CN118227770B

CN118227770B - Task processing method, legal question answering method and task processing model training method

Info

Publication number: CN118227770B
Application number: CN202410660148.1A
Authority: CN
Inventors: 陈湘楠; 肖谦; 方元成; 陈晓婷; 李俊成; 林君; 刘晓钟; 汤斯亮
Original assignee: Alibaba China Co Ltd
Current assignee: Alibaba China Co Ltd
Priority date: 2024-05-24
Filing date: 2024-05-24
Publication date: 2024-08-06
Anticipated expiration: 2044-05-24
Also published as: CN118227770A

Abstract

The embodiment of the specification provides a task processing method, a legal question-answering method and a task processing model training method, wherein the task processing method comprises the following steps: task data of a target task are obtained; and inputting the task data into a task processing model to obtain a task processing result of the target task, wherein the task processing model is trained based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively, and the sample anti-facts data and the sample anti-facts results are obtained by processing the sample task data by using anti-facts guiding information. And the sample task data is processed by using the anti-facts guiding information to obtain sample anti-facts data and sample anti-facts results, so that efficient and automatic generation of model training data is realized, the task processing model has anti-facts data processing capability, and the task processing efficiency is further improved.

Description

Task processing method, legal question answering method and task processing model training method

Technical Field

The embodiment of the specification relates to the technical field of computers, in particular to a task processing method, a legal question-answering method and a task processing model training method.

Background

With the development of computer technology, a large model starts to enlarge the wonderful colors, has remarkable capability in terms of language understanding, generation, interaction and reasoning, and is widely applied to the fields of natural language processing such as dialogue, translation, code generation and the like. When a large model is used for task processing, the model is generally only required to extract information from task data and infer, however, the task processing mode is too single, and the problem that people often ask for counterfacts in a real scene is not considered, so that the development of the model is restricted.

At present, training is usually performed on models by manually generating the task data of the counterfactual in a crowdsourcing manner, but a great deal of manual labeling and expert knowledge are required for manually generating the task data of the counterfactual, so that the training data production and model training process are extremely inefficient, and therefore, an efficient task processing scheme is needed.

Disclosure of Invention

In view of this, the present embodiment provides a task processing method. One or more embodiments of the present specification relate to a legal question-answering method, a task processing model training method, a task processing model-based information processing method, a model training platform, a task processing device, a legal question-answering device, a task processing model training device, a task processing model-based information processing device, a computing device, a computer readable storage medium and a computer program product, so as to solve the technical defects existing in the prior art.

According to a first aspect of embodiments of the present specification, there is provided a task processing method, including: task data of a target task are obtained; and inputting the task data into a task processing model to obtain a task processing result of the target task, wherein the task processing model is trained based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively, and the sample anti-facts data and the sample anti-facts results are obtained by processing the sample task data by using anti-facts guiding information.

According to the task processing method provided by the embodiment of the specification, the task processing model is obtained by training based on the plurality of sample anti-facts data and the sample anti-facts results, so that the task processing model has anti-facts data processing capability, sample task data are processed by using anti-facts guiding information to obtain the sample anti-facts data and the sample anti-facts results, and model training data do not need to be manually generated, so that efficient and automatic generation of the model training data is realized, and the task processing efficiency is further improved.

Drawings

FIG. 1 is an architecture diagram of a task processing system provided in one embodiment of the present description;

FIG. 2 is an architecture diagram of another task processing system provided by one embodiment of the present description;

FIG. 3 is a flow chart of a method of task processing provided in one embodiment of the present disclosure;

FIG. 4 is a flow chart of a legal question-answering method provided by one embodiment of the present description;

FIG. 5 is a flow chart of a task processing model training method provided in one embodiment of the present disclosure;

FIG. 6 is a flow chart illustrating the generation of model training data in a task processing model training method according to one embodiment of the present disclosure;

FIG. 7 is a flow chart of a method of task processing model-based information processing provided in one embodiment of the present disclosure;

FIG. 8 is a schematic structural diagram of a model training platform according to one embodiment of the present disclosure;

FIG. 9 is a schematic diagram of a task processing device according to one embodiment of the present disclosure;

FIG. 10 is a schematic diagram of a legal question answering device according to one embodiment of the present disclosure;

FIG. 11 is a schematic diagram of a task processing model training device according to an embodiment of the present disclosure;

FIG. 12 is a schematic diagram of an information processing apparatus based on a task processing model according to an embodiment of the present disclosure;

FIG. 13 is a block diagram of a computing device provided in one embodiment of the present description.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present description. This description may be embodied in many other forms than described herein and similarly generalized by those skilled in the art to whom this disclosure pertains without departing from the spirit of the disclosure and, therefore, this disclosure is not limited by the specific implementations disclosed below.

The terminology used in the one or more embodiments of the specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of the specification. As used in this specification, one or more embodiments and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in one or more embodiments of the present specification refers to and encompasses any or all possible combinations of one or more of the associated listed items.

It should be understood that, although the terms first, second, etc. may be used in one or more embodiments of this specification to describe various information, these information should not be limited by these terms. These terms are only used to distinguish one type of information from another. For example, a first may also be referred to as a second, and similarly, a second may also be referred to as a first, without departing from the scope of one or more embodiments of the present description. The word "if" as used herein may be interpreted as "at … …" or "at … …" or "in response to a determination" depending on the context.

Furthermore, it should be noted that, user information (including, but not limited to, user equipment information, user personal information, etc.) and data (including, but not limited to, data for analysis, stored data, presented data, etc.) according to one or more embodiments of the present disclosure are information and data authorized by a user or sufficiently authorized by each party, and the collection, use, and processing of relevant data is required to comply with relevant laws and regulations and standards of relevant countries and regions, and is provided with corresponding operation entries for the user to select authorization or denial.

In one or more embodiments of the present description, a large model refers to a deep learning model with large scale model parameters, typically including hundreds of millions, billions, trillions, and even more than one billion model parameters. The large Model can be called as a Foundation Model, the large Model is pre-trained through a large-scale unlabeled corpus, a pre-trained Model with more than one hundred million parameters is produced, the Model can adapt to a wide downstream task, and the Model has better generalization capability, such as a large-scale language Model (LLM, large Language Model), a multi-modal pre-trained Model (multi-modal pre-training Model) and the like.

When the large model is actually applied, the pretrained model can be applied to different tasks by fine tuning with a small amount of samples, the large model can be widely applied to the fields of natural language processing (NLP, natural Language Processing), computer vision and the like, and particularly can be applied to the tasks of the computer vision fields such as vision question and answer (VQA, visual Question Answering), image description (IC), image generation and the like, and the tasks of the natural language processing fields such as emotion classification based on texts, text abstract generation, machine translation and the like, and main application scenes of the large model comprise digital assistants, intelligent robots, searching, online education, office software, electronic commerce, intelligent design and the like.

First, terms related to one or more embodiments of the present specification will be explained.

Reverse fact reasoning: the counterfacts inference is also called counterfacts thinking, and refers to the action of negating and re-characterizing facts that have occurred in the past to construct a likelihood hypothesis.

Visual question-answering: the machine is given a picture and an open natural language question, which requires the machine to output natural language answers. The answer may be in any of the following forms: phrase, word, yes or no, correct answer is selected from several possible answers. The visual question and answer is a typical multi-modal question, the technology of image understanding and natural language processing is combined, a computer needs to learn to understand images and words at the same time, and the visual question and answer is referred to as a chart question and answer in the embodiment of the specification.

Chart question-answering: data visualization approaches such as bar graphs and line graphs have become popular in analyzing data and making informed decisions. In analyzing data, one often presents complex reasoning problems for graphs involving arithmetic and logical operations.

When a large model is used for task processing, taking a chart question-answer task as an example, the current chart question-answer task only needs to draw charts from a chart dataset by the model and do reasoning and answer questions, however, the task processing mode is too single, and the problem that people often ask for counter facts in a real scene is not considered, so that the development of the model is restricted. Therefore, how to perform the inverse task using the model is becoming an important point of research.

Currently, models can be trained by constructing counterfactual sample data in the following two ways. Firstly, the model is trained by manually generating the counterfactual sample data in a crowdsourcing mode, and the method can ensure the quality of the counterfactual sample data, but requires a great deal of manual labeling and expert knowledge, so that the production of the counterfactual sample data and the training process of the model are extremely low in efficiency; second, the inverse fact sample data is automatically generated using templates. The solution often uses machine reference templates to generate counterfactual sample data by formulating problem templates. Although this solution is simple and efficient, low cost, the problems generated using machines are of too low quality and serious homogenization.

In order to solve the above problems, the embodiments of the present disclosure provide an automatic synthesis strategy for sample counterfactual data, and further provide a task processing method based on the automatic synthesis strategy for sample counterfactual data, where the method may process sample task data according to the utilization of counterfactual guiding information to obtain sample counterfactual data and sample counterfactual results, thereby implementing automatic synthesis of sample counterfactual data, and since the counterfactual guiding information may be screened from a plurality of candidate guiding information based on sample task data, the plurality of candidate guiding information may be obtained based on task attribute information and reference guiding information, and the reference guiding information may be determined based on task types of sample tasks, and therefore, the sample counterfactual data is various and high-quality, and efficient and high-quality automatic generation of model training data is implemented.

