CN118133797A

CN118133797A - Concept verification method, device, equipment and medium based on pre-training language model

Info

Publication number: CN118133797A
Application number: CN202410203870.2A
Authority: CN
Inventors: 冯娟; 文水德; 李楠
Original assignee: Shenzhen International Graduate School of Tsinghua University
Current assignee: Shenzhen International Graduate School of Tsinghua University
Priority date: 2024-02-23
Filing date: 2024-02-23
Publication date: 2024-06-04

Abstract

The embodiment of the application provides a concept verification method, device, equipment and medium based on a pre-training language model, and relates to the field of artificial intelligence. The method comprises the following steps: obtaining scheme data to be verified, inputting the scheme data to be verified into a fine-tuned pre-training language model, obtaining an analysis result of the scheme data to be verified in a concept verification stage through the pre-training language model, and obtaining a concept verification report of the scheme data to be verified based on the analysis result. The pre-training language model is obtained through fine adjustment of the following steps: collecting domain corpus data; performing data preprocessing on the domain corpus data to obtain training sample data; and fine tuning the pre-training language model based on the training sample data to obtain the fine-tuned pre-training language model. Through the pre-training model, the intelligentization, the efficiency and the accuracy of the concept verification can be improved.

Description

Concept verification method, device, equipment and medium based on pre-training language model

Technical Field

The present invention relates to the field of artificial intelligence, and in particular, to a method, apparatus, device, and medium for concept verification based on a pre-training language model.

Background

In recent technological developments, artificial Intelligence (AI), especially in the field of machine learning and deep learning, has demonstrated its revolutionary potential. Especially large pre-trained models (e.g., GPT-3 and BERT), have achieved significant achievements in many areas of language processing, image recognition, and the like. These models are capable of learning patterns from large amounts of data, performing complex tasks ranging from simple classification to generating high quality text.

AI not only accelerates scientific and technical innovation, but also improves research and development flows. The method can process and analyze huge data sets, and provides a profound insight for the fields of new material development, drug discovery, market trend prediction and the like. In addition, AI also presents great potential in automating research flows, refining useful information in complex data, and aiding decisions. In particular, in the Proof of Concept (PoC) stage, the application of AI facilitates the rapid and accurate assessment of new technologies or achievements in scientific research. While the conventional art provides a standardized procedure to determine whether a product meets target requirements, it does not take full advantage of AI to accelerate the concept authentication process or to improve its efficiency and accuracy.

Disclosure of Invention

In view of the above, the present invention provides a method, apparatus, device and medium for concept verification based on a pre-training language model, which aims to improve the intelligentization, efficiency and accuracy of the concept verification.

In a first aspect, an embodiment of the present application proposes a concept verification method based on a pre-training language model, the method including the steps of:

obtaining scheme data to be verified;

Inputting the scheme data to be verified into a fine-tuned pre-training language model to obtain an analysis result of the scheme data to be verified in a concept verification stage through the pre-training language model;

obtaining a concept verification report of the scheme data to be verified based on the analysis result;

The pre-training language model is obtained through fine adjustment of the following steps:

Collecting domain corpus data;

Performing data preprocessing on the domain corpus data to obtain training sample data;

And fine tuning the pre-training language model based on the training sample data to obtain the fine-tuned pre-training language model.

With reference to the first aspect, in one possible implementation manner of the first aspect, the pre-training language model includes a plurality of concept-verification agents;

The step of inputting the scheme data to be verified into the fine-tuned pre-training language model to obtain an analysis result of the scheme data to be verified in the concept verification stage through the pre-training language model, comprising the following steps:

Inputting the scheme data to be verified into a fine-tuned pre-training language model to generate a plurality of analysis results of the scheme data to be verified in a concept verification stage through each concept verification agent;

The analysis result includes at least one of:

value claim canvas, macro analysis model, user trip map, blue sea warfare thumbnail, product effect map, business model canvas, stakeholder analysis model.

