and obtain at least one keyword and the key semantic components corresponding to the keyword from the word segmentation processing result; a condition generation unit, configured to determine product screening conditions according to the keyword and the key semantic components, and the product screening conditions include the category to which the product belongs and at least one product attribute condition.

[0137] In some possible implementation manners, the condition generation unit is further configured to: extract the semantic features of the keyword; based on a pre-trained scenario knowledge graph, according to the semantic features of the keyword, obtain the product usage scenarios associated with the keyword, and the product screening conditions associated with the product usage scenarios.

[0138] In some possible implementation manners, the condition generation unit is further configured to: use the correspondence between the keyword and the product screening conditions associated with the product usage scenario to expand the scenario knowledge graph.

[0139] In some possible implementation manners, the condition generation unit is further configured to: extract the semantic features of the keyword; according to the semantic features of the keyword and historical user interaction input, obtain the user intent associated with the keyword and the product screening conditions associated with the user intent through semantic reasoning.

[0140] In some possible implementations, the screening module includes: a global search unit, configured to perform a global fuzzy search on an e-commerce platform through an application programming interface of the e-commerce platform to obtain products that match the category to which the product belongs according to the category to which the product belongs; a filtering unit, configured to filter the products that match the category to which the product belongs according to product attribute conditions to obtain products that meet the product attribute conditions; and a product information unit, configured to obtain product information of the products that meet the product attribute conditions.

[0141] In some possible implementations, the large model-based data screening device further includes: in the case where the number of products that meet the product attribute conditions exceeds a preset number, filtering the products that meet the product attribute conditions based on a pre-trained user shopping preference model.

[0142] In some possible implementations, the large model-based data screening device further includes an optimization module, configured to: obtain at least one product detail input by a user, parse the product detail based on a pre-trained large language model to generate a current screening condition; update the product screening condition according to the current screening condition; and obtain product information that meets the updated product screening condition from the e-commerce platform through the application programming interface of the e-commerce platform according to the updated product screening condition.

[0143] In some possible implementations, the optimization module is further configured to: in the case where the number of products that meet the updated product screening condition is greater than a preset number, generate and feedback questions about the product detail; obtain an answer input of the user to the questions about the product detail, and update the product detail according to the answer input; and filter the products that meet the updated product screening condition according to the updated product detail.

[0144] It can be understood that the above-mentioned modules of the large model-based data screening device in the embodiments of the present disclosure have functions corresponding to the corresponding steps of the large model-based data screening method in the embodiments shown in Figure 1 The function can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. The above modules can be software and / or hardware, and the above modules can be implemented separately or multiple modules can be integrated. For the function descriptions of the above modules of the large model-based data screening device, reference can be specifically made to the corresponding descriptions of the large model-based data screening method in the embodiments shown in Figure 1 The corresponding descriptions in the embodiments shown in are not elaborated herein.

[0145] In the technical solution of the present disclosure, the acquisition, storage, and application of the user's personal information involved all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0146] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0147] The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the data screening method based on a large model provided in the embodiments of the present disclosure.

[0148] Compared with the prior art, the electronic device uses a large language model to perform semantic parsing on user interaction inputs, generates product screening conditions, and obtains products that meet user needs from the real-time data of the e-commerce platform according to the product screening conditions by calling an API, providing users with personalized and intelligent product screening and recommendation services.

[0149] The readable storage medium is a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to execute the data screening method based on a large model provided in the embodiments of the present disclosure.

[0150] Compared with the prior art, the readable storage medium uses a large language model to perform semantic parsing on user interaction inputs, generates product screening conditions, and obtains products that meet user needs from the real-time data of the e-commerce platform according to the product screening conditions by calling an API, providing users with personalized and intelligent product screening and recommendation services.

[0151] The computer program product includes a computer program, and the computer program, when executed by a processor, implements the data screening method based on a large model provided in the embodiments of the present disclosure.

[0152] Compared with the prior art, the computer program product uses a large language model to perform semantic parsing on user interaction inputs, generates product screening conditions, and obtains products that meet user needs from the real-time data of the e-commerce platform according to the product screening conditions by calling an API, providing users with personalized and intelligent product screening and recommendation services.

