Digital asset risk information processing method and system

Abstract

The invention provides a digital asset risk information processing method and system. The method comprises the following steps: acquiring web2 multi-modal source data (including but not limited to various modalities such as texts, images, voices, videos and the like) and web3 source data; converting the web2 non-text source data into web2 text processing data; filtering the web2 text source data and the web2 text processing data to obtain web2 related risk information related to the digital assets; web2 statistical information is obtained from the web2 text source data and the web2 text processing data; web3 related risk information related to the digital assets is extracted from the web3 source data; inputting the web2 statistical information and the web3 related risk information into a first machine learning model to obtain risk scores related to the digital assets and corresponding risk information features; and inputting the filtered web2 text data and the web3 related risk information into a second machine learning model to obtain a risk analysis and prevention and control strategy related to the digital assets.

Description

Technical Field

[0001] The present invention relates to an information processing method and system, and more particularly to a method and system for processing digital asset risk information. Background Art

[0002] Although the occurrence probability of digital asset risk events may be relatively low, once they occur, they will have a great impact on asset holders. For example, risk events in the business field may lead to corporate bankruptcy, stock market crashes, etc. Early warning signals before the occurrence of risk events are often easily overlooked, thus missing the time window for risk prevention and resulting in significant losses. Summary of the Invention

[0003] In one aspect, the present invention provides a method for processing digital asset risk information. According to an embodiment, the method for processing digital asset risk information of the present invention includes obtaining web2 text source data, web2 non-text source data, and web3 source data; converting the web2 non-text source data into web2 text processing data; filtering the web2 text source data and the web2 text processing data to obtain web2 related risk information related to the digital asset; obtaining web2 statistical information from the web2 text source data and the web2 text processing data; extracting web3 related risk information related to the digital asset from the web3 source data; and inputting the web2 statistical information and the web3 related risk information into a first machine learning model to obtain a risk score related to the digital asset and corresponding risk information features.

[0004] Preferably, the method further includes, after obtaining the risk score related to the digital asset, converting the risk score into a chart and displaying the chart.

[0005] Preferably or additionally, if the risk score exceeds a preset threshold, the method further includes generating and sending out a risk warning information.

[0006] According to another embodiment, the method for processing digital asset risk information of the present invention includes obtaining web2 text source data, web2 non-text source data, and web3 source data; converting the web2 non-text source data into web2 text processing data; filtering the web2 text source data and the web2 text processing data to obtain web2 related risk information related to the digital asset; embedding the web2 related risk information into a prompt template to generate first-level prompt information; extracting web3 related risk information related to the digital asset from the web3 source data; inputting the first-level prompt information into a second machine learning model to summarize key points; merging the web3 related information with the key points to generate second machine learning model prompt information; and generating risk insight information related to the digital asset based on the second machine learning model prompt information.

[0007] On the other hand, the present invention provides a digital asset risk information processing system. According to one embodiment, the digital asset risk information processing system of the present invention includes an information processing unit and a non-volatile information storage medium coupled to the information processing unit. Instructions are stored in the non-volatile information storage medium, and the instructions cause the information processing unit to perform the following steps: obtaining web2 text source data, web2 non-text source data, and web3 source data; converting the web2 non-text source data into web2 text processing data; filtering the web2 text source data and the web2 text processing data to obtain web2 related risk information related to the digital asset; obtaining web2 statistical information from the web2 text source data and the web2 text processing data; extracting web3 related risk information related to the digital asset from the web3 source data, and inputting the web2 statistical information and the web3 related risk information into a first machine learning model to obtain a risk score related to the digital asset.

[0008] According to another embodiment, the digital asset risk information processing system of the present invention includes an information processing unit and a non-volatile information storage medium coupled to the information processing unit. Instructions are stored in the non-volatile information storage medium, and the instructions cause the information processing unit to perform the following steps: obtaining web2 text source data, web2 non-text source data, and web3 source data; converting the web2 non-text source data into web2 text processing data; filtering the web2 text source data and the web2 text processing data to obtain web2 related risk information related to the digital asset; embedding the web2 related risk information into a prompt template to obtain a first prompt information; extracting web3 related risk information related to the digital asset from the web3 source data, inputting the first prompt information into a second machine learning model to summarize key points; merging the web3 related information with the key points to generate second machine learning model prompt information; and generating risk insight information related to the digital asset based on the second machine learning model prompt information.

