Information processing methods, store diagnosis method, intelligent assistant working method, system and device

Abstract

Provided in the embodiments of the present disclosure are information processing methods, a store diagnosis method, an intelligent assistant working method, a system and a device. A solution provided in the present disclosure uses a causal inference algorithm to build a text repository (for example, at least one factor of a preset question stored in the text repository, which is obtained by performing attribution analysis on the preset question by using the causal inference algorithm), and empowers a merchant-side management tool with a first preset model which is fine-tuned by using a training sample set related to a store operation scenario, so as to upgrade the level of intelligence of the management tool. The text repository provides data support to retrieve a store-diagnosis result and / or at least one factor of a user's query question, as well as suggestion content linked to said factor; the first preset model provides intelligent response support; and, on the basis of the store diagnosis result and / or the user's query question and the suggestion content linked to the at least one factor, the first preset model can output a text that better conforms to natural language logic and is suitable for the store operation scenario.

Description

Information processing, store diagnosis and intelligent assistant working method, system and device

[0001] The present disclosure claims priority to Chinese Patent Application No. 202410867589.9, filed on June 28, 2024, with the Chinese Patent Office, entitled "Information processing, store diagnosis and intelligent assistant working method, system and device", the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0002] The present disclosure relates to the field of computer technology, and particularly relates to an information processing, store diagnosis and intelligent assistant working method, system and device. BACKGROUND

[0003] There are many merchants on an e-commerce platform, and the merchants can manage the stores operated by the merchants on the e-commerce platform through a management tool provided by the platform. For example, the merchants can obtain recent traffic data (such as the number of visits, the number of orders, the number of payments, etc.) of the stores through the management tool, and can also set keywords for the store goods through the management tool to improve the search probability of the goods, etc.

[0004] However, the existing management tool is not intelligent enough, and the merchants can only understand some statistical data of the stores through the management tool. For experienced merchants, they can accurately locate the problems of the stores through these statistical data, and then adjust the operation strategy based on rich experience. However, for inexperienced merchants, seeing these statistical data cannot accurately locate the problems of the stores, and they do not know how to adjust the operation strategy. SUMMARY

[0005] Embodiments of the present disclosure provide an information processing, store diagnosis and intelligent assistant working method, system and device to help the merchants operate the stores.

[0006] In a first embodiment of the present disclosure, an information processing method is provided. The method comprises:

[0007] For a target store, a target problem is determined;

[0008] In a script database, at least one factor of the target problem is retrieved;

[0009] Based on the target problem and the suggestion content associated with the at least one factor, a prompt word Prompt is generated;

[0010] A script corresponding to the Prompt is output by using a first preset model, to help a user solve or improve the target problem;

[0011] The script database stores at least one factor of the preset question obtained by using a causal inference algorithm to perform attribution analysis on the preset question.

[0012] In a second embodiment of the present disclosure, an information processing method is provided. The method comprises:

[0013] A query template library is constructed, the query template library comprising a plurality of query templates, one query template being associated with one preset question related to shop operation;

[0014] Based on the query template library, a script database is constructed, wherein the script database stores at least one factor of the query template obtained by using a causal inference algorithm to perform attribution analysis on the preset question associated with the query template;

[0015] According to the query template library and the script database, a training sample set is generated;

[0016] The pre-trained model is adjusted and trained using the training sample set to obtain a first preset model capable of processing intelligent response tasks in a shop operation scenario.

[0017] In a third embodiment of the present disclosure, a shop diagnosis method is provided. The method is applicable to a client, and specifically, the method comprises:

[0018] In response to a diagnosis triggering event for a target shop, triggering shop diagnosis to present and / or play the diagnosed target problem;

[0019] Displaying and / or playing a script, the script comprising suggestion content for assisting a user to solve or improve the target problem;

[0020] The script is obtained by using a first preset model, the first preset model being obtained by adjusting and training a pre-trained model for intelligent response tasks in a shop operation scenario; the suggestion content is associated with at least one factor of the target problem, the at least one factor of the problem being obtained by using a causal inference algorithm to perform attribution analysis on the target problem.

[0021] In a fourth embodiment of the present disclosure, a shop intelligent assistant working method is provided. The method is applicable to a server, and specifically, the method comprises:

[0022] Displaying a management page of a target shop;

[0023] Listen to the operation of the user on the management page to capture the target problem of the user on the shop operation;

[0024] In response to the confirmation instruction triggered by the user for the target problem, a script is displayed and / or played, the script including suggestion content for assisting the user to solve or improve the target problem;

[0025] The script is obtained by using a first preset model, the first preset model being obtained by adjusting and training a pre-training model for an intelligent response task in a shop operation scenario; the suggestion content is associated with at least one factor of the target problem, and the at least one factor of the problem is obtained by performing attribution analysis on the target problem by using a causal inference algorithm.

[0026] In a fifth embodiment of the present disclosure, a service system is provided. The service system includes a client and a server. The client is configured to implement the steps of the shop diagnosis method or the steps of the shop intelligent assistant working method. The server is configured to implement the steps of the information processing method.

[0027] In a sixth embodiment of the present disclosure, an electronic device is provided. The electronic device includes a memory and a processor. The memory is configured to store executable instructions. The processor is configured to implement the steps of the method by running the executable instructions.

[0028] In a sixth embodiment of the present disclosure, a computer-readable storage medium is provided. The storage medium stores computer instructions, which are executed by a processor to implement the steps of the method.

[0029] The seventh embodiment of the present disclosure also provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, the processor executes the steps of the method.

[0030] The pre-training model mentioned in the embodiments of the present disclosure refers to a machine learning model with a large number of parameters generated by training with a large amount of data using a self-supervised method. It provides excellent distributed feature representation capability and model generalization capability for downstream tasks (such as the intelligent response task in the shop operation scenario mentioned in the embodiments of the present disclosure). The pre-training model can also be called a universal large model.

[0031] The technical solution provided by the embodiments of the present disclosure combines a large model with a causal inference technology, that is, a causal inference algorithm is used when constructing a copy database to perform attribution analysis on preset questions associated with a Query template to obtain at least one factor; the training sample used for adjusting and training a pre-training model is generated based on a Query template library and a copy database. A training sample set composed of multiple such training samples is used to adjust and train the pre-training model to adjust it into a first preset model suitable for a store operation scenario.

[0032] The solution provided by the embodiments of the present disclosure constructs a copy database (for example, at least one factor of a preset question obtained by performing attribution analysis on the preset question by using a causal inference algorithm) by using a causal inference algorithm, and enables a merchant-side management tool by using a first preset model adjusted by using a training sample set related to a store operation scenario, so that the management tool has a higher degree of intelligence. The copy database, the first preset model, and the like can be deployed on a server, and a corresponding product form on a client side can be an intelligent assistant. A user can diagnose a target store or ask questions related to store operation through the intelligent assistant. The copy database provides data support for the intelligent assistant, and at least one factor of a store diagnosis result and / or a question asked by a user and suggestion content associated with the factor can be retrieved. The first preset model adjusted by training provides intelligent response support for the intelligent assistant. Based on the store diagnosis result and / or the question asked by the user and the suggestion content associated with the at least one factor, the first preset model can output a copy more conforming to natural language logic and suitable for a store operation scenario to help the user operate the store.

[0033] It can be seen that the technical solution provided by the embodiments of the present disclosure uses a large model and a causal inference technology to improve the intelligent service capability of a platform and help a merchant operate a store. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0035] FIG. 1 is a schematic diagram of a technical architecture corresponding to the technical solution provided by the embodiments of the present disclosure;

[0036] FIG. 2 is a schematic diagram of a service system provided by an embodiment of the present disclosure;

[0037] FIG. 3 is a flowchart of an information processing method provided by an embodiment of the present disclosure;

[0038] FIG. 4 is a flowchart of an information processing method according to another embodiment of the present disclosure;

[0039] FIG. 5 is a flowchart of a store diagnosis method according to an embodiment of the present disclosure;

[0040] FIG. 6a is a first interface according to an embodiment of the present disclosure;

[0041] FIG. 6b is a second interface according to an embodiment of the present disclosure;

[0042] FIG. 6c is a third interface according to an embodiment of the present disclosure;

[0043] FIG. 7 is a flowchart of a store intelligent assistant working method according to an embodiment of the present disclosure;

[0044] FIG. 8 is a structural block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0045] In order to enable those skilled in the art to better understand the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure.

[0046] In some of the processes described in the specification, claims, and accompanying drawings, multiple operations are described in a specific order. These operations can be performed in a different order or in parallel, unless otherwise specifically noted. The numbering of the operations described in the specification, claims, and accompanying drawings is merely for the purpose of distinguishing between different operations, and does not necessarily correspond to the order in which the operations are performed. Additionally, the processes described in the specification, claims, and accompanying drawings can include more or fewer operations than those described, and the operations can be performed in a different order or in parallel. It should be noted that the terms "first", "second", and the like, used in the description and in the claims, are used to distinguish between different messages, devices, modules, and the like, and do not necessarily correspond to the order in which the messages, devices, modules, and the like are described. The embodiments described below are merely some of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present disclosure.

[0047] Before introducing the embodiments, some technical terms appearing in the text will be briefly introduced.

[0048] Query: the meaning of query, in order to find files, websites, records or a series of records in the database, the message sent by the search engine or the database.