[0153] Figure 9FIG. shows a schematic block diagram of an exemplary electronic device 800 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0154] As Figure 9 shown, the device 900 includes a computing unit 901 that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0155] A plurality of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a magnetic disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0156] The computing unit 901 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 executes the various methods and processes described above, such as the large model-based data screening method. For example, in some embodiments, the large model-based data screening method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the large model-based data screening method described above can be executed. Alternatively, in other embodiments, the computing unit 901 can be configured to execute the large model-based data screening method by any other suitable means (e.g., by means of firmware).

[0157] Various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), system-on-chip systems (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a dedicated or general-purpose programmable processor, receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0158] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program codes can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0159] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0160] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, speech input, or tactile input).

[0161] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0162] A computer system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The client-server relationship is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, a server of a distributed system, or a server incorporating blockchain.

[0163] It should be understood that the various forms of processes shown above can be used, with steps reordered, added or deleted. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and no limitations are imposed herein.

[0164] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub - combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A data screening method based on a large model, comprising: Obtain user interaction input, parse the user interaction input based on a pre-trained large language model, and generate product screening conditions; According to the commodity screening condition, obtaining commodity information that meets the commodity screening condition from the e-commerce platform through an application programming interface of the e-commerce platform; The product information is fed back so that the user can obtain the product information through interaction.

2. The method according to claim 1, wherein: The obtaining of user interaction input, parsing the user interaction input based on a pre-trained large language model, and generating product screening conditions includes: Performing word segmentation processing on the user interaction input, inputting the word segmentation processing result into a pre-trained large language model for parsing, and obtaining at least one keyword and key semantic components corresponding to the keyword from the word segmentation parsing result; The commodity screening condition is determined according to the keyword and the key semantic component, and the commodity screening condition includes the category to which the commodity belongs and at least one commodity attribute condition.

3. The method according to claim 2, wherein: The determining the product screening condition according to the keyword and the key semantic component includes: Extracting semantic features of the keywords; Based on the pre-trained scenario knowledge graph, the product usage scenario associated with the keyword and the product screening condition associated with the product usage scenario are obtained according to the semantic features of the keyword.

4. The method according to claim 3, further comprising: The scenario knowledge graph is expanded using the keywords and the corresponding relationship between the commodity screening conditions associated with the commodity usage scenario.

5. The method according to claim 2, wherein: The determining the product screening condition according to the keyword and the key semantic component includes: Extracting semantic features of the keywords; According to the semantic features of the keyword and historical user interaction input, the user intent related to the keyword and the commodity screening condition associated with the user intent are obtained through semantic reasoning.

6. The method according to claim 2, wherein: The step of obtaining commodity information that meets the commodity screening condition from the e-commerce platform through an application programming interface of the e-commerce platform according to the commodity screening condition includes: Through the application programming interface of the e-commerce platform, a global fuzzy search is performed on the e-commerce platform according to the category of the commodity to obtain commodities that match the category of the commodity; According to the commodity attribute conditions, commodities that meet the category to which the commodity belongs are screened to obtain commodities that meet the commodity attribute conditions; Get product information that meets the product attribute conditions.

7. The method according to claim 6, further comprising: When the number of acquired commodities meeting the commodity attribute conditions exceeds a preset number, the commodities meeting the commodity attribute conditions are screened based on a pre-trained user shopping preference model.

8. The method according to claim 1, further comprising: Obtain at least one product detail input by the user, parse the product detail based on a pre-trained large language model, and generate a current filtering condition; Update the product screening condition according to the current screening condition; According to the updated commodity screening conditions, commodity information that meets the updated commodity screening conditions is obtained from the e-commerce platform through the application programming interface of the e-commerce platform.

9. The method according to claim 8, further comprising: When the number of products that meet the updated product screening conditions is greater than the preset number, questions about product details are generated and feedback is given; Obtaining a user's answer input to a question about product details, and updating the product details according to the answer input; The products that meet the updated product screening conditions are screened according to the updated product details.

10. A data screening device based on a large model, comprising: A condition module is used to obtain user interaction input, parse the user interaction input based on a pre-trained large language model, and generate product screening conditions; A screening module, configured to obtain commodity information that meets the commodity screening conditions from the e-commerce platform through an application programming interface of the e-commerce platform according to the commodity screening conditions; The feedback module is used to feed back the product information so that the user can obtain the product information through interaction.

11. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-9.

13. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images