In the present specification, a task processing method, a legal question answering method, a task processing model training method, a task processing model-based information processing method, a model training platform, a task processing device, a legal question answering device, a task processing model training device, a task processing model-based information processing device, a computing device, a computer readable storage medium and a computer program product are provided, and the following embodiments are described in detail one by one.

Referring to fig. 1, fig. 1 illustrates an architecture diagram of a task processing system provided in one embodiment of the present disclosure, where the task processing system may include a client 100 and a server 200;

A client 100, configured to send task data of a target task to a server 200;

The server 200 is configured to input task data into a task processing model, and obtain a task processing result of a target task, where the task processing model is obtained by training sample anti-facts results corresponding to a plurality of sample anti-facts data and the plurality of sample anti-facts data, and the sample anti-facts data and the sample anti-facts results are obtained by processing the sample task data using anti-facts guiding information; sending a task processing result to the client 100;

The client 100 is further configured to receive a task processing result sent by the server 200.

By applying the scheme of the embodiment of the specification, the sample task data is processed by using the anti-facts guiding information to obtain the sample anti-facts data and the sample anti-facts results, so that the efficient and automatic generation of the model training data is realized, and the task processing efficiency is further improved.

Referring to fig. 2, fig. 2 illustrates an architecture diagram of another task processing system provided in one embodiment of the present disclosure, which may include a plurality of clients 100 and a server 200. Communication connection can be established between the plurality of clients 100 through the server 200, in a task processing scenario, the server 200 is used to provide task processing services between the plurality of clients 100, and the plurality of clients 100 can respectively serve as a transmitting end or a receiving end, so that communication is realized through the server 200.

The user may interact with the server 200 through the client 100 to receive data transmitted from other clients 100, or transmit data to other clients 100, etc. In the task processing scenario, it may be that the user issues a data stream to the server 200 through the client 100, and the server 200 generates a task processing result according to the data stream and pushes the task processing result to other clients that establish communications.

Wherein, the client 100 and the server 200 establish a connection through a network. The network provides a medium for a communication link between client 100 and server 200. The network may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others. The data transmitted by the client 100 may need to be encoded, transcoded, compressed, etc. before being distributed to the server 200.

The client 100 may be a browser, APP (Application), or a web Application such as H5 (HyperText Markup Language, hypertext markup language (htv) 5 th edition) Application, or a light Application (also called applet, a lightweight Application) or cloud Application, etc., and the client 100 may be based on a software development kit (SDK, software Development Kit) of a corresponding service provided by the server 200, such as a real-time communication (RTC, real Time Communication) based SDK development acquisition, etc. The client 100 may be deployed in an electronic device, need to run depending on the device or some APP in the device, etc. The electronic device may for example have a display screen and support information browsing etc. as may be a personal mobile terminal such as a mobile phone, tablet computer, personal computer etc. Various other types of applications are also commonly deployed in electronic devices, such as human-machine conversation type applications, model training type applications, text processing type applications, web browser applications, shopping type applications, search type applications, instant messaging tools, mailbox clients, social platform software, and the like.

The server 200 may include a server that provides various services, such as a server that provides communication services for multiple clients, a server for background training that provides support for a model used on a client, a server that processes data sent by a client, and so on. It should be noted that, the server 200 may be implemented as a distributed server cluster formed by a plurality of servers, or may be implemented as a single server. The server may also be a server of a distributed system or a server that incorporates a blockchain. The server may also be a cloud server for cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN, content Delivery Network), basic cloud computing services such as big data and artificial intelligence platforms, or an intelligent cloud computing server or an intelligent cloud host with artificial intelligence technology.

It should be noted that, the task processing method provided in the embodiments of the present disclosure is generally executed by the server, but in other embodiments of the present disclosure, the client may also have a similar function to the server, so as to execute the task processing method provided in the embodiments of the present disclosure. In other embodiments, the task processing method provided in the embodiments of the present disclosure may be performed by the client and the server together.

Referring to fig. 3, fig. 3 shows a flowchart of a task processing method according to an embodiment of the present disclosure, which specifically includes the following steps:

step 302: task data of a target task is obtained.

In one or more embodiments of the present disclosure, task data of a target task may be obtained, so that the task data is processed by using the task processing method provided in the embodiments of the present disclosure, and a task processing result of the target task is obtained.

Specifically, the target task may be a task in a different scene, such as a task in a legal scene, a task in an electronic market scene. The target tasks may be different types of tasks, such as intelligent question-answering tasks, text classification tasks. The target tasks may be different tasks of the same type, taking as an example the target task as an intelligent question-answering task, the target task may be a visual question-answering task, a chart question-answering task, a counter fact reasoning task, etc. The task data of the target task is a processing object of the task processing process. The task data may be data of different models, such as text data, image data, etc. The task data may be data in different languages, such as english task data, chinese task data, and so on.

In practical applications, there are various ways to obtain task data of the target task, and the embodiment of the present disclosure is not limited in this regard. In one possible implementation manner of the present disclosure, task data of a target task sent by a user may be received. In another possible implementation manner of the present specification, task data of the target task may be read from other data acquisition devices or databases.

Step 304: and inputting the task data into a task processing model to obtain a task processing result of the target task, wherein the task processing model is trained based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively, and the sample anti-facts data and the sample anti-facts results are obtained by processing the sample task data by using anti-facts guiding information.

In one or more embodiments of the present disclosure, after task data of a target task is acquired, further, the task data may be input into a task processing model to obtain a task processing result of the target task.

Specifically, the task processing model is used for processing task data to obtain task processing results. If the task data is anti-facts task data, the task processing model may be regarded as an anti-facts processing model. The task processing model may be obtained by training the pre-training large model based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data, or may be obtained by training the machine learning model based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data. Task processing models include, but are not limited to, a bi-directional coded representation deep self-attention model (BERT, bidirectional Encoder Representations from Transformers) model, a Text-to-Text deep self-attention model (T5, text-to-Text Transfer Transformer) model. And the task processing result corresponds to the target task, and if the target task is an intelligent question-answering task, the task processing result is a reply result of the question to be answered.

The task processing model includes an encoding unit and a decoding unit, the task data is input into the task processing model, firstly, the encoding unit captures complex dependency relationships among elements in the task data through a self-attention mechanism, and converts the task data into high-level semantic embedded representations layer by layer to obtain encoding vectors. Then, the decoding unit decodes the coding vector provided by the coding unit step by step through a self-attention mechanism and a masking mechanism to obtain a task processing result.

By applying the scheme of the embodiment of the specification, the task processing model is trained based on a plurality of sample anti-fact data and sample anti-fact results, so that the task processing model has anti-fact data processing capability, sample task data are processed by using anti-fact guiding information to obtain the sample anti-fact data and the sample anti-fact results, efficient automatic generation of model training data is realized, and task processing efficiency is further improved.

In practical application, when task data is input into the task processing model, in an alternative embodiment of the present disclosure, the task data may be directly input into the task processing model, so as to obtain a task processing result of the target task. In another optional embodiment of the present disclosure, in order to ensure that the task processing model can accurately process the task data, prompt information (Prompt) of the task processing model may also be obtained, and a processing procedure of the task processing model is normalized by using the Prompt information, so as to implement processing of the task data.

In an optional embodiment of the present disclosure, after the task data is input into the task processing model and the task processing result of the target task is obtained, the method may further include the following steps:

The task processing result is sent to the client;

Receiving result feedback information sent by a client, wherein the result feedback information is information for feeding back a task processing result based on task information of a target task;

according to the result feedback information, model optimization data are constructed;

And carrying out parameter adjustment on the task processing model by using the model optimization data.

Specifically, the result feedback information may be information that feeds back the content, quality and completion degree of the task processing result, and reflects the actual feeling and expectation of the client on the task processing result, where the result feedback information includes, but is not limited to, result quality evaluation information, corrected accurate task processing result, and optimization field of the model. Model optimization data refers to accurate optimization sample data for optimizing a task processing model.

In practical applications, there are various ways to construct model optimization data according to the result feedback information, and the embodiment of the present disclosure is not limited in any way. In one possible implementation manner of the present disclosure, the model optimization data may be automatically constructed directly according to the result feedback information. In another possible implementation manner of the present disclosure, the optimization prompt information may be generated based on the result feedback information, and model optimization data sent by the client terminal based on the optimization prompt information is received.

It should be noted that, taking directly according to the result feedback information, model optimization data is constructed as an example, if the result feedback information is the corrected accurate task processing result, the model optimization data can be constructed according to the task data of the target task and the corrected accurate task processing result. If the result feedback information is the optimization field of the model, such as the XXX field, sample inverse facts data of the XXX field can be obtained, and the sample inverse facts data of the XXX field is determined to be model optimization data. The process of parameter adjustment for the task processing model by using the model optimization data is the same as the training process of the task processing model, and the embodiments of the present disclosure will not be described again.

By applying the scheme of the embodiment of the specification, the performance of the task processing model is continuously optimized by collecting and utilizing the result feedback information, so that the actual requirements of a client are more accurately met, and the quality and accuracy of the final task processing result are improved.