[0049] Pre-trained model: refers to a machine learning model with a large number of parameters generated by training a large amount of data using a self-supervised method, which provides excellent distributed feature representation capability and model generalization capability for downstream tasks. The pre-trained model can be called a general-purpose large model.

[0050] SFT: Supervised Finetuning, supervised fine-tuning, based on a large pre-trained model, fine-tuning the pre-trained model for a specific task in a specific scenario. During fine-tuning, the pre-trained model adjusts the parameters and structure of the pre-trained model according to the characteristics of the task to improve the performance of the pre-trained model on the task.

[0051] LoRA: a model Fine Tuning method that freezes the model parameters of the pre-trained model and injects a trainable low-rank matrix into the attention layer of the Transformer architecture, greatly reducing the number of trainable parameters of the downstream task and improving the training speed.

[0052] SHAP: Shapley Value, a method for calculating the causal relationship between features and labels, by enumerating different feature combinations, calculating the impact on the prediction result when adding a feature in a certain feature combination, compared to not adding this feature. The impact is the marginal benefit of the feature to the feature combination.

[0053] Tars: ICUB large language model, a deep learning model trained based on a large amount of text data.

[0054] ab: inquiry, when a buyer or seller wants to buy or sell a certain commodity, the buyer or seller asks the other party for the transaction conditions.

[0055] rfq: request for quotation.

[0056] In order to facilitate the understanding of the technical solutions, the underlying technical architecture of each embodiment of the present disclosure is described. The technical solutions provided in each of the following embodiments are implemented based on the technical architecture. Referring to FIG. 1, a technical architecture diagram of the technical solution provided by the present disclosure is shown. As shown in FIG. 1, the technical architecture of the technical solution provided by the present disclosure includes:

[0057] I. Constructing a Query template library (label 101 in FIG. 1)

[0058] In an implementable technical solution, constructing a Query template library can include:

[0059] S11, determine a plurality of basic Query templates;

[0060] S12, extending the plurality of basic Query templates to obtain a plurality of extended Query templates by using a second preset model;

[0061] S13, evaluating the plurality of extended Query templates, and retaining the extended Query templates passing the evaluation;

[0062] The Query template library includes the plurality of basic Query templates and the plurality of extended Query templates passing the evaluation.

[0063] For example, the plurality of basic Query templates in S11 can include but are not limited to: a plurality of basic Query templates covering all traffic nodes, a plurality of basic Query templates related to store optimization items, a plurality of basic Query templates related to store operation problems, etc. In the e-commerce scenario, the traffic nodes can include but are not limited to: keyword search, user access, adding to shopping cart, inquiry, ordering, payment, etc. The store optimization items can include but are not limited to: in-store product keywords, in-store product pictures, store promoted products, product target audience, etc. The store operation problems can include but are not limited to: how to improve the search probability, how to improve the payment amount, how to improve the store access amount, etc. The store mentioned in this paper can be an online store on an e-commerce platform or an online store on a social platform, etc. The online store can also be called "online store" or "e-commerce store". The e-commerce platform or social platform provides technical support for merchants to operate stores on the Internet, display and sell goods or services in online stores, etc. The e-commerce platform or social platform uses Internet technology to allow consumers to browse product information, compare, place orders, pay for goods, and wait for goods to be delivered to the designated address.

[0064] The plurality of basic Query templates covering all traffic nodes can be divided into two categories: the first category is a plurality of basic Query templates corresponding to the differences between the store's own traffic data and the reference store's (such as peer stores and / or stores of the same level) traffic data; the second category is a plurality of basic Query templates corresponding to the differences between the store's first period traffic data and the second period historical traffic data.

[0065] After obtaining the plurality of basic Query templates, a plurality of extended Query templates can be extended offline by using a second preset model. The second preset model can be a large language model (LLM, Large Language Model). The large language model is a deep learning model trained based on massive text data. It can not only generate natural language text, but also deeply understand the meaning of the text and process various natural language tasks.

[0066] The multiple extended queries extended by the second preset model do not necessarily all meet the requirements. Therefore, the multiple extended queries output by the second preset model can be evaluated. The evaluation process can include but is not limited to: obtaining an evaluation rule; respectively evaluating each extended query by using the evaluation rule, and retaining the extended query template that passes the evaluation. Further, human intervention interfaces can also be added in the evaluation process, such as displaying the extended query template that passes the evaluation to the client interface, and the user can also edit, delete, etc. the extended query template that passes the evaluation.

[0067] In the present disclosure, the evaluation rule is not specifically limited, and can include multiple rule items, each of which can be artificially set.

[0068] Finally, a query template library including multiple basic query templates and multiple extended query templates that pass the evaluation is obtained. In the diagnosis of the store, the multiple query templates in the query template library can be sorted according to the diagnosis result of the store (such as low store exposure rate, low click rate, low inquiry volume, etc.). For example, the query templates closely related to the diagnosis result of the store are placed in the front. In this way, the first, second, third or fourth, etc. query templates placed in the front can be used as the query templates corresponding to the diagnosis result of the store. Alternatively, through big data statistics, the common problems of stores in the same industry or at the same level are counted, and the multiple query templates in the query template library are sorted based on the statistical result. In this way, the first, second, third or fourth, etc. query templates placed in the front can be used as the recommended query templates.

[0069] In summary, the query template library can include at least one of the following:

[0070] Query templates corresponding to the reasons for the data differences between any store on the platform and the reference store;

[0071] Query templates corresponding to the reasons for the data differences between any store on the platform at different periods;

[0072] Query templates corresponding to store management problems;

[0073] Query templates corresponding to store optimization items;

[0074] Query templates extended by the second preset model.

[0075] II. Constructing a script database (label 102 in FIG. 1)

[0076] 2.1, Collect basic data such as attribute information of shops on the platform, traffic data of the shop, operation data of the shop, etc. Among them, the attribute information of the shop can include but is not limited to: the industry to which the shop belongs, the level of the shop, etc. The traffic data of the shop can include but is not limited to: the historical period (such as the past 30 days, 60 days, etc.) exposure rate, click rate, inquiry quantity, order daily quantity, etc. The operation data of the shop can include but is not limited to: the number of shop windows, the number of good products, the number of self-marketing interactions, the rfd amount, the advertising consumption amount, the shop flow situation (including the number of shops through different channels such as search, activity, paid advertising, live broadcast, etc.), etc.

[0077] 2.2, Shop optimization item attribution

[0078] Causal inference is a statistical and reasoning method aimed at understanding and inferring the causal relationship between events, variables or behaviors. The present scheme is to use causal inference technology to analyze abnormal traffic nodes that cause shop problems, which are shop optimization items that can be optimized by the shop. Funnel analysis technology is used in attribution analysis. That is, the funnel analysis technology is used to analyze the traffic data and operation data of the shops on the platform to analyze the shop optimization items that solve or improve the shop problems.

[0079] Data funnel nodes are a key element in funnel analysis, representing various traffic nodes experienced from the beginning of an event to the end of an event. These traffic nodes include starting points, intermediate process nodes, and endpoints. Specifically, for different application scenarios, data funnel node settings will be different. In the e-commerce field, data funnel nodes can include customer visits, adding to the shopping cart, ordering, payment, etc. By analyzing the data of these nodes, the conversion of customers at each stage can be understood, and potential problems and optimization items can be found. For example, if the conversion rate of a certain node is low, it may mean that there is a problem with that node that needs to be further adjusted and optimized. That node is the shop optimization item that needs to be optimized by the shop.

[0080] Therefore, when performing funnel analysis, the nodes of the data funnel need to be carefully determined and defined to ensure the accuracy and effectiveness of the analysis. The determination and definition of the nodes of the data funnel in the present embodiment are not specifically limited and need to be analyzed according to specific scenarios. In specific practice, they can be determined and defined according to experience.

[0081] When performing funnel analysis, node optimization point attribution analysis in the traffic funnel can be performed based on the monotone stack method. Among them, the monotone stack is often used to find the position of the first element larger or smaller than a certain element on the left or right of the array, so as to identify the local maximum and minimum values in the sequence.

[0082] For example, two increasing monotonic stacks can be maintained to find the node with the largest difference between the index (access volume, inquiry number, etc.) associated with the problem inquired by the merchant or diagnosed by the merchant store and the reference store (another store of the same industry or same level as the merchant store) in the specific implementation, i.e., the position of the minimum value in front of each position in an array. By using two monotonic increasing stacks, the abnormal node position can be effectively found, and this position is the key point of optimization, i.e., the store optimization item.

[0083] 2.3 Factor sorting of store optimization items

[0084] A cause-effect tree model can be built based on multiple factors for modeling the store optimization item. The cause-effect tree model can be a LightGBM model (LGB model for short). In the specific implementation, the multiple factors can be selected from multiple search ranking features in an e-commerce search scenario. In the specific implementation, the cause-effect tree model can be trained based on a preset training set. The preset training set can be constructed according to store data of stores on a platform. For factors that cannot be changed, such as store types and statistical features, to avoid interference with the segmentation nodes of the cause-effect tree model and affect the feature importance effect, the factors will be deleted during training.