With reference to the first aspect, in one possible implementation manner of the first aspect, the inputting the to-be-verified solution data into the fine-tuned pre-training language model to generate, by each of the concept-verification agents, a plurality of analysis results of the to-be-verified solution data in a concept-verification stage includes:

Inputting scheme data to be verified into a first concept-verifying agent to generate a value-claiming canvas through the first concept-verifying agent;

Inputting the value claim canvas into a second concept-verifying agent to generate a macro-analysis model through the second concept-verifying agent;

inputting the value claim canvas into a third concept-verifying agent to generate a user-itinerary map through the third concept-verifying agent;

Inputting the fusion result of the value claiming canvas and the macro analysis model into a fourth concept-verifying agent to generate a blue sea warfare sketch through the fourth concept-verifying agent;

inputting the value claim canvas, the macro analysis model, the user trip map, and the blue sea warfare thumbnail into a fifth concept-verifying agent to generate a product effect diagram through the fifth concept-verifying agent;

Inputting the value claim canvas, the macro analysis model, the user trip map, and the blue sea warfare thumbnail into a sixth concept-verifying agent to generate a business model canvas through the sixth concept-verifying agent;

the business model canvas is input to a seventh concept-verifying agent to generate a stakeholder analysis model through the seventh concept-verifying agent.

With reference to the first aspect, in a possible implementation manner of the first aspect, the obtaining a proof of concept report of the scheme data to be verified based on the analysis result includes:

Carrying out data analysis and fusion processing on the value claiming canvas, the macro analysis model, the user journey map, the blue sea warfare sketch, the product effect diagram, the business mode canvas and the stakeholder analysis model to obtain a structured data frame;

simulating different markets and user behavior scenes based on the structured data framework, and generating a concept verification report of the scheme data to be verified according to a concept verification report template.

With reference to the first aspect, in a possible implementation manner of the first aspect, the generating a concept-verification report of the scheme data to be verified according to a concept-verification report template includes:

acquiring a predefined concept verification report template;

generating a concept verification report framework according to the concept verification report template, wherein the report framework comprises a plurality of chapter contents;

And generating concept verification content in each chapter content in the concept verification report framework to obtain a concept verification report of the scheme data to be verified.

With reference to the first aspect, in one possible implementation manner of the first aspect, the performing data preprocessing on the domain corpus data to obtain training sample data includes:

performing text cleaning on the domain corpus data to obtain the domain corpus data after text cleaning;

And generating structured training sample data according to the text-cleaned domain corpus data.

With reference to the first aspect, in one possible implementation manner of the first aspect, the performing fine tuning on the pre-training language model based on the training sample data to obtain the fine-tuned pre-training language model includes:

acquiring labeling information of the training sample data;

performing word segmentation and vectorization on the training sample data to obtain word vectors corresponding to the training sample data;

and fine tuning the pre-training language model according to word vectors corresponding to the training sample data and the labeling information of the training sample data so as to update model parameters in the pre-training language model and obtain the fine-tuned pre-training language model.

In a second aspect, an embodiment of the present application proposes a concept-verifying apparatus based on a pre-trained language model, the apparatus comprising:

the acquisition module is used for acquiring scheme data to be verified;

The concept verification module is used for inputting the scheme data to be verified into the fine-tuned pre-training language model so as to obtain an analysis result of the scheme data to be verified in a concept verification stage through the pre-training language model;

The report generation module is used for obtaining a concept verification report of the scheme data to be verified based on the analysis result; the pre-training language model is obtained through fine adjustment of the following steps: collecting domain corpus data; performing data preprocessing on the domain corpus data to obtain training sample data; and fine tuning the pre-training language model based on the training sample data to obtain the fine-tuned pre-training language model.

In a third aspect, an embodiment of the present application proposes an electronic device, including a memory and a processor, where the memory stores a computer program or instructions, and where the processor implements a method for concept verification based on a pre-trained language model as described in the first aspect above when the processor executes the computer program or instructions.

In a fourth aspect, an embodiment of the present application proposes a computer readable storage medium, on which a computer program or instructions is stored, which when executed by a processor, implements a method for concept verification based on a pre-trained language model as described in the first aspect above.

The method, the device, the equipment and the medium for verifying the concept based on the pre-training language model are provided by the embodiment of the application, firstly, the scheme data to be verified is acquired, the scheme data to be verified is input into the pre-training language model after fine adjustment, so that the analysis result of the scheme data to be verified in the concept verification stage is obtained through the pre-training language model, and the concept verification report of the scheme data to be verified is obtained based on the analysis result. The pre-training language model is obtained through fine adjustment of the following steps: collecting domain corpus data; performing data preprocessing on the domain corpus data to obtain training sample data; and fine tuning the pre-training language model based on the training sample data to obtain the fine-tuned pre-training language model. Through the pre-training model, the intelligentization, the efficiency and the accuracy of the concept verification can be improved.

Drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention.