In an alternative embodiment of the present disclosure, the building model optimization data according to the result feedback information may include the following steps:

Generating optimization prompt information according to the result feedback information, wherein the optimization prompt information is used for guiding a client to send model optimization data for optimizing a task processing model;

and sending the optimization prompt information to the client and receiving model optimization data sent by the client based on the optimization prompt information.

It should be noted that, according to the result feedback information, there are various ways to generate the optimization prompt information, and the embodiment of the present disclosure is not limited in this regard. In one possible implementation manner of the present disclosure, the preset prompt information may be directly obtained, and the result feedback information is added to the preset prompt information to obtain the optimized prompt information, for example, the preset prompt information is "very sorry brings inaccurate information to you". Please note where in particular is not accurate enough or provide the correct answer to the question concerned, i will correct and optimize my answer as soon as possible in order to better serve you. The result feedback information is ' inaccurate result ', and the optimized prompt information is ' aiming at the problem of inaccurate result fed back by you, so that inaccurate information is brought to you by very sorry. Please note where in particular is not accurate enough or provide the correct answer to the question concerned, i will correct and optimize my answer as soon as possible in order to better serve you. In another possible implementation manner of the present disclosure, the type of the result feedback information may be identified, the information type of the result feedback information may be determined, the information type may be further matched with the prompt type of each prompt information in the prompt information base, and the prompt information with the same prompt type as the information type may be determined as the optimized prompt information.

By applying the scheme of the embodiment of the specification, the model optimization data is acquired in an interactive guiding mode, so that the interactivity with the user is improved, and the user satisfaction is further improved.

marking key information in the task processing result to obtain an updated task processing result;

and sending the updated task processing result to the client.

Specifically, the key information refers to content in which the task processing result in the text is decisive for helping the user understand the result gist, capture the core viewpoint, recognize important facts or details. This information is typically a core component of the task processing results. The key information in the task processing result includes, but is not limited to, keywords, topics, entities.

In practical application, the key information in the task processing result is marked, and before the updated task processing result is obtained, the key information in the task processing result can be identified and determined. The method for identifying the key information of the task processing result is various, and the embodiment of the present disclosure is not limited in any way. In one possible implementation manner of the present disclosure, a task processing result may be matched with preset key information, and information that appears in the task processing result and is the same as the preset key information is determined as the key information. In another possible implementation manner of the present disclosure, the task processing result may be input into a key information recognition model, to obtain the key information, where the key information recognition model is obtained by training based on a plurality of training data and key information labels of each training data.

When the key information in the task processing result is marked, operations such as thickening, highlighting, tilting, adding frames and the like can be performed on the key information, and the marking mode of the key information is not limited in the embodiment of the specification.

By applying the scheme of the embodiment of the specification, the updated task processing result is sent to the client, so that the client can conveniently check the result, and the user experience is improved.

The task processing result is sent to the client;

receiving modification information sent by a client, wherein the modification information is used for modifying a task processing result;

And modifying the task processing result according to the modification information to obtain a modified task processing result.

Specifically, the modification information is used for describing the modification requirement of the client on the task processing result. Modification requirements include, but are not limited to, adding annotation requirements, deleting content requirements, replacing content requirements. Further, the modification requirement may also be a modification requirement for key information in the task processing result.

The task processing result is modified in various ways according to the modification information, and the embodiment of the present disclosure is not limited in any way. In one possible implementation manner of the present disclosure, a modification template (e.g., a content replacement template, a content deletion template) corresponding to the modification information may be obtained, and the task processing result is modified based on the modification template. In another possible implementation manner of the present disclosure, a deep learning model (e.g., a translation model, a classification model) corresponding to the modification information may be determined, and the task processing result is modified by using the deep learning model corresponding to the modification information.

By applying the scheme of the embodiment of the specification, the task processing result is modified according to the modification information sent by the client, so that the man-machine interaction is increased, and the adaptability and the flexibility of task processing are improved.

In an optional embodiment of the present disclosure, before the task data is input into the task processing model and the task processing result of the target task is obtained, the method may further include the following steps:

Acquiring a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data respectively;

inputting the multiple sample anti-facts data into a task processing model to obtain sample prediction results corresponding to the multiple sample anti-facts data respectively;

And adjusting model parameters of the task processing model according to the sample prediction result and the sample inverse result to obtain a task processing model with completed training.

In particular, the counterfactual data (Counterfactual Data) refers to one type of hypothetical data generated based on imaginary modifications to events that have occurred in the real world, for exploring results that might occur under different conditions or choices. The counterfactual data is not directly observed or actually recorded data, but rather is a specific hypothetical problem that meets the background of a particular factual problem, constructed by reasoning, modeling, or statistical adjustments to existing data, with the objective of discussing "how would … … be? "such a counterfactual problem, i.e., what affects the result if a certain key variable (e.g., decision, event, etc.) takes a value different from the actual case, while keeping other conditions unchanged. Sample anti-facts data refers to anti-facts data used to train a task processing model, and may also be referred to as sample anti-facts questions. The sample anti-facts result refers to the result corresponding to the sample anti-facts problem. For example, the sample counterfactual data is "what changes occur to the employment rate of teenagers if a policy is not implemented", and the sample counterfactual result is "employment rate will increase". The sample prediction result is a result obtained by predicting the sample anti-facts data by the task processing model. The sample anti-facts result is a prediction target of the task processing model and is used for evaluating the accuracy of the task processing model.

It should be noted that, there are various ways to obtain the plurality of sample anti-facts data and the sample anti-facts results corresponding to the plurality of sample anti-facts data, and the embodiment of the present disclosure is not limited in any way. In one possible implementation manner of the present disclosure, a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data may be read from other data acquisition devices or databases. In another possible implementation manner of the present disclosure, a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data sent by a user may be received.

In practical application, when training the task processing model according to the sample prediction result and the sample counter fact result, a loss value can be calculated according to the sample prediction result and the sample counter fact result, and model parameters of the task processing model are adjusted according to the loss value until the training process meets the preset stop condition, so that the task processing model after training is completed is obtained. Among these, there are many functions for calculating the loss value, such as a cross entropy loss function, an L1 norm loss function, a maximum loss function, a mean square error loss function, a logarithmic loss function, and the like. The preset stopping condition comprises, but is not limited to, that the loss value is smaller than or equal to a preset threshold value, and the iteration number reaches the preset iteration number, wherein the preset threshold value and the preset iteration number are specifically set according to actual conditions.

In one possible implementation of the present disclosure, after calculating the loss value, the loss value is compared with a preset threshold. Specifically, if the loss value is greater than a preset threshold, it indicates that the difference between the sample prediction result and the sample anti-fact result is greater, the task processing model has poor prediction capability on the sample anti-fact data, at this time, model parameters of the task processing model can be adjusted, and the steps of inputting a plurality of sample anti-fact data into the task processing model to obtain sample prediction results corresponding to the plurality of sample anti-fact data respectively are returned to be executed, and training is continued on the task processing model until the loss value is less than or equal to the preset threshold, which indicates that the difference between the sample prediction result and the sample anti-fact result is smaller, and a task processing model for completing training is obtained.

In another possible implementation manner of the present disclosure, in addition to comparing the magnitude relation between the loss value and the preset threshold, it may also be determined whether the current task processing model is trained and completed in combination with the iteration number. Specifically, if the loss value is greater than a preset threshold, the model parameters of the task processing model are adjusted, the step of inputting a plurality of sample anti-fact data into the task processing model and obtaining sample prediction results corresponding to the sample anti-fact data respectively is performed, training of the task processing model is continued until the preset iteration times are reached, iteration is stopped, and the task processing model with training completed is obtained.

By applying the scheme of the embodiment of the specification, training the task processing model according to the sample prediction result and the sample inverse facts result, and continuing training the task processing model under the condition that the preset stop condition is not met until the preset stop condition is reached, and completing training to obtain the task processing model. The model parameters of the task processing model are continuously adjusted, so that the finally obtained task processing model is more accurate.

In an optional embodiment of the present disclosure, in order to ensure effective training of the task processing model, the sample anti-facts results corresponding to the obtained plurality of sample anti-facts data and the plurality of sample anti-facts data may be verified, and the task processing model is trained by using the sample anti-facts results corresponding to the verified plurality of sample anti-facts data and the plurality of sample anti-facts data, that is, before the plurality of sample anti-facts data are input into the task processing model, and the sample prediction results corresponding to the plurality of sample anti-facts data are obtained, the method may further include the following steps:

Verifying the sample anti-facts data and the sample anti-facts results corresponding to the sample anti-facts data respectively to obtain verification results corresponding to the sample anti-facts data respectively;

Screening the multiple sample anti-facts data according to the verification result to obtain screened multiple sample anti-facts data;

inputting the plurality of sample anti-fact data into a task processing model to obtain sample prediction results respectively corresponding to the plurality of sample anti-fact data, wherein the method comprises the following steps of:

And inputting the screened multiple sample anti-facts data into a task processing model to obtain sample prediction results respectively corresponding to the screened multiple sample anti-facts data.

Specifically, the verification result may be a result of verification of whether the verification is passed or not, or may be a quality score obtained by verification, such as 69 points.