[0085] A data set is constructed based on store data (such as traffic data of a store and operation data of a store) of stores on a platform. The data set is different from the training set described above, and the data set can be referred to as a prediction set. After the cause-effect tree model is trained by using the training set, the data set is used to calculate the influence of each factor in the multiple factors on the prediction result of the cause-effect tree model, to obtain the influence degree corresponding to each factor in the multiple factors.

[0086] Specifically, the calculation of the influence degree can be specifically SHAP value (Shapley Additive exPlanations). SHAP is a "model explanation" package that can explain the output of any machine learning model. All features are regarded as "contributors". For each prediction sample, the model produces a prediction value, and the SHAP value is the value allocated to each feature in the sample. In the present disclosure, the SHAP value is the influence degree or contribution of each factor on the prediction result of the cause-effect tree model. The overall idea of calculating the SHAP value is to enumerate different factor combinations, calculate the influence of increasing a factor on the prediction result relative to not increasing the factor under a certain factor combination, and the influence is the marginal benefit of the factor for the factor combination, and the formula is as follows:

[0087] wherein V(i) represents the Shapley value of factor i in the multiple factors; N represents a set of the multiple factors; |N| represents the number of the multiple factors; S is a subset of N; |S| is the number of the subset of N; and v(S) represents the value of the multiple factors in the set S.

[0088] In the calculation of the Shapley value, all possible factor combination sets S are traversed, and the new contribution of factor i to the combination S is calculated, that is, v(S∪{i})-v(S). Then, the weighted average of the new contribution of all combinations is calculated, and the weight is determined according to the size of the combination.

[0089] The meaning of the above formula is that the Shapley value is equal to the average contribution of a certain factor in all possible cooperative combinations. It considers the marginal contribution of each factor and determines the final allocation through weighted average. The Shapley value of a factor is defined as the contribution of the factor to the prediction result of the causal tree model. In the case of a large number of factors, an approximate value of the Shapley value can be calculated using factor sampling or factor independence assumption.

[0090] Because the shop optimization items are modeled before, a causal tree model is constructed based on multiple factors; and then the SHAP value of each factor in the multiple factors is calculated by using the above algorithm. Therefore, after obtaining the SHAP values of the multiple factors, the multiple factors can be sorted according to the SHAP values; and then the factor sorting of the shop optimization items is obtained.

[0091] There are many shop optimization items, such as whether to use a set channel to improve the click rate, whether the keyword is accurate, etc. These are all shop optimization items. Through the above technical means, the factor sorting of each shop optimization item can be obtained.

[0092] Actually, shops belonging to different industries, shops at different levels (levels evaluated according to platform rating rules), and shops at different stages (such as growth stage or level upgrade preparation stage, etc.) have differences in shop operation. That is, the factor sorting of the same shop optimization item for shops in different industries, or different levels, or different stages will be different. For example, according to the first and second industries to which the shop belongs, the level to which the shop belongs, whether the shop advertises, whether the shop is in the growth stage, whether the shop is in the level upgrade preparation stage, etc., the shops on the platform are grouped.

[0093] The first and second industries are two levels of industry classification, which constitute a classification system from broad to specific. The first industry is the most basic level of industry classification, which can include agriculture, forestry, animal husbandry, fishing, industry, construction, etc. The second industry is a further subdivision of these first industries.

[0094] Therefore, the present disclosure proposes the concept of grouping stores on the platform, and calculating the factor ranking of the corresponding store optimization item for different groups of stores. That is, in the above technical solution, the data set is constructed based on the store data (such as traffic data and operation data of the store) of the stores belonging to the same group when constructing the data set. Then, the SHAP value of each factor calculated by using the data set constructed by the store data of the stores in the same group, and the factor ranking obtained based on the SHAP value ranking, is the factor ranking of the store optimization item of the group. In short, according to the above technical solution, the factor ranking of the store optimization item corresponding to different groups will be calculated. For the convenience of understanding, it can be represented as Table 1 as follows:

[0095] 2.4 Answer database

[0096] Respectively, the suggestion content associated with the factor. The suggestion content can be a task that the store can perform (such as changing keywords, placing advertisements, using auxiliary channels, etc.). Further, a copy template under different Query templates can also be constructed. For example, for service, advertising or transaction tasks, the copy template can stipulate that in addition to containing suggestion content, the comparison content of the target store and the reference store data should also be given, and so on. For goods, sub-account tasks, the copy template can stipulate that in addition to containing suggestion content, a list card of the optimization object (such as optimization target goods or sub-accounts, etc.) should also be given, and so on.

[0097] In summary, the copy database can store: attribute information of stores on the platform, traffic data and operation data of stores on the platform, store optimization items associated with Query templates, factor rankings of store optimization items, suggestion contents associated with factors, and corpus knowledge base of copy template information corresponding to different Query templates.

[0098] Three, model deployment (label 103 in FIG. 1)

[0099] 3.1, generating training samples

[0100] Based on the copy database obtained in the above (ii) and the Query template library obtained in the above (i), a third preset model is called to generate a plurality of training samples. The third preset model can be a large language model (LLM, Large Language Model) or the like, which is not limited in the present embodiment. For example, 4000 or more (such as 5000, 6000, etc.) valid data are generated offline by calling the third preset model as training samples. Some of the training samples contain sensitive data (such as merchant privacy information), and the sensitive information needs to be desensitized, such as using mask <mask>In the meantime, the first preset model output script can be in a specified format, such as a JSON format (JavaScript Object Notation, a lightweight data exchange format), by adjusting the prompt, so as to facilitate subsequent inter-group and jump connection access.

[0101] For example, it is explicitly stated in the prompt that the model is expected to return the script in JSON format. For example, a simple JSON format example is provided. An example of JSON output can be as follows: {"question": "What is the target question?", "answer": "This is the corresponding script."} Please provide your answer in this format.

[0102] 3.2, model adjustment training

[0103] Using the training samples generated in 3.1 above, the Fine Tuning (fine-tuning) pre-trained model can be trained using the fine-tuning method (such as the LoRA method) to enhance the logicality and accuracy of the pre-trained model output in the store operation scenario. The advantage of LoRA is that the LoRA method only trains a small number of parameters of the pre-trained model, and by adding a low-rank matrix in the attention layer of the Transformer, the model prediction accuracy is improved.

[0104] Among them, the pre-trained model can be a large language model (LLM). The pre-trained model has learned rich feature representations and pattern recognition capabilities, and through fine-tuning, these learned feature representations can be quickly and effectively adapted to specific scenarios (such as store operation scenarios). Fine-tuning is faster than training a model from scratch, which speeds up the training process.

[0105] Four, online link building (104 in FIG. 1)

[0106] In a specific implementation, the online deployment can be completed based on an Rlab (a high-level programming language for engineering application computing) tool. Of course, the online deployment can also be completed based on other tools, and the present disclosure does not make a specific limitation in this regard, and only the case of using the Rlab tool is exemplarily listed here. The online link building also needs to be supported by, but is not limited to, EAS, Igraph, a pre-trained model development tool (Tars Studio), PAI and Itag tools. Among them, EAS (Elastic Algorithm Service, model online service) is a model online service provided by the machine learning platform product for realizing one-stop model development and deployment application, supporting the deployment of model services in a public resource group or a dedicated resource group, realizing the real-time response of model loading and data request based on heterogeneous hardware (such as CPU or GPU). Igraph is an online graph storage and query system, which provides storage, query, update and calculation services for large-scale graph data. The pre-trained model development tool (Tars Studio) provides a visual development tool. PAI is a machine learning platform to enable developers to use artificial intelligence AI technology more efficiently, simply and standardly. Itag provides data labeling services to help users quickly and efficiently complete data labeling tasks.

[0107] The online built link can include but is not limited to:

[0108] S21, determining a target Query.

[0109] S22, rewriting the target Query.

[0110] The rewriting of the target Query can include but is not limited to:

[0111] Query normalization: rewriting the target Query through simple string processing, such as removing spaces, converting cases, processing special characters, etc.;

[0112] Query correction: actively correcting the target Query with spelling errors;

[0113] Query expansion: expanding according to the semantics or intent of the target Query, finding a group of Queries consistent with the target Query, and searching together with the target Query;

[0114] and so on.

[0115] S23, obtaining data based on Igraph.

[0116] The obtained data can be data retrieved from a script database. The data can include, but is not limited to, store data of the target store (including traffic data and operation data of the target store), store data of the reference store (including traffic data and operation data of the reference store), and script information. The script information can include, but is not limited to, a script template, a query template matched with the target query after the rewriting process in S22, a store optimization item associated with the matched query, at least one factor of the store optimization item, and suggested content associated with the factor.

[0117] S24, preprocessing the obtained data.

[0118] The present disclosure does not make specific limitations on "preprocessing", which can include, but is not limited to, data cleaning, standardization, format conversion, etc. For example, useful data can be filtered from the obtained data through a data cleaning process. For unstructured data (such as text), it may be necessary to perform structured processing through steps such as word segmentation and stem extraction. According to the input requirements of the pre-trained model, the data needs to be converted into an appropriate format, such as JSON format or text format.

[0119] S25, prompt concatenation

[0120] The matched query template and the data after the preprocessing can be concatenated into a prompt. The output (or response) quality of the first preset model (which can be a large language model fine-tuned) depends on the prompt (i.e., Prompt). In simple terms, Prompt is the question the user wants to ask, which is input into the first preset model. The first preset model will try to understand this input and then output an appropriate answer or response. Here, concatenating Prompt is to make the first preset model output more accurate and useful scripts.