FIG. 1 is a schematic flow diagram of a concept verification method based on a pre-training language model according to an embodiment of the present application;

FIG. 2 is a schematic diagram of a process for fine tuning a pre-trained language model according to an embodiment of the present application;

FIG. 3 is a schematic diagram of a process for generating a plurality of analysis results according to an embodiment of the present application;

FIG. 4 is a schematic diagram of a process for generating a proof of concept report provided by an embodiment of the present application;

FIG. 5 is a schematic diagram of a process for generating a fine-tuned pre-trained language model according to an embodiment of the present application;

FIG. 6 is a schematic structural diagram of a concept-verifying device based on a pre-training language model according to an embodiment of the present application;

fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

It should be noted that although functional block division is performed in a device diagram and a logic sequence is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the block division in the device, or in the flowchart. The terms first, second and the like in the description and in the claims and in the above-described figures, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.

First, several nouns involved in the present application are parsed:

Artificial intelligence (ARTIFICIAL INTELLIGENCE, AI for short) is a branch of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. This includes the fields of machine learning (MACHINE LEARNING), deep learning (DEEP LEARNING), natural language processing (Natural Language Processing), computer Vision (Computer Vision), and the like. Machine learning: techniques for a computer system to utilize data to improve its performance without explicit programming. It allows the computer to learn from the data and make predictions or decisions. Deep learning: a sub-field of machine learning, which uses neural network structures similar to the human brain to learn complex patterns of data. Deep learning has achieved significant achievements in the fields of image recognition, speech recognition, natural language processing, and the like. Natural language processing: techniques to enable a computer to understand, interpret, and generate human language. This includes speech recognition, machine translation, emotion analysis, etc. Computer vision: enabling a computer to obtain information from an image or multidimensional data and understand the content of such information. This has wide application in the fields of automatic driving automobiles, face recognition, image classification, and the like.

The goal of artificial intelligence is to create intelligent systems capable of performing complex tasks that typically require human intelligence such as visual perception, language understanding, decision making, and learning. With the development of technology, artificial intelligence has had profound effects in many areas, including medical, financial, educational, recreational, and manufacturing industries, among others.

Text cleansing (Text Cleaning) is a key step in data preprocessing that involves performing a series of processes on raw Text data to improve data quality, ready for subsequent analysis, modeling, or machine learning tasks. The main purpose of text cleansing is to remove noise, correct errors, standardize data formats, and extract useful information. This process typically includes the following steps: removing noise, removing stop words, extracting word stems (Stemming), reducing word shapes (Lemmatization), removing punctuation marks, converting cases and cases, identifying entities, removing HTML labels and normalizing texts. Emotion analysis: in some cases, emotion analysis may be required on the text to determine its emotion tendencies, which is particularly important for emotion analysis tasks. The purpose of text cleansing is to ensure the quality and consistency of the data set, thereby improving the accuracy and efficiency of the data analysis and machine learning model.

Word segmentation (Tokenization): word segmentation is the process of breaking a text string into smaller units (called "tokens" or "token"). These elements may be words, phrases, symbols, or other meaningful pieces of text. In east asian languages such as chinese, word segmentation is particularly important because the writing habit of these languages does not separate words by spaces like english, and thus the boundary of each word needs to be determined by word segmentation. The accuracy of word segmentation has direct influence on subsequent text analysis tasks such as emotion analysis, topic modeling and the like. Incorrect segmentation may lead to misunderstanding of the model, thereby affecting the analysis results.

Vectorization (Vectorization): vectorization is the process of converting text data into numerical form so that a computer can process and analyze the data. In NLP, text data typically needs to be converted into numerical vectors for input into a machine learning model. Common vectorization methods include Bag of Words model (Bag of Words), TF-IDF (Term Frequency-Inverse Document Frequency), word2Vec, gloVe, BERT, and the like. The bag of words model simply represents text as a fixed-size vector with each element in the vector corresponding to the frequency of occurrence of a word element. TF-IDF takes into account the frequency of the tokens in the document and the rarity in the whole corpus. Models such as Word2Vec and GloVe then attempt to capture contextual relationships between tokens, generating a richer semantic representation. Vectorization applies not only to words, but also to sentences, paragraphs and even whole documents, depending on the requirements of the analysis task.

Preprocessing (Preprocess ing) is an important step in data analysis and machine learning, which involves performing a series of operations on raw data to improve data quality, simplify model training processes, enhance model performance, or meet the requirements of a particular algorithm. The purpose of the preprocessing is to convert the data into a form more suitable for analysis for more efficient subsequent data processing and modeling. Common steps of pretreatment include: data cleaning,

Data conversion, feature engineering, data integration, data protocol, data discretization and data enhancement. Preprocessing is critical to ensure the success of data analysis and machine learning projects, as it directly affects the final performance of the model. Good preprocessing strategies can improve the accuracy, robustness and interpretability of the model.