In practical applications, there are various ways to verify the sample anti-facts data and the sample anti-facts results corresponding to the sample anti-facts data, and the embodiment of the present disclosure is not limited in any way. In one possible implementation manner of the present disclosure, a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data may be sent to a verifier, and the verifier may manually verify the sample anti-facts data and the sample anti-facts results and receive verification results sent by the verifier. In another possible implementation manner of the present disclosure, a preset verification rule may be obtained, and a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data respectively are verified based on the verification rule, so as to obtain a verification result. For example, if the validation rule is that the percentage cannot exceed 100%, if the sample anti-fact data is 120%, it is indicated that the sample anti-fact data is not validated.

It should be noted that, there are various ways to screen the inverse facts data of the plurality of samples according to the verification result, and the embodiment of the present disclosure is not limited in this way. In one possible implementation manner of the present disclosure, sample anti-facts data that are not verified and corresponding sample anti-facts results thereof may be deleted, so as to obtain a plurality of sample anti-facts data after screening. In another possible implementation manner of the present disclosure, a verification screening threshold may be obtained, and sample counterfactual data with a quality score lower than the verification screening threshold and corresponding sample counterfactual results thereof are deleted, so as to obtain a plurality of sample counterfactual data after screening.

By applying the scheme of the embodiment of the specification, the accuracy of the task processing model obtained by training based on the screened multiple sample anti-facts data is higher because the screened multiple sample anti-facts data obtained by verification screening is the high-quality sample anti-facts data.

In an optional embodiment of the present disclosure, the acquiring a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data respectively may include the following steps:

acquiring a plurality of sample data pairs, wherein the sample data pairs comprise sample task data and counterfactual guide information corresponding to the sample task data;

and generating a model by inputting a plurality of sample data pairs into a task, and obtaining a plurality of sample anti-fact data and sample anti-fact results corresponding to the plurality of sample anti-fact data respectively, wherein the sample anti-fact data and the sample data pairs are in one-to-one correspondence.

Specifically, the task generating model is used for carrying out anti-fact reasoning on the sample task data by utilizing the anti-fact guiding information to obtain sample anti-fact data and a corresponding sample anti-fact result. The task generation model may be a pre-trained large model; the machine learning model can also be obtained based on training sample data pairs and result labels corresponding to the training sample data pairs, wherein the training sample data pairs comprise training task data and training counterfactual guiding information corresponding to the training task data, and the result labels comprise training counterfactual data of the training sample data pairs and training counterfactual results corresponding to the training counterfactual data; the machine learning model can be obtained by training based on training task attribute information, training reference guide information and training guide information labels. The anti-facts guide information is screened from the plurality of candidate guide information based on the sample task data, so that the anti-facts guide information is candidate guide information with high matching degree with the sample task data in the plurality of candidate guide information. Sample anti-fact data corresponding to the sample task data can be generated in high quality through the anti-fact guidance information. Sample task data in the sample data pair is data conforming to actual conditions, and sample anti-fact data can be obtained by carrying out anti-fact reasoning on the sample task data through anti-fact guiding information.

Illustratively, assume that a pie chart of sample data versus "what is the employment rate" including sample task data includes three sectors, the first sector "student" accounts for 20%, the second sector "graduate" corresponds to a value of 30%, and the third sector "employment person" corresponds to a value of 50%. The sample data pair also comprises the counterfactual guiding information corresponding to the sample task data, namely "select one sector, supposing that M is added, the sample data pair is input into the task generation model, the sample counterfactual data is obtained, namely" supposing that M is added to the second sector, what change happens to the employment rate ", and the sample counterfactual result is" employment rate increase ".

In practical applications, there are various ways to obtain multiple pairs of sample data. In one possible implementation of the present description, a plurality of sample data pairs may be read from other data acquisition devices or databases. In another possible implementation of the present disclosure, a plurality of sample data pairs sent by a user may be received.

By applying the scheme of the embodiment of the specification, high-quality sample counterfactual data and sample counterfactual results are automatically generated by utilizing the language understanding capability of the task generation model. Moreover, as the sample anti-facts data is obtained by carrying out anti-facts reasoning based on the sample task data, the problem that the anti-facts data is difficult to obtain is solved, and the feasibility of the scheme is improved.

In practical application, in a graph question-answering scene, because the current data set contains a single graph type, only a certain graph type, such as a histogram and a line graph, is often contained, and this cannot reflect the rich graph types in the real scene and cannot fully train the graph understanding capability of the model, in the embodiment of the specification, various graph sample data pairs can be obtained to train the task processing model, so that the task processing model has higher graph understanding capability.

In an alternative embodiment of the present disclosure, taking a sample data pair of a chart question-answer scene as an example, the sample task data includes a sample chart question and a sample chart annotation, and the counterfactual guide information includes chart counterfactual guide information; the generating a model for the input task by using the plurality of sample data to obtain a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data respectively may include the following steps:

and inputting the sample chart questions, the sample chart notes and the chart anti-facts guiding information into a task generating model to obtain sample chart anti-facts data and sample chart anti-facts results corresponding to the sample chart anti-facts data.

In particular, the sample chart questions, sample chart annotations, and chart counterfactual guidance information may be information of different charts. The different charts include, but are not limited to, bar charts, line charts, pie charts, histograms, scatter charts, thermodynamic diagrams, geographic maps, and the embodiments of the present specification are not limited in any way. Sample chart questions are sample questions of the chart, such as "how much people are not on the internet," sample chart notes refer to written descriptions on or closely associated with the chart, and sample chart notes are used to provide necessary explanation, supplementary information or context for the chart, helping the reader to better understand the content, meaning of data, analysis method or conclusion of the chart. Sample chart annotations include, but are not limited to, legends, titles. The chart counterfactual guide information is the counterfactual guide information of the pointer to the chart, such as "select one sector, assume it is merged with another sector".

It should be noted that, in order to save the cost of calling the task generation model application programming interface (API, application Programming Interface), specific contents of the chart, such as sample chart annotation, may be represented in the manner of JSON file.

By applying the scheme of the embodiment of the specification, various sample chart problems, sample chart notes and chart counter fact guiding information are processed by utilizing the language understanding capability of the task generation model, so that sample chart counter fact data and sample chart counter fact results are obtained, and the diversity and accuracy of model training data of the task processing model are improved.

In an alternative embodiment of the present disclosure, the acquiring a plurality of pairs of sample data may include the following steps:

acquiring sample task data and task attribute information of a plurality of sample tasks;

Inputting task attribute information of a sample task and reference guide information corresponding to the sample task into a task generation model to obtain a plurality of candidate guide information corresponding to sample task data, wherein the reference guide information corresponding to the sample task is determined based on the task type of the sample task;

screening out the counterfactual guide information of the sample task from a plurality of candidate guide information corresponding to the sample task data according to the sample task data of the sample task;

and constructing a plurality of sample data pairs according to the sample task data and the counterfactual guiding information of the plurality of sample tasks.

Specifically, the description of the sample task may refer to the description of the target task, and the embodiments of the present disclosure will not be repeated. Sample task data refers to data conforming to actual conditions. The task attribute information refers to detailed data describing various characteristics and settings of the sample task. For example, the sample task is a chart task, and the task attribute information is chart attribute information. The chart attribute information is used to describe the visual presentation of the chart, the manner in which the data is processed, the interactive functionality, and the details of the interaction with the user. Chart attribute information includes, but is not limited to, chart type, chart data source, chart color. The candidate guidance information refers to guidance information having a relationship with task attribute information of the sample task. The candidate guiding information is obtained based on task attribute information and reference guiding information, and the reference guiding information can be guiding information set by a user according to priori knowledge in advance or guiding information determined based on task types of sample tasks. Task types of sample tasks include, but are not limited to, chart types, text types.

In practical applications, there are various ways to obtain sample task data and task attribute information of a plurality of sample tasks, which are not limited in any way in the embodiments of the present disclosure. In one possible implementation manner of the present disclosure, sample task data and task attribute information of a plurality of sample tasks may be read from other data acquisition devices or databases. In another possible implementation manner of the present disclosure, sample task data and task attribute information of a plurality of sample tasks sent by a user may be received.

The method for screening the first anti-facts guide information of the first sample task from the plurality of first candidate guide information according to the first sample task data of the first sample task is not limited in any way. The first sample task is any one of a plurality of sample tasks, and the first sample task data, the first candidate guide information and the first anti-facts guide information are sample task data, candidate guide information and anti-facts guide information of the first sample task respectively. In one possible implementation manner of the present disclosure, the similarity between the first sample task data and the plurality of first candidate guiding information may be calculated, and the first candidate guiding information with the higher similarity is determined as the first counterfactual guiding information. In another possible implementation manner of the present disclosure, the first sample task data and the plurality of first candidate guide information may be input into an information filtering model, and the first inverse facts guide information of the first sample task may be filtered out by using a language understanding capability of the information filtering model, where the information filtering model may be a pre-training large model, or may be a machine learning model obtained by training based on training task data, the plurality of training guide information to be filtered, and the inverse facts guide information label.

By applying the scheme of the embodiment of the specification, sample task data and task attribute information of a plurality of sample tasks are obtained; inputting first task attribute information and first reference guide information of a first sample task into a task generation model to obtain a plurality of first candidate guide information corresponding to first sample task data; screening first inverse fact guide information of a first sample task from a plurality of first candidate guide information according to first sample task data of the first sample task; and constructing a plurality of sample data pairs according to the sample task data and the counterfactual guiding information of the plurality of sample tasks. By utilizing the language understanding capability of the task generating model, various and high-quality sample data pairs are generated based on the task attribute information and the reference guide information, and the accuracy of the task processing model is improved.