[0121] Prompt is an important part. Prompt is used to guide the behavior of the model. By adjusting the Prompt, the quality of the model output script can be improved. In specific implementation, the designer can manually design the Prompt concatenation rule to concatenate according to the preset concatenation rule when performing Prompt concatenation.

[0122] Five, front-end function deployment

[0123] With the support of the technical architecture of the above-mentioned first, second, third, and fourth parts, corresponding function interfaces can be deployed on the front-end application (APP) or browser web page, so that users can call corresponding services of the back-end (such as a server with the above technical architecture) through the function interfaces, such as store diagnosis services, intelligent question and answer services, etc.

[0124] The form of the function interface on the APP page or the browser webpage can be a function control, and the user can call the corresponding service of the back end by clicking the function control. Alternatively, the APP page or the browser webpage is provided with a specific identifier, such as an intelligent assistant identifier. The identifier can be any pattern or animation, which is not limited here. The user can wake up the intelligent assistant function through voice or gesture, and the intelligent assistant identifier changes the display style to make the user perceive the successful wake-up of the intelligent assistant. The user can call the corresponding service of the back end by interacting with the intelligent assistant. That is, the product form presented by the above technical architecture on the client side can be an intelligent assistant identifier.

[0125] The store diagnosis service can include diagnosing the store and giving a store optimization item that is strongly related to a problem of the store when the problem exists. Further, the store diagnosis service can also give action suggestion content for improving the store optimization item, and the like.

[0126] The intelligent question and answer service can include responding to a behavior suggestion question input by the user and outputting an action strategy and a target corresponding to the behavior suggestion question, and the like.

[0127] In summary, the technical solution provided by the present disclosure has at least the following technical effects:

[0128] 1. The technical solution provided by the present disclosure can rewrite and polish the output script by adjusting and training the pre-training model suitable for the downstream task (such as the intelligent response task in the store operation scenario), so as to realize content personalization.

[0129] 2. A plurality of factors are taken as starting points to respectively establish a causal tree model corresponding to different store optimization items, and then the factor ranking of different store optimization items is determined. The factor association has suggestion content, such as the suggestion content of the task that can be executed by the merchant.

[0130] 3. Because there are differences between different industries and different levels of stores, the factor ranking of the same store optimization item corresponding to different stores in different industries is different for the same problem (such as how to improve the access volume). For example, the factor ranking corresponding to two stores in the mechanical equipment industry and the clothing industry is different when solving the access volume problem. Therefore, the technical solution of the present disclosure groups the stores on the platform; when determining the factor ranking of different store optimization items, the factor ranking of different store optimization items in different groups is determined respectively according to the grouping.

[0131] The technical solution provided by the present disclosure combines a pre-training model with a causal inference technology, and realizes the landing of an e-commerce scenario. Specifically, the present disclosure constructs a Query template library suitable for a shop operation scenario; and further constructs a copy database suitable for the shop operation scenario based on the Query template library. Moreover, when constructing the copy database, the causal inference technology is used to attribute analysis at least one factor of each shop operation related problem, wherein the factor is associated with a suggestion content. In this way, a training sample set can be generated based on the Query template library and the copy database, and the pre-training model is adjusted and trained based on the training sample set to enhance the logicality and accuracy of the output copy in the shop operation scenario.

[0132] The technical solution corresponding to the technical architecture mentioned above (such as the solutions corresponding to the first, second, third and fourth parts in FIG. 1) can be deployed on a service side. The service side can be a server, a service cluster, a virtual server or a cloud, etc., and the present embodiment does not make specific limitation thereon. The service side provides corresponding function services for the client side, such as shop diagnosis, intelligent question and answer services, etc. The user can enter the management page provided by the platform to the user through the browser or the client application (APP) on the client device. The client device can be, but is not limited to, a smart phone, a smart wearable device, a tablet computer, a notebook computer, a desktop computer, etc. As shown in FIG. 2, an embodiment of the present disclosure provides a service system, which includes a service side and a client side. Wherein,

[0133] The service side 11 is configured to construct a Query template library, wherein the Query template library includes a plurality of Query templates, and one Query template is associated with one preset problem related to shop operation; based on the Query template library, a copy database is constructed; wherein the copy database stores at least one factor of the Query template obtained by using a causal inference algorithm to attribute analysis on the preset problem associated with the Query template; a training sample set is generated based on the Query template library and the copy database; and the pre-training model is adjusted and trained based on the training sample set to obtain a first preset model capable of processing intelligent response tasks in the shop operation scenario.

[0134] The client side 12 is configured to deploy a corresponding interface, so that the user can call the service side through the interface, retrieve target copy data in the copy database, and output the copy corresponding to the copy data by using the first preset model.

[0135] Further, the service side 11 and the client side 12 included in the service system are further configured to implement the following functions. Specifically:

[0136] The client 12 is further configured to trigger store diagnosis to present and / or play the diagnosed target problem in response to a diagnosis trigger event for the target store, and send information about the target problem to the server; or display a management page of the target store; listen to operations of a user on the management page to capture a target problem of the user on store operation; and send information about the target problem to the server in response to a confirmation instruction triggered by the user for the target problem.

[0137] The server 11 is further configured to determine a target problem for a target store; retrieve at least one factor of the target problem in a script database; generate a prompt word Prompt based on suggested content associated with the target problem and the at least one factor; and output a script corresponding to the Prompt by using a first preset model to help the user solve or improve the target problem.

[0138] The client 12 is further configured to receive the script fed back by the server and display and / or play the script.

[0139] It should be noted that the execution subject of the above-mentioned "store diagnosis" can be the client 12, which can obtain store data of a reference store and / or historical store data of the target store from the server 11 side after triggering store diagnosis, and then call a diagnosis program locally stored in the client 12 to perform store diagnosis on the target store. Alternatively, the execution subject of the "store diagnosis" can be the server 11. After the client 12 triggers store diagnosis, the client 12 sends a target store diagnosis request to the server 11, so that the server 11 obtains current store data, historical store data of the target store, store data of a reference store, and the like, and performs diagnosis on the target store based on the obtained data to determine a target problem. The server 11 feeds back the determined target store to the client 12.

[0140] It can be seen that the service system provided by the embodiments of the present disclosure uses the powerful computing power of the server 11 side to construct a Query template library and find at least one factor causing a store problem by using a causal inference technology, and further realizes the logic and accuracy of outputting scripts in the store operation scenario by using a large model (i.e., a first preset model), to help the user (or store merchant) grow.

[0141] It should be noted that the above is an introduction to the technical solutions of the present disclosure by taking the store operation scenario as an example. In fact, the technical architecture provided by the present disclosure can also be applied to other specific scenarios, such as enterprise management scenarios, project management scenarios, and the like.

[0142] The technical solutions provided by the present disclosure will be described below in the form of method embodiments.

[0143] FIG. 3 shows a flowchart of an information processing method according to an embodiment of the present disclosure. The execution subject of the method provided in the embodiment can be the server in the service system described above. As shown in FIG. 3, the method can include:

[0144] 301. Determine a target problem for a target store.

[0145] 302. Retrieve at least one factor of the target problem from a script database.

[0146] 303. Generate a prompt based on the target problem and the at least one factor associated with the recommended content.

[0147] 304. Output the script corresponding to the prompt using a first preset model to help the user solve or improve the target problem.

[0148] Wherein, the script database stores at least one factor of a preset problem obtained by attributing analysis of the preset problem using a causal inference algorithm; the first preset model is obtained by adjusting and training a pre-training model for intelligent response tasks in a store operation scenario based on the pre-training model.

[0149] The above technical architecture lists an example of constructing a Query template library. In fact, the Query template corresponds to the problem. That is, the Query template can be directly understood as a problem template corresponding to it. Or, in specific implementation, the above construction of the Query template library can also be "construction of the problem library".

[0150] The above 301 can determine the target problem of the target store by diagnosing the target store, or determine the target problem based on user operation. Among them, determining the target problem based on user operation can include but is not limited to: determining the target problem existing in the target store based on user input information, or determining the target problem of the target store based on the request triggered by the user for a setting item, or recommending at least one recommended problem based on big data analysis on the network side, and the user selects one from them. The selected recommended problem is the target problem.

[0151] That is, the above 301 "determining a target problem for a target store" can include:

[0152] 3011. In response to user input, determine the target problem for the target store based on user input information; and / or

[0153] 3012. In response to a request triggered by a user for a setting item on a management page of a target store, determine the target problem for the setting item; and / or

[0154] 3013. diagnosing the target store, and determining the target problem based on the diagnosis result; and / or

[0155] 3014. obtaining at least one recommended problem recommended by the network side based on big data analysis, and in response to the selection event, selecting one of the recommended problems as the target problem.

[0156] In one embodiment, the above-mentioned 3013 "diagnosing the target store, and determining the target problem based on the diagnosis result" can include:

[0157] diagnosing the gap between the target store and the reference store, and / or the data difference between different time periods of the target store, and / or the gap between the store data of the target store and the set index, to obtain a diagnosis result;

[0158] determining the target problem based on the diagnosis result.