In view of the above, the present application provides a method, apparatus, device and medium for concept verification based on a pre-training language model, which aims to improve the intelligentization, efficiency and accuracy of the concept verification.

The embodiment of the application provides a concept verification method, device, equipment and medium based on a pre-training language model, and specifically describes the method based on the pre-training language model in the embodiment of the application through the following embodiment.

The embodiment of the application provides a concept verification method of a pre-training language model, and relates to the field of artificial intelligence. The concept verification method based on the pre-training language model provided by the embodiment of the application can be applied to a terminal, a server side and software running in the terminal or the server side. In some embodiments, the terminal may be a smart phone, tablet, notebook, desktop, etc.; the server side can be configured as an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, basic cloud computing services such as big data and artificial intelligent platforms and the like; the software may be an application or the like that implements a concept-verification method based on a pre-trained language model, but is not limited to the above form.

The application is operational with numerous general purpose or special purpose computer system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like. The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

Referring to fig. 1, fig. 1 is a flowchart of a concept verification method based on a pre-training language model according to an embodiment of the present application. As shown in fig. 1, the concept verification method based on the pre-training language model according to the embodiment of the present application includes, but is not limited to, steps S110 to S130, and each step is described below in turn.

Step S110: and obtaining scheme data to be verified.

It can be understood that the scheme data to be verified is a target object of the extraction process. The scheme data to be verified may be automatically identified data. The acquired scheme data to be verified at least comprises one of the following data: text data, image data, sound data, and video data.

It will be appreciated that the scheme data to be verified is obtained and the resulting scheme data to be verified is used for processing. And collecting different types of scheme data to be verified so as to ensure the integrity of the scheme data to be verified.

It should be noted that, after obtaining the scheme data to be verified, the data to be verified is processed, and the processing is to preprocess the data, including data cleaning, filtering, denoising, normalization and the like, so as to ensure the accuracy and consistency of the data. The embodiment of the application is not limited to a specific method of processing.

Step S120: and inputting the scheme data to be verified into the fine-tuned pre-training language model to obtain an analysis result of the scheme data to be verified in the concept verification stage through the pre-training language model.

It can be understood that the scheme data to be verified is used as initial input data and is input into the fine-tuned pre-training language model for processing, so that an analysis result of the scheme data to be verified in the concept verification stage is obtained.

It should be noted that, the pre-training language model adopted in step S120 is pre-trained, specifically may be a pre-training language model obtained by training based on a large-scale text data set, and the training process of the pre-training language model is not limited in the embodiment of the present application.

Step S130: and obtaining a concept verification report of the scheme data to be verified based on the analysis result.

It can be understood that the concept verification report of the scheme data to be verified is obtained through the analysis result.

Referring to fig. 2, fig. 2 is a schematic diagram illustrating a process for fine tuning a pre-training language model according to an embodiment of the present application. The process of fine tuning a pre-trained language model according to an embodiment of the present application is described below with reference to the process of fine tuning a pre-trained language model provided in fig. 2.

In some embodiments, fine tuning the pre-trained language model may specifically include the following steps S210-S230:

Step S210: and collecting the corpus data of the field.

It is understood that the domain corpus data may be patent literature and scientific paper specific corpus.

Step S220: and carrying out data preprocessing on the domain corpus data to obtain training sample data.

It should be noted that, performing data preprocessing on the domain corpus data to obtain training sample data, including: performing text cleaning on the domain corpus data to obtain text cleaned domain corpus data; and generating structured training sample data according to the text-cleaned domain corpus data.

Step S230: and fine tuning the pre-training language model based on the training sample data to obtain a fine-tuned pre-training language model.

Referring to fig. 5, fig. 5 is a schematic process diagram of a fine-tuned pre-training language model according to an embodiment of the present application. The process of the trimmed pre-training language model of the embodiment of the application is exemplarily described below in connection with the process of the trimmed pre-training language model provided in fig. 5.