It should be noted that, in the embodiment of the present specification, the reference guide information may be extended by bootstrapping (bootstrapping) to help the task generating model generate more various candidate guide information in the next round. Specifically, after the candidate guide information is obtained, the candidate guide information may be determined as reference guide information to participate in the generation of the candidate guide information of the next round, the counterfactual guide information may be determined as reference guide information to participate in the generation of the candidate guide information of the next round, and the counterfactual guide information corresponding to the verified sample counterfactual data may be determined as reference guide information to participate in the generation of the candidate guide information of the next round, which is specifically selected according to the actual situation, and the embodiment of the present specification does not limit any limitation. Preferably, in order to ensure accuracy of the reference guidance information, the anti-fact guidance information corresponding to the sample anti-fact data that passes the verification is determined as the reference guidance information.

The task processing method provided in the present specification will be further described with reference to fig. 4 by taking an application of the task processing method in a legal question-answering scenario as an example. Fig. 4 shows a flowchart of a legal question-answering method according to an embodiment of the present disclosure, which specifically includes the following steps:

Step 402: and obtaining the questions to be answered of the target legal task.

Step 404: and inputting the questions to be answered into a back facts question and answer model to obtain answer results of the questions to be answered, wherein the back facts question and answer model is obtained based on training of a plurality of sample back facts data and sample back facts results corresponding to the sample back facts data respectively, and the sample back facts data and the sample back facts results are obtained by processing sample task data by using back facts guiding information.

It should be noted that, the implementation manner of step 402 to step 404 is the same as the implementation manner of step 302 to step 304, and the description of the embodiment of the present disclosure is omitted. The questions to be answered refer to legal questions to be answered, such as interpretation questions of legal principles, legal qualitative questions of new cases, etc. The relevant description of the inverse question-answer model can refer to the relevant description of the task processing model.

By applying the scheme of the embodiment of the specification, the anti-fact question-answering model is trained based on a plurality of sample anti-fact data and sample anti-fact results, so that the anti-fact question-answering model has anti-fact data processing capability, sample task data are processed by using anti-fact guiding information to obtain sample anti-fact data and sample anti-fact results, efficient automatic generation of model training data is realized, and legal question-answering efficiency is further improved.

In an optional embodiment of the present disclosure, after the inputting the to-be-answered question into the anti-fact question answering model and obtaining the answer result of the to-be-answered question, the method may further include the following steps:

Sending the reply result to the client;

Receiving result feedback information sent by a client, wherein the result feedback information is information for feeding back a reply result based on task information of a target legal task;

And sending the result feedback information to a model training platform, wherein the model training platform is used for carrying out parameter adjustment on the anti-fact question-answer model by utilizing the result feedback information.

Specifically, the model training platform may be deployed at a terminal device or may be deployed at a cloud device, and specifically set according to an actual situation, which is not limited in any way in the embodiment of the present disclosure. The model training platform provides large-scale data processing and high-performance computing resources for training, optimizing and deploying various machine learning models, in particular deep learning models. On the model training platform, users can upload data, select or customize algorithm models, and efficiently perform model training and verification through distributed computing capabilities. The model training platform can receive a task generation request from the terminal equipment, acquire a corresponding task processing model according to the request information, and generate task information based on the task processing model. The model training platform can also rapidly respond to the demands of different tasks, and call a proper task processing model to perform model training, so that an accurate task processing model is finally generated.

In practical application, after receiving the result feedback information sent by the client, the result feedback information can be sent to a model training platform, and the model training platform can carry out parameter adjustment on the task processing model based on the result feedback information. The mode of the model training platform for carrying out parameter adjustment on the task processing model based on the result feedback information can refer to the mode of constructing model optimization data according to the result feedback information; the implementation manner of performing parameter adjustment on the task processing model by using the model optimization data will not be described in detail in the embodiments of the present specification.

By applying the scheme of the embodiment of the specification, the parameter adjustment is carried out on the task processing model based on the result feedback information through the model training platform, so that the model training cost can be reduced while the accuracy and the training efficiency of the task processing model are ensured, and a convenient and efficient model training service is provided for users.

Referring to fig. 5, fig. 5 shows a flowchart of a task processing model training method according to an embodiment of the present disclosure, which specifically includes the following steps:

step 502: and obtaining a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data respectively, wherein the sample anti-facts data and the sample anti-facts results are obtained by processing sample task data by using anti-facts guiding information.

Step 504: and inputting the plurality of sample anti-facts data into a task processing model to obtain sample prediction results respectively corresponding to the plurality of sample anti-facts data.

Step 506: and adjusting model parameters of the task processing model according to the sample prediction result and the sample inverse result to obtain a task processing model with completed training.

It should be noted that, the implementation manners of the steps 502 to 506 may refer to the training manner of the task processing model, and the embodiments of the present disclosure will not be repeated.

In practical application, after obtaining the task processing model after training, the model parameters of the task processing model after training can be sent to the terminal equipment, and the client can build the task processing model locally based on the model parameters, and the task processing model is utilized to process tasks.

Referring to fig. 6, fig. 6 shows a flowchart for generating model training data in a task processing model training method according to an embodiment of the present disclosure, which specifically includes three stages, namely, an instruction proposal synthesis stage, a model training data generation stage, and a verification stage, where an instruction proposal (candidate guidance information) describing a general counterfactual operation may be generated using a task generation model; in the model training data generation stage, candidate guide information generated in the reference instruction proposal synthesis stage can be utilized to generate model training data; in the verification stage, the model training data can be verified, and high-quality model training data can be obtained through screening. Next, the implementation procedure of the above three phases will be described in detail respectively:

Instruction proposal synthesis stage: firstly, initializing an instruction suggestion library, wherein the instruction suggestion library comprises at least one piece of reference guide information of different task types; then, sample task data (comprising sample chart problems and sample chart notes) and chart attribute information of a sample chart task are obtained, the chart attribute information and reference guide information of the sample chart task are input into a task generation model, and a plurality of candidate guide information are obtained, wherein the reference guide information of the sample chart task is extracted from an instruction suggestion library based on the task type of the sample chart task;

model training data generation stage: screening out the inverse fact guide information from the candidate guide information according to the sample chart problem and the sample chart annotation; inputting sample chart questions, sample chart notes and counterfactual guiding information into a task generating model to obtain sample counterfactual data and sample counterfactual results corresponding to the sample counterfactual data;

Verification: verifying the sample anti-fact data and sample anti-fact results corresponding to the sample anti-fact data to obtain verification results; and determining sample anti-fact data and corresponding sample anti-fact results which are determined to pass through verification based on the verification results as model training data, and determining anti-fact guide information corresponding to the sample anti-fact data which is determined to pass through verification as reference guide information in a bootstrapping mode to be put back into an instruction suggestion library, so that the model is helped to generate more various instruction suggestions in the next round.

In practical application, after model training data is obtained, the task processing model can be trained by using the model training data, so as to obtain a task processing model after training.

By applying the scheme of the embodiment of the specification, the sample anti-fact data is automatically generated by designing the framework for automatically synthesizing the anti-fact data, so that the diversity and quality of model training data are improved; candidate guide information is generated by using the task generation model through given chart attribute information, and the counterfactual guide information is put back into the instruction suggestion library in a bootstrapping mode, so that the model is helped to generate more various instruction suggestions in the next round.

Referring to fig. 7, fig. 7 shows a flowchart of an information processing method based on a task processing model according to an embodiment of the present disclosure, which specifically includes the following steps:

Step 702: a task generation request is received, wherein the task generation request includes request information.

Specifically, the information processing method based on the task processing model can be applied to terminal equipment and also can be applied to a model training platform. The task generation request is for requesting task information of a generation target task. The task generation request typically contains the task type, the desired output format, and the request information. For example, when the user selects the "intelligent question and answer" function on the front-end interface of the model training platform and uploads the task data, a task generation request may be constructed, where the task generation request includes information such as task data, task type (i.e., intelligent question and answer), language type of the task data, and the like. The request information refers to parameters or description information related to the target task carried in the task generation request. The request information includes, but is not limited to, a task scenario identification of the target task, a task model identification, or model training data of the target task (a plurality of sample anti-fact data and sample anti-fact results corresponding to the plurality of sample anti-fact data, respectively).

Step 704: and acquiring a task processing model based on the request information, wherein the task processing model is trained based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively, and the sample anti-facts data and the sample anti-facts results are obtained by processing the sample task data by using anti-facts guiding information.

In practical applications, the manner of acquiring the task processing model based on the request information is various, and is specifically selected according to practical situations, which is not limited in any way in the embodiment of the present specification. In an optional embodiment of the disclosure, the request information includes a task scenario identifier of the target task, or a task model identifier; the task processing model acquiring step may include the following steps:

Determining a target scene template from a plurality of preset scene templates based on the task scene identification, and searching a task processing model from a model library based on the target scene template, wherein the model library stores a plurality of processing models; or alternatively

Based on the task model identification, a task processing model is searched from a model library.