[0159] The reference store can be a store that belongs to the same industry as the target store and has store data at the median or high end of all stores in the industry. Alternatively, the reference store can be a store that belongs to the same industry as the target store, is of the same level, and has store data at the median or high end of all stores in the industry. Alternatively, the reference store can be a store that belongs to the same group as the target store, is of the same level, and has store data at the median or high end of all stores in the industry. Alternatively, the reference store can be a store specified by the user, and so on.

[0160] The above-mentioned "data difference between different time periods of the target store" can be a data difference between store data of the target store in a first time period and store data of the target store in a second time period. The second time period can not overlap the first time period, and the second time period can be earlier than the first time period. Alternatively, the second time period can partially overlap the first time period, the start time of the second time period can be earlier than the start time of the first time period, and the end time of the second time period can be earlier than the end time of the first time period. Alternatively, the second time period can include the first time period, i.e., the start time of the first time period is a time point within the second time period, and the end time of the first time period is the same as the end time of the second time period.

[0161] The above-mentioned "set index" can be a target index set by the user. For example, the user expects the future store visit volume to increase by 10%, and this increase of 10% is the set index. Alternatively, the above-mentioned "set index" can also be an index possessed by the reference store. The reference store is described above.

[0162] The indicators can include store indicators and product indicators. The store indicators include, but are not limited to, store page view, conversion rate, fan activity, member activity, average single price (average purchase amount of each product purchased by a user), and the like. The conversion rate includes search conversion rate, UV (unique visitor, referring to a person accessing and browsing a store page through the Internet) order conversion rate, and the like. The product indicators can include, but are not limited to, product attention, product add-to-cart rate, product conversion rate, and the like.

[0163] From the above, it can be seen that the copy database stores at least one factor of the Query template obtained by using a causal inference algorithm to perform attribution analysis on a preset problem associated with the Query template. Correspondingly, the step 302 "retrieving at least one factor of the target problem in the copy database" can include the following steps:

[0164] 3021. determining a target Query according to the target problem;

[0165] 3022. retrieving a Query template matching the target Query in the copy database, and obtaining at least one factor of the matched Query template;

[0166] Wherein, the suggestion content associated with the target Query and at least one factor of the matched Query template is generated when generating the Prompt (i.e., the step 303).

[0167] It is mentioned above that a Query template library is pre-constructed, and the Query template library includes a plurality of Query templates. Correspondingly, one implementation scheme of the step 3021 "determining a target Query according to the target problem" is:

[0168] According to the target problem, the plurality of Query templates in the Query template library are sorted;

[0169] The first Query template in the sorting is used to determine the target Query, or the first number of Query templates in the front of the sorting are used as candidate Querys for the user to select, and the target Query is determined based on the candidate Query selected by the user.

[0170] In a specific implementation, the first number of Query templates in the front of the sorting can be 2, 3, or 5, and the like, which is not limited in the embodiment.

[0171] Further, the copy database stores: a store optimization item and at least one factor of the store optimization item, which can solve or improve a preset problem associated with the Query template, obtained by using a causal inference algorithm to perform attribution analysis on the preset problem.

[0172] retrieving, in the copy database, a Query template matching the target Query;

[0173] obtaining a store optimization item associated with the matched Query template and at least one factor of the store optimization item;

[0174] wherein the suggestion content is generated based on the target Query, the store optimization item associated with the matched Query template, and at least one factor of the store optimization item when generating the Prompt (i.e., step 303).

[0175] In constructing the copy database, the present disclosure uses a causal inference algorithm. That is, in the scheme provided by the present embodiment, when constructing the copy database, the following steps are included: using a causal inference algorithm to perform attribution analysis on a preset problem associated with a Query template, to obtain a store optimization item and at least one factor of the store optimization item that can solve or improve the preset problem. This part is an important content of the present embodiment. One implementable way is that the step "using a causal inference algorithm to perform attribution analysis on a preset problem associated with a Query template, to obtain a store optimization item and at least one factor of the store optimization item that can solve or improve the preset problem" can include the following steps:

[0176] 305. For a preset problem associated with a Query template, using funnel analysis technology to analyze traffic data and operation data of stores on the platform to analyze a store optimization item that can solve or improve the preset problem;

[0177] 306. Modeling with the store optimization item as the target, and constructing a causal tree model based on multiple factors;

[0178] 307. Constructing a data set based on traffic data and operation data of stores on the platform;

[0179] 308. Using the data set to calculate the influence of each factor in the multiple factors on the prediction result of the causal tree model, to obtain the corresponding influence degree of each factor in the multiple factors;

[0180] 309. Based on the influence of each of the multiple factors, sort the multiple factors to obtain the factor ranking of the store optimization items;

[0181] Wherein, at least one factor of the store optimization item is: the second-to-last factor in the factor ranking of the store optimization item.

[0182] For details on the implementation of steps 305 to 309 above, please refer to the description above, which will not be repeated here.

[0183] Furthermore, considering the differences between stores in different industries and at different levels, the factor ranking (factor priority) corresponding to the same optimization item for a store is different; the method provided in this embodiment also groups the stores on the platform to determine the factor ranking of store optimization items for different groups. That is, the method provided in this disclosure embodiment may further include the following steps:

[0184] 310. Group the stores on the platform;

[0185] 311. When constructing the dataset, it is built based on the traffic and operational data of stores within a group; and

[0186] 312. Using the factor ranking of the store optimization items obtained from the dataset, the factor ranking of the store optimization items corresponding to the group is obtained.

[0187] Specifically, in the step of "obtaining the store optimization items associated with the matched query template and at least one factor of the store optimization items" (i.e., a subordinate step included in step 3022), the group to which the target store belongs is determined; the store optimization items associated with the matched query template and at least one factor of the store optimization items corresponding to the group to which the target store belongs are obtained.

[0188] The method provided in this embodiment is written from the perspective of how to apply the support provided by the technical architecture shown in Figure 1 to help users solve or improve store problems. In fact, the method provided in this embodiment also includes the following steps, which are part of the preliminary deployment tasks. Specifically, the method provided in this embodiment may further include the following steps:

[0189] 313. Construct a query template library, which includes multiple query templates;

[0190] 314. Based on the Query template library, construct the copywriting database;

[0191] 315. Generate the training sample set based on the Query template library and the copywriting database;

[0192] 316、using the training sample set, performing adjustment training on the pre-training model.

[0193] For specific implementation of steps 313-316, refer to the foregoing description, which will not be repeated here.

[0194] In an implementable embodiment, the Query template library comprises at least one of the following: a Query template corresponding to a reason for data difference between any store on the platform and a reference store, a Query template corresponding to a reason for data difference between any store on the platform in different periods, a Query template corresponding to a store management problem, a Query template corresponding to a store optimization item, and a Query template extended by the second preset model.

[0195] The copywriting database stores: attribute information of stores on the platform, traffic data and operation data of stores on the platform, store optimization items associated with Query templates, factor sorting of store optimization items, suggestion content associated with factors, and a corpus knowledge base of copywriting template information corresponding to different Query templates.

[0196] The technical solution provided in this embodiment combines a large model with causal inference technology, that is, a causal inference algorithm is used when building the copywriting database to perform attribution analysis on the preset problems associated with the Query templates to obtain at least one factor. The training samples used to perform adjustment training on the pre-training model are generated based on the Query template library and the copywriting database. A training sample set composed of multiple such training samples is used to perform adjustment training on the pre-training model to adjust it into a first preset model suitable for the store operation scenario. The copywriting database provides data support, and at least one factor of a store diagnosis result and / or a question asked by a user and suggestion content associated with the factor can be retrieved. The first preset model that has been adjusted and trained provides intelligent response support. Based on the store diagnosis result and / or the question asked by the user and the suggestion content associated with the at least one factor, the first preset model can output copywriting that is more in line with natural language logic and suitable for the store operation scenario, to help the user operate the store.

[0197] Another embodiment of the present disclosure also provides an information processing method, more specifically, a model training method. The execution subject of the method provided in the embodiment of the present disclosure can be a server in the service system. Specifically, as shown in FIG. 4, the method comprises:

[0198] 401、constructing a Query Query template library, the Query template library comprising a plurality of Query templates, one Query template being associated with one preset problem related to store operation.

[0199] 402、constructing a script database based on the Query template library, wherein the script database stores at least one factor of the Query template obtained by performing attribution analysis on a preset problem associated with the Query template by using a causal inference algorithm.

[0200] 403、generating a training sample set according to the Query template library and the script database.

[0201] 404、adjusting and training a pre-trained model by using the training sample set to obtain a first preset model capable of processing intelligent response tasks in a store operation scenario.

[0202] In the step 402 of constructing the script database, the at least one factor of the Query template obtained by performing attribution analysis on a preset problem associated with the Query template by using a causal inference algorithm can include:

[0203] 4021, for a preset problem associated with a Query template, analyzing traffic data and operation data of stores on a platform by using funnel analysis technology to analyze store optimization items for solving or improving the preset problem;

[0204] 4022, modeling the store optimization items as targets and constructing a causal tree model based on multiple factors;

[0205] 4023, constructing a data set based on traffic data and operation data of stores on a platform;

[0206] 4024, calculating the influence of each factor in the multiple factors on the prediction result of the causal tree model by using the data set to obtain an influence degree corresponding to each factor in the multiple factors;

[0207] 4025, sorting the multiple factors according to the influence degree corresponding to each factor to obtain a factor sorting of the store optimization items.