In some embodiments, the training sample data is used to fine tune the pre-training language model, so as to obtain a fine-tuned pre-training language model, which specifically includes the following steps S510-S530:

step S510: acquiring labeling information of training sample data;

step S520: word segmentation and vectorization are carried out on the training sample data, so that word vectors corresponding to the training sample data are obtained;

It should be noted that word segmentation is a process of decomposing training sample data into smaller units. These elements include, but are not limited to, words, phrases, symbols, or other meaningful text fragments. The segmentation can improve the accuracy of subsequent text analysis tasks, such as emotion analysis, topic modeling and the like, so that the intelligentization, efficiency and accuracy of the concept verification are improved.

Step S530: and fine tuning the pre-training language model according to word vectors corresponding to the training sample data and the labeling information of the training sample data so as to update model parameters in the pre-training language model and obtain a fine-tuned pre-training language model.

It should be noted that, the pre-training language model includes a plurality of concept verification agents, and the method includes inputting the scheme data to be verified into the fine-tuned pre-training language model to obtain the analysis result of the scheme data to be verified in the concept verification stage through the pre-training language model, including: inputting the scheme data to be verified into the fine-tuned pre-training language model to generate a plurality of analysis results of the scheme data to be verified in a concept verification stage through each concept verification agent, wherein the analysis results comprise at least one of the following: value claim canvas, macro analysis model, user trip map, blue sea warfare thumbnail, product effect map, business model canvas, stakeholder analysis model.

Wherein the value claim canvas (Value Proposition Canvas) characterizes value claims for designing and analyzing a product or service. The value claim canvas includes two key aspects: customer (Customer): including market segments, customer work, challenges, desires, and wishes, etc. Value claim (Value Proposition): including product or service features, advantages, solutions, benefits, and the like. The value claim canvas helps enterprises to better understand customer needs and provide guidance for innovations and improvements by clearly presenting the value of a product or service to a customer.

The macro analysis model is a model for researching and explaining the operation and development of the whole economic system. The macro-analysis model generally relates to global economic indicators of countries or regions, such as global national production (GDP), expansion rates of currency, rates of loss, trade balances, etc., as well as influencing factors and outcomes of macro-economic policies. The macro analysis model can help decision makers, investors and analysts identify trends in the external environment for better strategy and planning.

A user itinerary map (Customer Journey Map) is used to visualize the interaction process between the user and the product or service. The user trip map describes the entire process that the user undergoes when using the product or service from the user's perspective, from an initial contact and awareness phase to a use, purchase and possibly feedback phase. The user trip map generally describes the user experience process from left to right on a time axis. The user trip map typically includes the following key components: user goals, user behaviors, user experiences, points of contact, points of pain and opportunities, and timelines. The user itinerary map can help the enterprise understand the behavior, feelings, demands and pain points of the user at different contact points, thereby optimizing the user experience (UX) and improving the user satisfaction.

Blue Ocean wars outline (Blue Ocean STRATEGY CANVAS) is used to visualize the difference between the existing market (red sea) and the future innovation market (Blue sea). The blue sea warfare sketch is usually a two-dimensional chart comprising two axes of a Value Curve (Value Curve) and Key Factors (Key Factors), wherein the Value Curve (Value Curve) is a horizontal axis representing the performance of a product or service on different Key Factors, which are usually representatives of industry standard or customer Value ideas such as price, performance, convenience, etc. The key element is the vertical axis, representing the key element in evaluating a product or service, which is typically the factor most appreciated by customers in the industry, such as product characteristics, service level, brand reputation, etc. The blue sea warfare sketch can help enterprises clearly see the conditions of each key element in the current market, so that opportunities of innovation and differentiation are found, brand new market space is created, the enterprises are helped to avoid the red sea market with strong competition and enter the blue sea market with relatively no competition, and long-term competitive advantage and profit space are obtained.

The product effect diagram is used to show the appearance, function and characteristics of the product. The goal of the product effect map is to convey clear, accurate information to potential consumers so that consumers can understand the appearance, characteristics, and use of the product and make purchasing decisions.

A business mode canvas (Business Model Canvas) is used to visualize a business mode, which typically includes the following nine elements: customer segments (Customer Segments), value claims (Value Propositions), channels (Channels), customer relationships (Customer Relationships), revenue sources (Revenue Streams), key Resources (Key Resources), key activities (KEY ACTIVITIES), key partners (KEY PARTNERSHIPS), cost structures (Cost structures), business model canvas is intended to help businesses understand, design and innovate business models more clearly, discover potential opportunities and threats, and better plan business development strategies.

The stakeholder analysis model is used for identifying, evaluating and managing relevant stakeholders in the project, and helps enterprises to know who will be influenced by their decisions or actions and the influence degree of the enterprises on the project, so that the enterprises can better communicate with the stakeholders, manage the expectations of the stakeholders, and ensure the success and sustainability of the project.