In particular, task scenario identification refers to a unique or specific tag used to distinguish between different task application scenarios. In the present embodiment, the task scenario identification is part of the request information. The task information can be generated by selecting a target scene template matched with the request information from a series of preset scene templates through task scene identification. For example, the task scenario identifier is "intelligent question-answer", which means that the client wants to answer the uploaded task data, and the intelligent question-answer scenario template can be selected from a plurality of preset scenario templates according to the task scenario identifier.

The preset scene templates are standard configuration scene templates predefined for different task application scenes, and each template comprises model information matched with the task application scene, task processing flow and other information. The task generation requests of different scenes can be responded quickly through a series of preset scene templates. Different preset scene templates correspond to different task types, model information and process flows. For example, there may be an intelligent question-answer scene template specific to intelligent question-answer, which includes model information and process flow of the trained intelligent question-answer model.

The target scene template refers to a scene template matched with the task scene identification. When the task generating request is analyzed, the task generating request can be positioned to a corresponding target scene template based on the task scene identification, and a corresponding task processing model and other relevant configuration information are selected from a model library according to model information included in the target scene template. For example, when the task scene is identified as "intelligent question-answering", the target scene template is a template containing model information and relevant configuration parameters of the intelligent question-answering model.

The model library is a resource library that centrally stores deep learning models that are trained and optimized to address different processing tasks. The model library stores process models including, but not limited to, task process models, intelligent question-answer models, anti-fact process models, anti-fact question-answer models. Moreover, the processing models in the model library can be divided into different versions according to different application scenes, for example, task processing models of multiple versions, such as task processing models for text data and task processing models for image data, can be stored in the model library.

Task model identification refers to a unique or specific tag used to distinguish between models to which different tasks apply. For example, the task model identification may be "intelligent question-answering", and based on the task model identification, an intelligent question-answering model applicable to the intelligent question-answering task may be searched from a model library.

By applying the scheme of the embodiment of the specification, the task processing model acquisition process is more flexible, efficient and standard through the predefined task scene template, the task model identification and the model library resource.

In another optional embodiment of the present disclosure, besides selecting a task processing model that is trained in advance from a model library, the task processing model may be obtained by targeted training according to model training data in the request information, that is, the request information includes a plurality of sample anti-facts data of the target task and sample anti-facts results corresponding to the plurality of sample anti-facts data respectively; the task processing model acquiring step may include the following steps:

Training a task processing model corresponding to the target task based on the sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively to obtain a task processing model after training.

It should be noted that, the implementation manner of "training the task processing model corresponding to the target task based on the plurality of sample anti-facts data and the sample anti-facts results corresponding to the plurality of sample anti-facts data to obtain the task processing model after training" is the same as the training manner of the task processing model, and the embodiments of the present disclosure will not be repeated.

By applying the scheme of the embodiment of the specification, the task processing model corresponding to the target task is trained based on the sample anti-fact data and the sample anti-fact results corresponding to the sample anti-fact data, so that the task processing model after training is obtained, the precision of the task processing model is ensured, and the task processing model meets the requirements of users.

Step 706: and generating task information based on the task processing model, wherein the task information is used for executing the target task.

Specifically, the task information includes model configuration and process flows required to execute the target task. The terminal device or other server component can correctly process the target task using the task processing model based on the task information.

The task information can be generated by directly packing model parameters of the task processing model based on the task processing model. Other model information of the task processing model, such as the processing mode of model input data, the specification of expected output results, and possibly related intermediate steps and other auxiliary information, can also be acquired, and task information is constructed based on the other model information.

For example, in an intelligent question-answering task, the task information may include parameters such as address information of the selected task processing model, storage locations for input task data, target paths for output of task processing results, and other environmental configurations required for the task processing model to run, which information enables the task processing model to be properly loaded and the intelligent question-answering task to be executed on a local or remote server.

By applying the scheme of the embodiment of the specification, the task information of the target task is generated, so that the system deployment and operation cost can be reduced while the processing quality and efficiency of the target task are ensured, and a convenient and efficient task processing service is provided for a user.

Referring to fig. 8, fig. 8 is a schematic structural diagram of a model training platform according to an embodiment of the present disclosure, where the model training platform includes a request interface unit 802 and a model unit 804;

a request interface unit 802, configured to receive a model training request, where the model training request includes request information;

a model unit 804, configured to obtain a task processing model; and training the task processing model based on the request information to obtain a trained task processing model, wherein the request information comprises a plurality of sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively, and the sample anti-facts data and the sample anti-facts results are obtained by processing the sample task data by using anti-facts guiding information.

The request information includes, but is not limited to, a task scenario identifier of the target task, a task model identifier, or model training data of the target task (a plurality of sample anti-fact data and sample anti-fact results corresponding to the plurality of sample anti-fact data, respectively). The processing manner of the model unit is the same as the training manner of the task processing model, and the description of the embodiment of the present disclosure will not be repeated.

In an alternative embodiment of the present disclosure, the model training platform further includes a model library;

and the model unit is also used for storing the trained task processing model into a model library.

It should be noted that, the model unit trains the task processing model based on the request information, and after obtaining the task processing model after training, the task processing model after training can also be stored in the model library, so that the task processing model after training can be directly obtained from the model library without repeated training. The method for obtaining the task processing model after training from the model library can refer to the task scene identification, determine a target scene template from a plurality of preset scene templates, and search the task processing model from the model library based on the target scene template, wherein the model library stores a plurality of processing models; or based on the task model identification, searching the implementation manner of the task processing model from the model library, and will not be described in detail in the embodiment of the present specification.

Corresponding to the task processing method embodiment, the present disclosure further provides a task processing device embodiment, and fig. 9 shows a schematic structural diagram of a task processing device provided in one embodiment of the present disclosure. As shown in fig. 9, the apparatus includes:

A first obtaining module 902 configured to obtain task data of a target task;

The first input module 904 is configured to input task data into a task processing model to obtain a task processing result of the target task, where the task processing model is trained based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data, and the sample anti-facts data and the sample anti-facts results are obtained by processing the sample task data using anti-facts guidance information.

Optionally, the apparatus further comprises: the model training module is configured to acquire a plurality of sample anti-facts data and sample anti-facts results respectively corresponding to the plurality of sample anti-facts data; inputting the multiple sample anti-facts data into a task processing model to obtain sample prediction results corresponding to the multiple sample anti-facts data respectively; and adjusting model parameters of the task processing model according to the sample prediction result and the sample inverse result to obtain a task processing model with completed training.

Optionally, the model training module is further configured to acquire a plurality of sample data pairs, wherein the sample data pairs include sample task data and counterfactual guide information corresponding to the sample task data; and generating a model by inputting a plurality of sample data pairs into a task, and obtaining a plurality of sample anti-fact data and sample anti-fact results corresponding to the plurality of sample anti-fact data respectively, wherein the sample anti-fact data and the sample data pairs are in one-to-one correspondence.

Optionally, the model training module is further configured to obtain sample task data and task attribute information of a plurality of sample tasks; inputting task attribute information of a sample task and reference guide information corresponding to the sample task into a task generation model to obtain a plurality of candidate guide information corresponding to sample task data, wherein the reference guide information corresponding to the sample task is determined based on the task type of the sample task; screening out the counterfactual guide information of the sample task from a plurality of candidate guide information corresponding to the sample task data according to the sample task data of the sample task; and constructing a plurality of sample data pairs according to the sample task data and the counterfactual guiding information of the plurality of sample tasks.

Optionally, the sample task data includes sample chart questions and sample chart annotations, and the counterfactual guide information includes chart counterfactual guide information; the model training module is further configured to input the sample chart problem, the sample chart annotation and the chart counterfactual guiding information into the task generating model to obtain sample chart counterfactual data and sample chart counterfactual results corresponding to the sample chart counterfactual data.

Optionally, the apparatus further comprises: the verification module is configured to verify the sample anti-fact data and the sample anti-fact results respectively corresponding to the sample anti-fact data to obtain verification results respectively corresponding to the sample anti-fact data; screening the multiple sample anti-facts data according to the verification result to obtain screened multiple sample anti-facts data; the model training module is further configured to input the screened multiple sample anti-facts data into the task processing model to obtain sample prediction results corresponding to the screened multiple sample anti-facts data respectively.

Optionally, the apparatus further comprises: the first sending module is configured to send the task processing result to the client; receiving result feedback information sent by a client, wherein the result feedback information is information for feeding back a task processing result based on task information of a target task; according to the result feedback information, model optimization data are constructed; and carrying out parameter adjustment on the task processing model by using the model optimization data.

Optionally, the first sending module is further configured to generate optimization prompt information according to the result feedback information, where the optimization prompt information is used to guide the client to send model optimization data for optimizing the task processing model; and sending the optimization prompt information to the client and receiving model optimization data sent by the client based on the optimization prompt information.

Optionally, the apparatus further comprises: the marking module is configured to mark key information in the task processing result and obtain an updated task processing result; and sending the updated task processing result to the client.

Optionally, the apparatus further comprises: the second sending module is configured to send the task processing result to the client; receiving modification information sent by a client, wherein the modification information is used for modifying a task processing result; and modifying the task processing result according to the modification information to obtain a modified task processing result.