[0208] Further, the method provided in the embodiment can further include the following steps:

[0209] 405, grouping stores on a platform.

[0210] 406, constructing the data set based on traffic data and operation data of stores in a group.

[0211] 407, obtaining a factor sorting of store optimization items by using the data set, and sorting the store optimization items corresponding to the group.

[0212] In one specific embodiment, the step 401 of "constructing a script database based on the Query template library" can include:

[0213] obtaining and storing attribute information of the stores on the platform;

[0214] obtaining and storing traffic data and operation data of the stores on the platform;

[0215] performing attribution analysis on the preset questions associated with the Query templates in the Query template library by using a causal inference algorithm to obtain at least one factor of the Query template;

[0216] determining and storing the suggestion content associated with the factor;

[0217] constructing a corpus knowledge base of the copy template information corresponding to different Query templates.

[0218] In one specific embodiment, the step 403 "generating a training sample set according to the Query template library and the copy database" can include:

[0219] outputting a plurality of training samples based on the Query template library and the copy database by using a third preset model;

[0220] The training sample set includes the plurality of training samples.

[0221] The step 401 "constructing a Query template library" can include:

[0222] 4011. determining a plurality of basic Query templates;

[0223] 4012. extending the plurality of basic Query templates by using a second preset model to obtain a plurality of extended Query templates;

[0224] 4013. evaluating the plurality of extended Query templates, and retaining the extended Query templates that pass the evaluation;

[0225] The Query template library includes the plurality of basic Query templates and the extended Query templates that pass the evaluation.

[0226] For specific contents of each step in this embodiment, please refer to the description above, which will not be repeated here.

[0227] The present disclosure also provides a store diagnosis method. The execution subject of the method provided in this embodiment can be a client in the service system described above. As shown in FIG. 5, the method includes:

[0228] 501. In response to a diagnosis triggering event for a target store, triggering store diagnosis to present and / or play the diagnosed target problems;

[0229] 502, display and / or play the script, the script including suggestion content for assisting the user to solve or improve the target problem;

[0230] The script is obtained by using a first preset model, the first preset model being obtained by adjusting and training a pre-training model for an intelligent response task in a shop operation scenario; the suggestion content is associated with at least one factor of the target problem, the at least one factor being obtained by performing attribution analysis on the target problem by using a causal inference algorithm.

[0231] Referring to Figs. 6a-6c, the diagnosis triggering event for the target shop can include:

[0232] In response to the user's operation on the diagnosis control on the interface, the diagnosis triggering event is generated; and / or

[0233] In response to the user's shop diagnosis voice, the diagnosis triggering event is generated; and / or

[0234] In response to the user's input of a shop diagnosis instruction through the intelligent assistant interaction interface, the diagnosis triggering event is generated.

[0235] Referring to the interface example shown in Fig. 6a, the interface can be a shop management interface, and the interface displays a diagnosis control, such as "shop diagnosis" shown in Fig. 6a. The user clicks the "shop diagnosis" control to trigger diagnosis.

[0236] Referring to the interface example shown in Fig. 6b, the user can first wake up the intelligent assistant through voice (such as "hello, intelligent assistant"). At this time, the interface can display an intelligent assistant identifier 1. Then the user can issue a "shop diagnosis" voice to trigger diagnosis. The intelligent assistant identifier can be of any shape or pattern, which is not limited in the embodiment.

[0237] Referring to the interface example shown in Fig. 6c, the display screen displays an intelligent assistant interaction interface 2, and the intelligent interaction interface 2 displays an input box 3. The user can input "shop diagnosis" in the input box 3, and then click a confirmation or sending control 4 (such as the "send" control in Fig. 6c) to trigger shop diagnosis.

[0238] ​The execution subject of the above-mentioned "store diagnosis" can be a client. After triggering the store diagnosis, the client can obtain the store data of the reference store and / or the historical store data of the target store from the server side, and then call the diagnosis program locally to diagnose the target store. Alternatively, the execution subject of the "store diagnosis" is the server. After the client triggers the store diagnosis, the client sends a target store diagnosis request to the server, so that the server obtains the current store data, historical store data of the target store, store data of the reference store, etc., and diagnoses the target store based on the obtained data to determine the target problem. The server feeds back the determined target store to the client.

[0239] Specifically, "diagnosing the target store" can include diagnosing the gap between the target store and the reference store, and / or the data difference between different periods of the target store, and / or the gap between the store data of the target store and the set index, to obtain a diagnosis result.

[0240] The script displayed and / or played in the above-mentioned 502 can be fed back by the server, and the specific content can be referred to the description above, which will not be repeated here.

[0241] The present disclosure also provides a store intelligent assistant working method. The execution subject of the method provided by the present embodiment can be a client in the above-mentioned service system. As shown in FIG. 7, the method comprises:

[0242] 601, displaying a management page of a target store;

[0243] 602, listening to the operation of the user on the management page to capture the target problem of the user in the store operation;

[0244] 603, in response to the confirmation instruction triggered by the user for the target problem, displaying and / or playing a script, the script including suggestion content for helping the user to solve or improve the target problem;

[0245] The script is obtained by using a first preset model, and the first preset model is obtained by adjusting and training a pre-training model for an intelligent response task in a store operation scenario. The suggestion content is associated with at least one factor of the target problem, and the at least one factor is obtained by attributing analysis of the target problem by using a causal inference algorithm.

[0246] In the above-mentioned 602, listening to the operation of the user on the management page to capture the target problem of the user in the store operation can include:

[0247] 6021, in response to the input operation of the user on the management page, determining the target problem based on the user input information; and / or

[0248] 6022、in response to a touch operation of a user on a control on the management page, determining a target question (such as how to set the setting item, etc.) according to a setting item associated with the control; and / or

[0249] 6023、monitoring a stay event of a mouse cursor at a content item on the management page, and determining a target question based on the content item; and / or

[0250] 6024、listening to a shop data viewing operation, and determining a target question according to shop data.

[0251] In the above 6021, the user can directly input question information in the intelligent assistant interaction area on the management page.

[0252] In the above 6022, for example, the setting item can be a keyword setting item. The user can set a keyword for the product at the keyword setting item. As shown, a control associated with the setting item is provided near the keyword setting item on the management page. After the user touches the control, the target question can be determined as "how to set the keyword?"

[0253] In the above 6023, the user stays at a certain content item for a long time, and the possibility of having doubts is greater. Therefore, when the mouse cursor stays at a content item for more than a certain time (such as 10s, 30s, etc.), the target question can be determined based on the content item.

[0254] In the above 6024, the user views the recent access of the shop through the management page, and then the target question can be determined according to the shop access currently seen by the user, such as "why is the access reduced?"

[0255] Further, the method provided by the embodiment can further include the following steps:

[0256] 604、displaying an intelligent assistant identifier on the management page corresponding to the target shop;

[0257] 605、in response to an intelligent assistant wake-up operation, presenting an intelligent assistant page containing an input box and at least one recommended query; wherein the at least one recommended query is obtained by diagnosing the target shop.

[0258] The user can input a target question through the input box, and also interact with the intelligent assistant to obtain required data, question solutions, etc.

[0259] The methods provided in the above two embodiments are written from the perspective of an intelligent assistant corresponding to the product form of the technical solutions provided by the present disclosure. Based on the technical support provided by the above server, the intelligent assistant can help merchants improve the efficiency of analysis, provide effective strategies, and grasp market opportunities to help merchants grow better. The intelligent assistant can diagnose problems of a store, analyze store optimization items and at least one factor of the store optimization items for solving or improving the problems by using a causal inference technology, and then output response scripts in accordance with human language logic and store operation scenarios by means of a large model to help the merchants grow.

[0260] An embodiment of the present disclosure further provides an information processing device. The information processing device comprises a determination module, a retrieval module, a generation module, and a calculation module. The determination module is configured to determine a target problem for a target store. The retrieval module is configured to retrieve at least one factor of the target problem in a script database. The generation module is configured to generate a prompt word (Prompt) based on suggestion content associated with the target problem and the at least one factor. The calculation module is configured to output a script corresponding to the Prompt by using a first preset model to help a user solve or improve the target problem.

[0261] The script database stores at least one factor of a preset problem obtained by performing attribution analysis on the preset problem by using a causal inference algorithm. The first preset model is obtained by adjusting and training a pre-training model for an intelligent response task in a store operation scenario on the basis of the pre-training model.

[0262] Further, the script database stores at least one factor of a Query template associated with a preset problem obtained by performing attribution analysis on the preset problem by using a causal inference algorithm. The retrieval module is configured to determine a target Query according to the target problem, retrieve a Query template matching the target Query in the script database, and obtain at least one factor of the matching Query template. The suggestion content associated with the target Query and the at least one factor of the matching Query template is used to generate the Prompt.

[0263] Further, when determining the target Query according to the target problem, the determination module is configured to sort a plurality of Query templates in a Query template library according to the target problem, determine the target Query based on a first Query template in the sorting, or select a first number of Query templates in the front of the sorting as candidate Queries for a user to select, and determine the target Query based on one candidate Query selected by the user.