Referring to fig. 3, fig. 3 is a schematic diagram illustrating a process of generating a plurality of analysis results according to an embodiment of the application. The process of generating a plurality of analysis results according to the embodiment of the present application will be exemplarily described with reference to the process of generating a plurality of analysis results provided in fig. 3.

In some embodiments, the to-be-verified scheme data is input into the fine-tuned pre-training language model to generate a plurality of analysis results of the to-be-verified scheme data in the concept verification stage through each concept verification agent, which may specifically include the following steps S310-S370:

step S310: the scheme data to be verified is entered into the first concept-verifying agent to generate a value claim canvas through the first concept-verifying agent.

Step S320: the value claim canvas is input to a second concept-verifying agent to generate a macro analysis model through the second concept-verifying agent.

Step S330: the value claim canvas is input into a third concept-verifying agent to generate a user-itinerary map through the third concept-verifying agent.

Step S340: and inputting the fusion result of the value claiming canvas and the macroscopic analysis model into a fourth concept-verifying agent to generate the blue sea warfare sketch through the fourth concept-verifying agent.

Step S350: the value claim canvas, the macro analysis model, the user trip map, and the blue sea warfare thumbnail are input to a fifth concept-verifying agent to generate a product effect map through the fifth concept-verifying agent.

Step S360: the value claim canvas, the macro analysis model, the user trip map, and the blue sea warfare thumbnail are input to a sixth concept-verifying agent to generate a business model canvas through the sixth concept-verifying agent.

Step S370: the business mode canvas is input to the seventh concept-verifying agent to generate a stakeholder analysis model through the seventh concept-verifying agent.

In a specific embodiment, obtaining a concept-verification report of the scheme data to be verified based on the analysis result includes:

A concept-verification report of the scheme data to be verified is generated based on the value claim canvas, the macro analysis model, the user trip map, the blue sea warfare thumbnail, the product effect map, the business model canvas, and the stakeholder analysis model.

It should be noted that, based on the value claim canvas, the macro analysis model, the user trip map, the blue sea warfare thumbnail, the product effect map, the business model canvas, and the stakeholder analysis model, generating the concept verification report of the scheme data to be verified includes: carrying out data analysis and fusion processing on a value claiming canvas, a macroscopic analysis model, a user journey map, a blue sea warfare sketch, a product effect map generation, a business mode canvas and a stakeholder analysis model to obtain a structured data frame; simulating different markets and user behavior scenes based on the structured data framework, and generating a concept verification report of scheme data to be verified according to the concept verification report template.

In one embodiment, the results generated by different agents, including but not limited to, value claim canvas, business model canvas, macro analysis model, stakeholder analysis model, user trip map, blue sea warfare thumbnail, and product effect map are first collected. Each portion is stored in a specific format, which may be a Json file, for convenient subsequent processing. Next, a high-level data processing Agent parses, compares, and fuses all the data collected. The method can identify the relevance among different materials, extract key information points and fuse the data and the information into a unified structured data frame according to a preset algorithm and logic. And finally, based on the fused data frames, the system simulates different markets and user behavior scenes. With the ability to generate an AI, product performance, user feedback, and market opportunities in various possible commercial environments are simulated as the basis for the content of the report.

Referring to fig. 4, fig. 4 is a schematic diagram of a process for generating a proof of concept report according to an embodiment of the present application. The process of generating a proof of concept report of an embodiment of the present application will be exemplarily described with reference to the process of generating a proof of concept report provided in fig. 4.

In some embodiments, the method for generating the concept-verification report of the scheme data to be verified according to the concept-verification report template may specifically include the following steps S410 to S430:

step S410: a predefined proof of concept report template is obtained.

Step S420: a concept-verification report framework is generated from the concept-verification report template, the report framework including a plurality of chapter content.

Step S430: and generating concept verification content in each chapter content in the concept verification report framework to obtain a concept verification report of scheme data to be verified.