The above is a schematic solution of a task processing device of the present embodiment. It should be noted that, the technical solution of the task processing device and the technical solution of the task processing method belong to the same concept, and details of the technical solution of the task processing device, which are not described in detail, can be referred to the description of the technical solution of the task processing method.

Corresponding to the above legal question-answering method embodiment, the present disclosure further provides a legal question-answering device embodiment, and fig. 10 shows a schematic structural diagram of a legal question-answering device provided in one embodiment of the present disclosure. As shown in fig. 10, the apparatus includes:

A second obtaining module 1002 configured to obtain a question to be answered of the target legal task;

And a second input module 1004 configured to input the question to be answered into a back facts question and answer model to obtain a answer result of the question to be answered, wherein the back facts question and answer model is obtained based on a plurality of sample back facts data and sample back facts results respectively corresponding to the sample back facts data, and the sample back facts data and the sample back facts results are obtained by processing sample task data by using back facts guiding information.

Optionally, the apparatus further comprises: a third sending module configured to send the reply result to the client; receiving result feedback information sent by a client, wherein the result feedback information is information for feeding back a reply result based on task information of a target legal task; and sending the result feedback information to a model training platform, wherein the model training platform is used for carrying out parameter adjustment on the anti-fact question-answer model by utilizing the result feedback information.

By applying the scheme of the embodiment of the specification, the sample task data is processed by using the anti-fact guide information to obtain the sample anti-fact data and the sample anti-fact result, so that the efficient automatic generation of the model training data is realized, and the legal question-answering efficiency is further improved.

The above is an exemplary scheme of a legal question-answering device of this embodiment. It should be noted that, the technical solution of the legal question answering device and the technical solution of the legal question answering method belong to the same concept, and details of the technical solution of the legal question answering device which are not described in detail can be referred to the description of the technical solution of the legal question answering method.

Corresponding to the task processing model training method embodiment, the present disclosure further provides a task processing model training device embodiment, and fig. 11 shows a schematic structural diagram of a task processing model training device provided in one embodiment of the present disclosure. As shown in fig. 11, the apparatus includes:

A third obtaining module 1102, configured to obtain a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data, where the sample anti-facts data and the sample anti-facts results are obtained by processing sample task data using anti-facts guidance information;

A third input module 1104 configured to input a plurality of sample inverse facts data into the task processing model, to obtain sample prediction results corresponding to the plurality of sample inverse facts data, respectively;

The adjustment module 1106 is configured to adjust model parameters of the task processing model according to the sample prediction result and the sample inverse result, so as to obtain a task processing model with completed training.

The above is a schematic scheme of a task processing model training device of the present embodiment. It should be noted that, the technical solution of the task processing model training device and the technical solution of the task processing model training method belong to the same concept, and details of the technical solution of the task processing model training device which are not described in detail can be referred to the description of the technical solution of the task processing model training method.

Corresponding to the above-mentioned information processing method embodiment based on the task processing model, the present disclosure further provides an embodiment of an information processing device based on the task processing model, and fig. 12 is a schematic structural diagram of an information processing device based on the task processing model according to one embodiment of the present disclosure. As shown in fig. 12, the apparatus includes:

a receiving module 1202 configured to receive a task generation request, wherein the task generation request includes request information;

A fourth obtaining module 1204, configured to obtain a task processing model based on the request information, where the task processing model is obtained by training sample counterfactual results corresponding to the plurality of sample counterfactual data and the plurality of sample counterfactual data respectively, and the sample counterfactual data and the sample counterfactual results are obtained by processing the sample task data using the counterfactual guiding information;

the generating module 1206 is configured to generate task information based on the task processing model, wherein the task information is used for executing the target task.

Optionally, the request information includes a task scene identifier of the target task, or a task model identifier; the fourth obtaining module 1204 is further configured to determine a target scene template from a plurality of preset scene templates based on the task scene identifier, and search a task processing model from a model library based on the target scene template, where the model library stores a plurality of processing models; or searching a task processing model from a model library based on the task model identification.

Optionally, the request information includes a plurality of sample anti-facts data of the target task and sample anti-facts results respectively corresponding to the plurality of sample anti-facts data; the fourth obtaining module 1204 is further configured to train the task processing model corresponding to the target task based on the plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data, respectively, to obtain a trained task processing model.

The above is a schematic scheme of an information processing apparatus based on a task processing model of the present embodiment. It should be noted that, the technical solution of the information processing apparatus based on the task processing model and the technical solution of the information processing method based on the task processing model belong to the same concept, and details of the technical solution of the information processing apparatus based on the task processing model, which are not described in detail, can be referred to the description of the technical solution of the information processing method based on the task processing model.

FIG. 13 illustrates a block diagram of a computing device provided in one embodiment of the present description. The components of computing device 1300 include, but are not limited to, a memory 1310 and a processor 1320. Processor 1320 is coupled to memory 1310 via bus 1330, and database 1350 is used to store data.

Computing device 1300 also includes an access device 1340, which access device 1340 enables computing device 1300 to communicate via one or more networks 1360. Examples of such networks include public switched telephone networks (PSTN, public Switched Telephone Network), local area networks (LAN, local Area Network), wide area networks (WAN, wide Area Network), personal area networks (PAN, personal Area Network), or combinations of communication networks such as the internet. The access device 1340 may include one or more of any type of Network interface, wired or wireless, such as one or more of a Network interface card (NIC, network INTERFACE CARD), such as an IEEE802.11 wireless local area Network (WLAN, wireless Local Area Networks) wireless interface, a worldwide interoperability for microwave access (Wi-MAX, world Interoperability for Microwave Access) interface, an ethernet interface, a universal serial bus (USB, universal Serial Bus) interface, a cellular Network interface, a bluetooth interface, a Near Field Communication (NFC) interface, and so forth.

In one embodiment of the present description, the above-described components of computing device 1300, as well as other components not shown in FIG. 13, may also be connected to each other, such as by a bus. It should be understood that the block diagram of the computing device illustrated in FIG. 13 is for exemplary purposes only and is not intended to limit the scope of the present description. Those skilled in the art may add or replace other components as desired.

Computing device 1300 may be any type of stationary or mobile computing device including a mobile computer or mobile computing device (e.g., tablet, personal digital assistant, laptop, notebook, netbook, etc.), mobile phone (e.g., smart phone), wearable computing device (e.g., smart watch, smart glasses, etc.), or other type of mobile device, or a stationary computing device such as a desktop computer or personal computer (PC, personal Computer). Computing device 1300 may also be a mobile or stationary server.

Wherein the processor 1320 is configured to execute a computer program/instruction which, when executed by the processor, implements the steps of the task processing method or legal question answering method or task processing model training method or task processing model based information processing method described above.

The foregoing is a schematic illustration of a computing device of this embodiment. It should be noted that, the technical solution of the computing device belongs to the same concept as the technical solution of the task processing method, the legal question answering method, the task processing model training method and the information processing method based on the task processing model, and details of the technical solution of the computing device which are not described in detail can be referred to the description of the technical solution of the task processing method, the legal question answering method, the task processing model training method or the information processing method based on the task processing model.

An embodiment of the present disclosure also provides a computer-readable storage medium storing a computer program/instruction that, when executed by a processor, implements the steps of the task processing method or legal question-answering method or task processing model training method or task processing model-based information processing method described above.

The above is an exemplary version of a computer-readable storage medium of the present embodiment. It should be noted that, the technical solution of the storage medium belongs to the same concept as the technical solution of the task processing method, the legal question answering method, the task processing model training method and the information processing method based on the task processing model, and details of the technical solution of the storage medium which are not described in detail can be referred to the description of the technical solution of the task processing method, the legal question answering method, the task processing model training method or the information processing method based on the task processing model.

An embodiment of the present disclosure further provides a computer program product, including a computer program/instruction, which when executed by a processor, implements the steps of the task processing method or legal question-answering method or task processing model training method or information processing method based on a task processing model.

The foregoing is a schematic version of a computer program product of this embodiment. It should be noted that, the technical solution of the computer program product and the technical solution of the task processing method, the legal question answering method, the task processing model training method and the information processing method based on the task processing model belong to the same concept, and the details of the technical solution of the computer program product, which are not described in detail, can be referred to the description of the technical solution of the task processing method, the legal question answering method, the task processing model training method or the information processing method based on the task processing model.

The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

The computer instructions include computer program code that may be in source code form, object code form, executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the content of the computer readable medium can be increased or decreased appropriately according to the requirements of the patent practice, for example, in some areas, according to the patent practice, the computer readable medium does not include an electric carrier signal and a telecommunication signal.

It should be noted that, for simplicity of description, the foregoing method embodiments are all expressed as a series of combinations of actions, but it should be understood by those skilled in the art that the embodiments are not limited by the order of actions described, as some steps may be performed in other order or simultaneously according to the embodiments of the present disclosure. Further, those skilled in the art will appreciate that the embodiments described in the specification are all preferred embodiments, and that the acts and modules referred to are not necessarily all required for the embodiments described in the specification.

In the foregoing embodiments, the descriptions of the embodiments are emphasized, and for parts of one embodiment that are not described in detail, reference may be made to the related descriptions of other embodiments.