[0264] Further, the copy database stores: a shop optimization item and at least one factor of the shop optimization item obtained by using a causal inference algorithm to perform attribution analysis on a preset problem associated with a Query template, and capable of solving or improving the preset problem. The retrieval module, when retrieving, in the copy database, the Query template matching the target Query and obtaining at least one factor of the matched Query template, is specifically configured to: retrieve, in the copy database, the Query template matching the target Query; obtain the shop optimization item associated with the matched Query template and at least one factor of the shop optimization item; and generate the suggestion content based on the target Query, the shop optimization item associated with the matched Query template, and at least one factor of the shop optimization item when generating the Prompt.

[0265] Further, the device provided in the embodiment further includes a copy database construction module. The copy database construction module can use a causal inference algorithm to perform attribution analysis on a preset problem associated with a Query template, and obtain a shop optimization item and at least one factor of the shop optimization item that can solve or improve the preset problem. Specifically, the copy database construction module can be specifically configured to:

[0266] For a preset problem associated with a Query template, a funnel analysis technology is used to analyze traffic data and operation data of shops on a platform, to analyze a shop optimization item that can solve or improve the preset problem; a causal tree model is constructed based on the shop optimization item as a target and multiple factors; a data set is constructed based on traffic data and operation data of shops on the platform; the data set is used to calculate an influence degree of each factor in the multiple factors on a prediction result of the causal tree model; the multiple factors are sorted according to the influence degree of each factor, to obtain a factor sorting of the shop optimization item; and at least one factor of the shop optimization item is the first second number of factors in the factor sorting of the shop optimization item.

[0267] Further, the copy database construction module in the device provided in the embodiment is further configured to: group shops on the platform; construct, when constructing the data set, based on traffic data and operation data of shops in a group; and obtain, using the data set, a factor sorting corresponding to the shop optimization item, as a factor sorting of the shop optimization item corresponding to the group. When obtaining the shop optimization item associated with the matched Query template and at least one factor of the shop optimization item, the group to which the target shop belongs is determined; the shop optimization item associated with the matched Query template and at least one factor of the shop optimization item corresponding to the group to which the target shop belongs are obtained.

[0268] Further, the device provided by the embodiment further comprises a Query template library construction module, a sample generation module and an adjustment training module. The Query template library construction module is configured to construct a Query template library, wherein the Query template library comprises a plurality of Query templates. The copy material construction module is configured to construct the copy material library based on the Query template library. The sample generation module is configured to generate the training sample set according to the Query template library and the copy material library. The adjustment training module is configured to adjust and train the pre-training model by using the training sample set.

[0269] Further, the determination module is specifically configured to:

[0270] determine a target problem for the target store based on user input information in response to user input; and / or

[0271] determine the target problem for a setting item in response to a request triggered by the user for the setting item on a management page of the target store; and / or

[0272] diagnose the target store and determine the target problem based on a diagnosis result; and / or

[0273] obtain at least one recommended problem recommended by a network side based on big data analysis, and select one of the recommended problems as the target problem in response to a selection event.

[0274] In the process of diagnosing the target store and determining the target problem based on a diagnosis result, the determination module is specifically configured to: diagnose a gap between the target store and the reference store, and / or a data difference existing in different periods of the target store, and / or a gap between store data of the target store and a set index, to obtain a diagnosis result; and determine the target problem based on the diagnosis result.

[0275] Another embodiment of the present disclosure provides an information processing device, comprising: a query template library construction module, a copy library construction module, a sample generation module, and an adjustment training module. The query template library construction module is configured to construct a query template library, wherein the query template library comprises a plurality of query templates, and each query template is associated with a preset question related to store operation. The copy library construction module is configured to construct a copy library based on the query template library, wherein the copy library stores at least one factor of a query template, which is obtained by using a causal inference algorithm to perform attribution analysis on a preset question associated with the query template. The sample generation module is configured to generate a training sample set according to the query template library and the copy library. The adjustment training module is configured to adjust and train a pre-trained model using the training sample set to obtain a first preset model capable of processing intelligent response tasks in a store operation scenario.

[0276] In a specific embodiment, when the copy library construction module constructs the copy library and uses a causal inference algorithm to perform attribution analysis on a preset question associated with a query template to obtain at least one factor of the query template, the copy library construction module is specifically configured to:

[0277] For a preset question associated with a query template, a funnel analysis technique is used to analyze traffic data and operation data of stores on a platform to analyze store optimization items for solving or improving the preset question. A causal tree model is constructed based on the store optimization items as a target and a plurality of factors. A data set is constructed based on traffic data and operation data of stores on a platform. The influence of each factor in the plurality of factors on the prediction result of the causal tree model is calculated using the data set to obtain an influence degree corresponding to each factor in the plurality of factors. The plurality of factors are sorted according to the influence degree corresponding to each factor to obtain a factor sorting of the store optimization items.

[0278] Further, the copy library construction module is further configured to: group stores on a platform; construct a data set based on traffic data and operation data of stores in a group when constructing the data set; and use the factor sorting of store optimization items obtained from the data set to sort factors of store optimization items corresponding to the group.

[0279] Further, the copy database construction module is configured to: acquire and store attribute information of the stores on the platform; acquire and store traffic data and operation data of the stores on the platform; perform attribution analysis on preset problems associated with the Query templates in the Query template library by using a causal inference algorithm to obtain at least one factor of the Query templates; determine and store suggestion content associated with the factors; and construct a corpus knowledge base of copy template information corresponding to different Query templates.

[0280] The sample generation module is configured to: output a plurality of training samples based on the Query template library and the copy database by using a third preset model, when generating a training sample set according to the Query template library and the copy database; and wherein the training sample set includes the plurality of training samples.

[0281] Further, the Query template library construction module is configured to: determine a plurality of basic Query templates; perform extension on the plurality of basic Query templates by using a second preset model to obtain a plurality of extended Query templates; and perform evaluation on the plurality of extended Query templates, and retain the extended Query templates that pass the evaluation; and wherein the Query template library includes the plurality of basic Query templates and the plurality of extended Query templates that pass the evaluation.

[0282] It should be noted that the information processing apparatus provided in the above two embodiments can implement the technical solutions described in the above corresponding method embodiments, and the principles of implementation of the above modules or units can be found in the corresponding content in the above method embodiments, which will not be described here again.

[0283] An embodiment of the present disclosure provides a store diagnosis apparatus, which includes a triggering module and an output module. The triggering module is configured to trigger store diagnosis in response to a diagnosis triggering event for a target store to present and / or play a diagnosed target problem. The output module is configured to display and / or play a copy, which includes suggestion content for helping a user to solve or improve the target problem. The copy is obtained by using a first preset model, which is obtained by adjusting and training a pre-training model for an intelligent response task in a store operation scenario. The suggestion content is associated with at least one factor of the target problem, which is obtained by performing attribution analysis on the target problem by using a causal inference algorithm.

[0284] It should be noted that the store diagnosis apparatus provided in the above embodiments can implement the technical solutions described in the corresponding method embodiments, and the principles of the implementation of the above modules or units can be referred to the corresponding content in the method embodiments, which will not be described here.

[0285] The store intelligent assistant apparatus provided in an embodiment of the present disclosure includes a display apparatus, a listening and capturing module, and an output module. The display module is configured to display a management page of a target store. The listening and capturing module is configured to listen to an operation of a user on the management page to capture a target problem of the user on store operation. The output module is configured to, in response to a confirmation instruction triggered by the user for the target problem, display and / or play a script, the script including suggestion content for assisting the user to solve or improve the target problem.

[0286] The script is obtained by using a first preset model, the first preset model is obtained by adjusting and training a pre-training model for an intelligent response task in a store operation scenario on the basis of the pre-training model, and the suggestion content is associated with at least one factor of the target problem, the at least one factor being obtained by performing attribution analysis on the target problem by using a causal inference algorithm.

[0287] It should be noted that the store intelligent assistant apparatus provided in the above embodiments can implement the technical solutions described in the corresponding method embodiments, and the principles of the implementation of the above modules or units can be referred to the corresponding content in the method embodiments, which will not be described here.

[0288] An electronic device is also provided in an embodiment of the present disclosure. As shown in FIG. 8, the electronic device includes a processor 42 and a memory 41. The memory 41 is configured to store one or more computer programs, and the processor 42 is coupled with the memory 41 and configured to execute the one or more computer programs to implement the steps in the method provided in the embodiments of the present disclosure.

[0289] The memory 41 can be implemented by any type of volatile or nonvolatile storage devices or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk, or an optical disk.

[0290] Further, the electronic device also includes a communication component 43, a display 44, a power component 45, an audio component 46, and other components. Here, only some components are shown schematically, and the electronic device does not necessarily include only these components.

[0291] Correspondingly, the embodiments of the present disclosure further provide a computer readable storage medium, which stores a computer program. The computer program is executed by a computer and can implement the method steps or functions provided by the embodiments of the present disclosure.

[0292] The method in the present disclosure can be implemented by software, hardware, firmware or any combination thereof, in whole or in part. When implemented by software, it can be implemented in the form of a computer program product, in whole or in part. Thus, the present disclosure also provides a computer program product. The computer program product includes computer programs / instructions, which, when executed by an electronic element such as a processor, can execute the steps or functions in the method provided by the embodiments of the present disclosure, in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user equipment, a core network device, an OAM or other programmable devices.< / mask>

Claims

1. An information processing method, wherein, include: For the target store, identify the target problem; Retrieve at least one factor of the target question from the document database; Based on the target question and the suggested content associated with at least one factor, a prompt word is generated. The first preset model is used to output the text corresponding to the Prompt, so as to help users solve or improve the target problem; The document database stores at least one factor of the preset question obtained by attribution analysis of the preset question using a causal inference algorithm; the first preset model is obtained by adjusting and training the pre-trained model based on the pre-trained model for intelligent response tasks in the store operation scenario.