Before the report is started to be generated, a framework of the concept verification report is designed according to a preset template or a dynamically generated structure. This framework includes, but is not limited to, covering a plurality of parts such as summaries, analysis result interpretations, chart interpretations, model predictions, scene simulation results, and the like. According to the reporting framework, the content of each chapter is automatically populated using a generative AI technique. This includes visualizing data (e.g., charts, images), generating text description analysis results and predictions, and interpreting the meaning of complex models. Finally, based on the report content, multi-modal outputs such as video and PPT are generated using AI generation techniques. Videos include, but are not limited to, dynamic charts, narrative text, and animations that simulate scenes; the PPT presents the report content in a more formal and static form. After the report is generated, the system performs a round of verification, evaluates the integrity, accuracy and understandability of the report, and performs necessary adjustment and optimization according to feedback. Through the concept verification of the pre-training language model, a large amount of complex data and model information can be converted into a visual and easily understood concept verification report, and the efficiency and effect of cross-domain concept verification are greatly improved.

In an embodiment, the patent text CN116874803a is used as concept solution data, the patent text is input into a fine-tuned pre-training language model, so that an analysis result of the patent text in a concept verification stage is obtained through the pre-training language model, a concept verification report of the patent text is obtained based on the analysis result, and the concept verification report includes a report version, a patent name, basic information (application number, applicant, inventor, application date), types (utility model patent, appearance patent), the field and a plurality of specific sections, such as brief description, description of innovation points of the patent, description of what problem the utility model solves, closest technology or patent description, detailed information, obvious difference between the utility model and the prior art or products, value of the utility model on companies and markets, how the utility model is applied commercially, best application suggestion and detailed reason and best application product effect diagram.

For example, in the chapter content briefly described, the content of the concept verification thereof is: discloses an antibacterial tree-like polymer for a store, a preparation method and application thereof, and belongs to the field of biomedicine. The invention prepares an ideal antibacterial polyester polymer by grafting amino groups on the branched chains of dendritic polymers: the modification of the biopolymer is realized by utilizing an antibacterial mechanism that the protonation of the amino group contains positive charges and can be effectively inserted into a bacterial cell membrane structure to cause bacterial lysis, so that the grafting abundance of the amino group of the macromolecule is improved, and the antibacterial effect of the polyester macromolecule is enhanced. The innovation point section of the patent is that the grafting is prepared on a branched chain of a tree-like polymer, and a positive charge structure is effectively inserted into a bacterial thin thermal film structure through protonation of amino groups, so that bacteria are cracked, and the antibacterial effect of a ship-gathering polymer is improved by a modified well of the biopolymer. The invention solves the problems of improving the grafting abundance of the amino groups of the polymers and enhancing the antibacterial effect of the polyester polymers, and solves the defects of the prior biological polymers in antibacterial performance through the modification of specific structures.

For example, in the section of how the present invention is commercially applied, its concept-verifying content includes the fields of commercial application in various aspects of medical instruments and consumables, antibacterial textiles, packaging materials, personal care products, biomedical research, environmental protection materials, and the like.

For example, in the section content of best application advice and detailed reasons, applications in the medical instrument and consumable fields are included as major reasons for best business applications, such as high demand, technical adaptability, social impact, business prospects, policy support.

Referring to fig. 6, in some possible embodiments of the present application, there is further provided a concept-verifying apparatus 600 based on a pre-training language model, where the method for performing the above-mentioned concept-verification may be implemented, where the apparatus 600 includes:

an obtaining module 610, configured to obtain scheme data to be verified.

The concept verification module 620 is configured to input the to-be-verified solution data into the fine-tuned pre-training language model, so as to obtain an analysis result of the to-be-verified solution data in the concept verification stage through the pre-training language model.

A report generating module 630, configured to obtain a concept verification report of the scheme data to be verified based on the analysis result; the pre-training language model is obtained through fine adjustment of the following steps: collecting domain corpus data; carrying out data preprocessing on the domain corpus data to obtain training sample data; and fine tuning the pre-training language model based on the training sample data to obtain a fine-tuned pre-training language model.

The concept verification device based on the pre-training language model belongs to the same inventive concept as the concept verification method based on the pre-training language model provided in the above embodiment, and the specific step flow and beneficial effects thereof can be referred to the description of the above embodiment, which is not repeated here

The embodiment of the application also provides electronic equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the concept verification method based on the pre-training language model when executing the computer program. The electronic equipment can be any intelligent terminal including a tablet personal computer, a vehicle-mounted computer and the like.

Referring to fig. 7, fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the application. The electronic device includes:

the processor 710 may be implemented by a general purpose central processing unit (Central Processing Unit, CPU), a microprocessor, an Application SPECIFIC INTEGRATED Circuit (ASIC), or one or more integrated circuits, etc. for executing related programs to implement the technical solutions provided by the embodiments of the present application.