The preferred embodiments of the present specification disclosed above are merely used to help clarify the present specification. Alternative embodiments are not intended to be exhaustive or to limit the invention to the precise form disclosed. Obviously, many modifications and variations are possible in light of the teaching of the embodiments. The embodiments were chosen and described in order to best explain the principles of the embodiments and the practical application, to thereby enable others skilled in the art to best understand and utilize the invention. This specification is to be limited only by the claims and the full scope and equivalents thereof.

Claims

1. A task processing method, comprising:

task data of a target task are obtained;

Inputting the task data into a task processing model to obtain a task processing result of the target task, wherein the task processing model is trained based on a plurality of sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively, the sample anti-facts data and the sample anti-facts results are obtained by performing anti-fact reasoning on the sample task data by using anti-fact guiding information, the anti-fact guiding information is screened from a plurality of candidate guiding information based on the sample task data, the candidate guiding information is guiding information which has a relation with task attribute information of the sample task, the candidate guiding information is obtained based on the task attribute information and reference guiding information, and the reference guiding information is at least one of guiding information determined based on task types of the sample task and guiding information preset by a user according to priori knowledge.

2. The method according to claim 1, wherein the inputting the task data into a task processing model, before obtaining the task processing result of the target task, further comprises:

Inputting the plurality of sample anti-fact data into a task processing model to obtain sample prediction results respectively corresponding to the plurality of sample anti-fact data;

3. The method of claim 2, the obtaining a plurality of sample counterfactual data and sample counterfactual results respectively corresponding to the plurality of sample counterfactual data, comprising:

acquiring a plurality of sample data pairs, wherein the sample data pairs comprise sample task data and inverse fact guide information corresponding to the sample task data;

And inputting the plurality of sample data pairs into a task generation model to obtain a plurality of sample anti-fact data and sample anti-fact results respectively corresponding to the plurality of sample anti-fact data, wherein the sample anti-fact data and the sample data pairs are in one-to-one correspondence.

4. A method according to claim 3, the acquiring a plurality of pairs of sample data comprising:

Inputting task attribute information of the sample task and reference guide information corresponding to the sample task into a task generation model to obtain a plurality of candidate guide information corresponding to the sample task data, wherein the reference guide information corresponding to the sample task is determined based on the task type of the sample task;

And constructing a plurality of sample data pairs according to the sample task data and the inverse fact guiding information of the plurality of sample tasks.

5. The method of claim 3, the sample task data comprising sample chart questions and sample chart annotations, the counterfactual guide information comprising chart counterfactual guide information;

The step of generating a model for the input task by the plurality of sample data, obtaining a plurality of sample anti-facts data and sample anti-facts results corresponding to the plurality of sample anti-facts data respectively, includes:

And inputting the sample chart problem, the sample chart annotation and the chart anti-facts guiding information into a task generating model to obtain sample chart anti-facts data and sample chart anti-facts results corresponding to the sample chart anti-facts data.

6. The method of claim 2, wherein before inputting the plurality of sample anti-facts data into the task processing model to obtain sample prediction results corresponding to the plurality of sample anti-facts data, further comprises:

Screening the plurality of sample anti-facts data according to the verification result to obtain screened plurality of sample anti-facts data;

Inputting the plurality of sample anti-fact data into a task processing model to obtain sample prediction results respectively corresponding to the plurality of sample anti-fact data, wherein the method comprises the following steps:

7. The method according to claim 1, wherein the inputting the task data into a task processing model, after obtaining the task processing result of the target task, further comprises:

The task processing result is sent to a client;

receiving result feedback information sent by the client, wherein the result feedback information is information for feeding back the task processing result based on the task information of the target task;

and carrying out parameter adjustment on the task processing model by utilizing the model optimization data.

8. The method of claim 7, wherein constructing model optimization data based on the result feedback information comprises:

generating optimization prompt information according to the result feedback information, wherein the optimization prompt information is used for guiding the client to send model optimization data for optimizing the task processing model;

And sending the optimization prompt information to the client, and receiving the model optimization data sent by the client based on the optimization prompt information.

9. The method according to claim 1, wherein the inputting the task data into a task processing model, after obtaining the task processing result of the target task, further comprises:

marking the key information in the task processing result to obtain an updated task processing result;

and sending the updated task processing result to the client.

10. The method according to claim 1, wherein the inputting the task data into a task processing model, after obtaining the task processing result of the target task, further comprises:

The task processing result is sent to a client;

Receiving modification information sent by the client, wherein the modification information is used for modifying the task processing result;

11. A legal question-answering method, comprising:

acquiring a question to be answered of a target legal task;

Inputting the questions to be answered into a back facts question answering model to obtain answer results of the questions to be answered, wherein the back facts question answering model is trained based on a plurality of sample back facts data and sample back facts results corresponding to the sample back facts data respectively, the sample back facts data and the sample back facts results are obtained by carrying out back facts reasoning on sample task data by using back facts guiding information for a task generating model, the back facts guiding information is screened from a plurality of candidate guiding information based on the sample task data, the candidate guiding information is guiding information which has a relation with task attribute information of a sample task, the candidate guiding information is obtained based on the task attribute information and reference guiding information, and the reference guiding information is at least one of guiding information determined based on task types of the sample task and guiding information preset by a user according to priori knowledge.

12. The method of claim 11, wherein the inputting the question to be answered into a back facts question answering model, after obtaining the answer result of the question to be answered, further comprises:

The reply result is sent to the client;

Receiving result feedback information sent by the client, wherein the result feedback information is information for feeding back the reply result based on task information of the target legal task;

and sending the result feedback information to a model training platform, wherein the model training platform is used for carrying out parameter adjustment on the anti-fact question-answering model by utilizing the result feedback information.

13. A task processing model training method, comprising:

obtaining a plurality of sample anti-fact data and sample anti-fact results corresponding to the sample anti-fact data respectively, wherein the sample anti-fact data and the sample anti-fact results are obtained by performing anti-fact reasoning on sample task data by using anti-fact guide information for a task generation model, the anti-fact guide information is obtained by screening a plurality of candidate guide information based on the sample task data, the candidate guide information is guide information which has a relation with task attribute information of a sample task, the candidate guide information is obtained based on the task attribute information and reference guide information, and the reference guide information is at least one of guide information determined based on task types of the sample task and guide information preset by a user according to priori knowledge;

14. An information processing method based on a task processing model, comprising:

receiving a task generation request, wherein the task generation request comprises request information;

Acquiring a task processing model based on the request information, wherein the task processing model is trained based on a plurality of sample anti-fact data and sample anti-fact results corresponding to the sample anti-fact data respectively, the sample anti-fact data and the sample anti-fact results are obtained by performing anti-fact reasoning on sample task data by a task generating model through anti-fact guiding information, the anti-fact guiding information is screened from a plurality of candidate guiding information based on the sample task data, the candidate guiding information is guiding information which has a relation with task attribute information of a sample task, the candidate guiding information is obtained based on the task attribute information and reference guiding information, and the reference guiding information is at least one of guiding information determined based on task types of the sample task and guiding information preset by a user according to priori knowledge;

And generating task information based on the task processing model, wherein the task information is used for executing a target task.

15. The method of claim 14, the request information comprising a task scenario identification of a target task, or a task model identification;

the task processing model obtaining step comprises the steps of:

And searching a task processing model from the model library based on the task model identification.

16. The method of claim 14, the request information comprising a plurality of sample anti-facts data for a target task and sample anti-facts results for the plurality of sample anti-facts data, respectively;

the task processing model obtaining step comprises the steps of:

And training the task processing model corresponding to the target task based on the sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively to obtain a task processing model after training.

17. A model training platform comprising a request interface unit and a model unit;

the request interface unit is used for receiving a model training request, wherein the model training request comprises request information;

The model unit is used for acquiring a task processing model; training the task processing model based on the request information to obtain a task processing model with completed training, wherein the request information comprises a plurality of sample anti-facts data and sample anti-facts results corresponding to the sample anti-facts data respectively, the sample anti-facts data and the sample anti-facts results are obtained by carrying out anti-fact reasoning on sample task data by using anti-fact guiding information for a task generating model, the anti-fact guiding information is obtained by screening a plurality of candidate guiding information based on the sample task data, the candidate guiding information is guiding information which has a relation with task attribute information of a sample task, the candidate guiding information is obtained based on the task attribute information and reference guiding information, and the reference guiding information is at least one of guiding information determined based on task types of the sample task and guiding information preset by a user according to priori knowledge.

18. The model training platform of claim 17, further comprising a model library;

the model unit is further used for storing the trained task processing model to the model library.

19. A computing device, comprising:

A memory and a processor;

the memory is configured to store a computer program/instruction, the processor being configured to execute the computer program/instruction, which when executed by the processor, performs the steps of the method of any one of claims 1 to 10 or any one of claims 11 to 12 or claim 13 or any one of claims 14 to 16.

20. A computer readable storage medium storing a computer program/instruction which when executed by a processor performs the steps of the method of any one of claims 1 to 10 or any one of claims 11 to 12 or claim 13 or any one of claims 14 to 16.

21. A computer program product comprising computer programs/instructions which when executed by a processor implement the steps of the method of any one of claims 1 to 10 or any one of claims 11 to 12 or claim 13 or any one of claims 14 to 16.