2. The method according to claim 1, wherein, The document database stores at least one factor of the query template obtained by attribution analysis of preset questions associated with the query template using a causal inference algorithm. as well as Retrieve at least one factor of the target question from the document database, including: Based on the target question, determine the target query; In the document database, retrieve query templates that match the target query and obtain at least one factor of the matched query template; Specifically, when generating the Prompt, suggested content is generated based on at least one factor associated with the target Query and the matched Query template.

3. The method according to claim 2, wherein, The target query is determined based on the target question, including: Based on the target problem, sort the multiple query templates in the query template library; The target query is determined based on the first ranked query template, or the first number of ranked query templates are provided as candidate queries for the user to choose from, and the target query is determined based on the candidate query selected by the user.

4. The method according to claim 2 or 3, wherein, The copywriting database stores: store optimization items that can solve or improve the preset problems obtained by using a causal inference algorithm to perform attribution analysis on preset problems associated with the query template, and at least one factor of the store optimization items. as well as In the document database, retrieve query templates that match the target query, and obtain at least one factor of the matched query templates, including: Retrieve a query template that matches the target query from the document database; Obtain the store optimization items associated with the matched query template, and at least one factor of the store optimization items; Specifically, when generating the Prompt, the suggested content is generated based on the target Query, the store optimization items associated with the matched Query template, and at least one factor associated with the store optimization items.

5. The method according to claim 4, wherein, Using a causal inference algorithm, attribution analysis is performed on a preset problem associated with a query template to obtain store optimization items that solve or improve the preset problem, and at least one factor of the store optimization items, including: To address the pre-defined issues associated with query templates, funnel analysis techniques are used to analyze the traffic and operational data of stores on the platform in order to identify store optimization items that can solve or improve the pre-defined issues. Modeling is performed using the aforementioned store optimization items as targets, and a causal tree model is constructed based on multiple factors; A dataset is constructed based on the traffic and operational data of stores on the platform; Using the dataset, the influence of each of the multiple factors on the prediction results of the causal tree model is calculated, and the influence degree of each of the multiple factors is obtained. Based on the influence of each of the multiple factors, the multiple factors are sorted to obtain the factor ranking of the store optimization items; Wherein, at least one factor of the store optimization item is: the second-to-last factor in the factor ranking of the store optimization item.

6. The method according to claim 5, wherein, Also includes: Group the stores on the platform; The dataset was constructed based on traffic and operational data of stores within a group. as well as The factor ranking corresponding to the store optimization item obtained using the dataset is the factor ranking of the store optimization item corresponding to this group. Specifically, when obtaining the store optimization items associated with the matched query template and at least one factor of the store optimization items, the group to which the target store belongs is determined; the store optimization items associated with the matched query template and at least one factor of the store optimization items corresponding to the group to which the target store belongs are obtained.

7. The method according to any one of claims 1 to 6, wherein, Also includes: Build a query template library, which includes multiple query templates; Based on the Query template library, the copywriting database is constructed; The training sample set is generated based on the Query template library and the copywriting database; The pre-trained model is adjusted and trained using the training sample set.

8. The method according to claim 7, wherein, The Query template library includes at least one of the following: The query template for the reason why the data of any store on the platform differs from that of a reference store; Query templates for explaining why data discrepancies exist for any store on the platform at different times; Query templates for store management issues; Query templates corresponding to store optimization items; The query template is extended using the second preset model; The document database stores: The attribute information of stores on the platform; Traffic and operational data of stores on the platform; Store optimization items associated with the query template; Factor sorting of store optimization items; Suggested content regarding factor correlation; A corpus knowledge base of copywriting template information corresponding to different query templates.

9. The method according to claim 8, wherein, For the target store, identify the target issues, including: Responding to user input, determining the target question for the target store based on the user input information; and / or Responding to a user request triggered on the management page of a target store regarding a setting item, and for that setting item, determining the target issue; and / or The target store is diagnosed, and the target problem is determined based on the diagnosis results; and / or Obtain at least one recommendation question from the network side based on big data analysis, and in response to a selection event, take the selected recommendation question as the target question.

10. The method according to claim 9, wherein, The target store is diagnosed, and the target problem is determined based on the diagnosis results, including: Diagnose the gap between the target store and the reference store, and / or the data differences of the target store at different times, and / or the gap between the store data of the target store and the set indicators, and obtain the diagnostic results; The target problem is determined based on the diagnostic results.

11. An information processing method, wherein, include: Construct a query template library, which includes multiple query templates, and each query template is associated with a preset question related to store operations; Based on the Query template library, a copywriting database is constructed; wherein, the copywriting database stores at least one factor of the Query template obtained by attribution analysis of preset questions associated with the Query template using a causal inference algorithm; A training sample set is generated based on the Query template library and the copywriting database; Using the training sample set, the pre-trained model is adjusted and trained to obtain a first preset model that can handle intelligent response tasks in store operation scenarios.

12. The method according to claim 11, wherein, When constructing the document database, a causal inference algorithm is used to perform attribution analysis on preset questions associated with the query template to obtain at least one factor of the query template, including: To address the pre-defined issues associated with query templates, funnel analysis techniques are used to analyze the traffic and operational data of stores on the platform in order to identify store optimization items that can solve or improve the pre-defined issues. Modeling is performed using the aforementioned store optimization items as targets, and a causal tree model is constructed based on multiple factors; A dataset is constructed based on the traffic and operational data of stores on the platform; Using the dataset, the influence of each of the multiple factors on the prediction results of the causal tree model is calculated, and the influence degree of each of the multiple factors is obtained. Based on the influence of each of the multiple factors, the factors are ranked to obtain the factor ranking of the store optimization items.

13. The method according to claim 12, wherein, Also includes: Group the stores on the platform; The dataset was constructed based on traffic and operational data of stores within a group. as well as The factor ranking of the store optimization items obtained using the dataset is used to rank the store optimization items corresponding to this group.

14. The method according to any one of claims 11 to 13, wherein, Based on the aforementioned query template library, a copywriting resource library is constructed, including: Acquire and store the attribute information of stores on the platform; Acquire and store traffic and operational data of stores on the platform; Attribution analysis of preset questions associated with query templates in the query template library is performed using a causal inference algorithm to obtain at least one factor of the query template; Identify and store the suggested content for factor association; Construct a corpus knowledge base of copywriting template information corresponding to different query templates.

15. The method according to any one of claims 11 to 14, wherein, Based on the Query template library and the copywriting database, a training sample set is generated, including: The third preset model is used to output multiple training samples based on the Query template library and the copywriting database; The training sample set includes the plurality of training samples.

16. The method according to any one of claims 11 to 15, wherein, Build a query template library, including: Identify multiple basic query templates; Using the second preset model, the multiple basic query templates are expanded to obtain multiple extended query templates; The multiple extended query templates are evaluated, and the extended query templates that pass the evaluation are retained; The Query template library includes multiple basic Query templates and multiple extended Query templates that have passed evaluation.

17. A store diagnostic method, wherein, For the client, the method includes: Respond to diagnostic trigger events for the target store, trigger store diagnostics to present and / or play the diagnosed target issues; Display and / or play text, which includes suggestions to help users solve or improve the target problem; The text is obtained using a first preset model, which is obtained by adjusting and training the pre-trained model based on a pre-trained model for intelligent response tasks in a store operation scenario; the suggested content is associated with at least one factor of the target problem, and the at least one factor is obtained by attribution analysis of the target problem using a causal inference algorithm.

18. A method for operating a smart store assistant, wherein, For the client, the method includes: Display the target store's management page; Monitor user actions on the management page to identify user-defined issues related to store operations; In response to a user's confirmation command triggered for the target problem, display and / or play text, which includes suggestions to help the user solve or improve the target problem; The text is obtained using a first preset model, which is obtained by adjusting and training the pre-trained model based on a pre-trained model for intelligent response tasks in a store operation scenario; the suggested content is associated with at least one factor of the target problem, and the at least one factor is obtained by attribution analysis of the target problem using a causal inference algorithm.

19. A service system, wherein, include: A client for implementing the method described in claim 17 or 18 above; A server-side component for implementing the method described in any one of claims 1 to 16.

20. An electronic device, wherein, Includes memory and processor; among which, The memory is used to store executable instructions; The processor executes the executable instructions to implement the steps in the information processing method as described in any one of claims 1 to 16, the steps in the store diagnosis method as described in claim 17, or the steps in the store smart assistant working method as described in claim 18.

21. A computer-readable storage medium, wherein, The storage medium stores computer instructions that, when executed by a processor, can implement the steps in the information processing method as described in any one of claims 1 to 16, the steps in the store diagnosis method as described in claim 17, or the steps in the store intelligent assistant working method as described in claim 18.

22. A computer program product, wherein, The computer program product includes a computer program or instructions that, when executed by a processor, cause the processor to perform steps in the information processing method as described in any one of claims 1 to 16, or steps in the store diagnosis method as described in claim 17, or steps in the store smart assistant working method as described in claim 18.

Images