The Memory 720 may be implemented in the form of a Read Only Memory (ROM), a static storage device, a dynamic storage device, or a random access Memory (Random Access Memory, RAM). Memory 720 may store an operating system and other application programs, and when implementing the technical solutions provided in the embodiments of the present specification by software or firmware, relevant program codes are stored in memory 720 and invoked by processor 710 to perform the concept-verification method of the present application based on a pre-trained language model.

Input/output interface 730 for implementing information input and output.

The communication interface 740 is configured to implement communication interaction between the device and other devices, and may implement communication in a wired manner (e.g., USB, network cable, etc.), or may implement communication in a wireless manner (e.g., mobile network, WIFI, bluetooth, etc.).

Bus 750 conveys information between various components of the device (e.g., processor 710, memory 720, input/output interface 730, and communication interface 740).

Wherein processor 710, memory 720, input/output interface 730, and communication interface 740 implement a communication connection among each other within the device via bus 750.

The embodiment of the application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program realizes the concept verification method based on the pre-training language model when being executed by a processor.

The memory, as a non-transitory computer readable storage medium, may be used to store non-transitory software programs as well as non-transitory computer executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory remotely located relative to the processor, the remote memory being connectable to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The method, the device, the equipment and the medium for verifying the concept based on the pre-training language model are provided by the embodiment of the application, firstly, the scheme data to be verified is acquired, the scheme data to be verified is input into the pre-training language model after fine adjustment, so that the analysis result of the scheme data to be verified in the concept verification stage is obtained through the pre-training language model, and the concept verification report of the scheme data to be verified is obtained based on the analysis result. The pre-training language model is obtained through fine adjustment of the following steps: collecting domain corpus data; carrying out data preprocessing on the domain corpus data to obtain training sample data; and fine tuning the pre-training language model based on the training sample data to obtain a fine-tuned pre-training language model. Through the pre-training model, the intelligentization, the efficiency and the accuracy of the concept verification can be improved.

The embodiments described in the embodiments of the present application are for more clearly illustrating the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application, and as those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are applicable to similar technical problems.

It will be appreciated by persons skilled in the art that the embodiments of the application are not limited by the illustrations, and that more or fewer steps than those shown may be included, or certain steps may be combined, or different steps may be included.

The above described apparatus embodiments are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Those of ordinary skill in the art will appreciate that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof.

The terms "first," "second," "third," "fourth," and the like in the description of the application and in the above figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: "a", "b", "c", "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the above-described division of units is merely a logical function division, and there may be another division manner in actual implementation, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other forms.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including multiple instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method of the various embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a random access Memory (Random Access Memory RAM), a magnetic disk, an optical disk, or other various media capable of storing a program.

The preferred embodiments of the present application have been described above with reference to the accompanying drawings, and are not thereby limiting the scope of the claims of the embodiments of the present application. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and spirit of the embodiments of the present application shall fall within the scope of the claims of the embodiments of the present application.

Claims

1. A method of concept verification based on a pre-trained language model, the method comprising:

obtaining scheme data to be verified;

Collecting domain corpus data;

2. The method of claim 1, wherein the pre-trained language model comprises a plurality of concept-verifying agents;

The analysis result includes at least one of:

3. The method of claim 2, wherein said inputting the solution data to be verified into the fine-tuned pre-trained language model to generate a plurality of analysis results of the solution data to be verified in a concept-verification phase by the respective concept-verification agents, respectively, comprises:

4. A method according to claim 3, wherein said obtaining a proof of concept report of said scheme data to be verified based on said analysis result comprises:

5. The method of claim 4, wherein generating the proof of concept report of the scheme data to be verified according to the proof of concept report template comprises:

acquiring a predefined concept verification report template;

6. The method according to claim 1, wherein the performing data preprocessing on the domain corpus data to obtain training sample data includes:

7. The method of claim 1, wherein said fine tuning the pre-training language model based on the training sample data results in a fine tuned pre-training language model, comprising:

acquiring labeling information of the training sample data;

8. A concept-verifying apparatus based on a pre-trained language model, the apparatus comprising:

the acquisition module is used for acquiring scheme data to be verified;

the report generation module is used for obtaining a concept verification report of the scheme data to be verified based on the analysis result;

Collecting domain corpus data;

9. An electronic device comprising a memory storing a computer program or instructions that when executed implement the pre-trained language model-based concept-verification method of any one of claims 1 to 7, and a processor.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program or instructions which, when executed by a processor, implements the pre-trained model-based concept verification method according to any one of claims 1 to 7.