[0009] In some embodiments of the method and / or system according to the present invention, the web2 related risk information includes at least one of the following: negative tweets, negative comments, and negative reports related to the digital asset.

[0010] In some embodiments of the method and / or system according to the present invention, filtering the web2 text source data and the web2 text processing data includes performing with at least one of the following filters: negative sentiment score, popularity index, authenticity index, and time range.

[0011] In some embodiments of the method and / or system according to the present invention, extracting web3 related risk information related to the digital asset from the web3 source data includes primary feature extraction and secondary feature extraction. The primary feature extraction includes directly generating web3 related risk information from the on-chain transaction records, and the secondary feature extraction includes detecting abnormal transaction information from the on-chain transaction records and according to the statistical information of the abnormal transaction information. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The following describes various embodiments of the present invention with reference to the drawings, wherein:

[0013] Figure 1 is a flowchart of a digital asset risk information processing method according to an embodiment of the present invention;

[0014] Figure 2 is a flowchart of a digital asset risk information processing method according to another embodiment of the present invention;

[0015] Figure 3It is a framework diagram of a digital asset risk information processing system according to an embodiment of the present invention;

[0016] Figure 4 It is a schematic flow diagram of a digital asset risk information processing system and a processing method according to an embodiment of the present invention. Detailed implementation manners

[0017] The present invention provides a method and a system for obtaining valuable digital asset risk information and risk indicators through artificial intelligence methods based on data and information obtained from web2 data sources and web3 data sources.

[0018] The technical solution provided by the present invention can automatically complete relevant data collection, collation, aggregation, risk indicator formulation, and risk insight presentation. According to the method and system of the present invention, structured data and unstructured data in real-time data from web2 data sources and web3 data sources are aggregated to provide an integrated risk data set. Based on these data sets, through the combination of machine learning technologies such as, for example, Large Language Model (LLM) or Graph Neural Network (GNN), more in-depth digital asset risk indicators are provided.

[0019] On the one hand, the technical solution provided by the present invention extracts risk indicators from various structured data through a dedicated machine learning model, obtains an overall risk score for digital asset risk identification, prediction, and control, and can also present the visualization information processing results in the form of charts and the like. On the other hand, the technical solution provided by the present invention summarizes the key points in the negative feedback in the online information through a large language model to understand the mainstream negative information about entities or assets. Subsequently, the technical solution provided by the present invention combines the key points in web2 data and the web3 statistical features extracted from web3 data sources, and uses a large language model to generate and provide digital asset risk insights, such as potential risks, comparable historical events, and risk mitigation action strategies. In some embodiments, the method and system according to the present invention provide at least one of the following solutions: (1) Notification and early warning function: When the obtained risk score exceeds a preset threshold, the method and system according to the present invention can provide a risk early warning notification / information to assist users in taking corresponding measures and actions to reduce financial losses; (2) Visualization function: The method and system according to the present invention can present digital risk information in the form of charts for easy access by users; (3) Understanding function: Users can view key negative arguments, potential risks, comparable events, and risk response and mitigation strategies to obtain actionable risk early warning and risk prevention and control insights.

[0020] The present invention provides a method and system for integrating the processing flows of web2 data and web3 data to comprehensively evaluate the digital asset risks of entities. The method and system according to the present invention also provide for converting multi-modal data (including but not limited to various modalities such as text, image, voice, video, audio, etc.) into a text format, and using natural language processing techniques to filter out feed information related to digital asset risks from large amounts of data with different structures. The method and system according to the present invention process heterogeneous data structures and complete information available online, thereby providing users with more valuable insights into digital asset risks. By adopting big data analysis techniques and real-time detection techniques, the method and system according to the present invention can automatically monitor information related to digital asset risks and detect potential risks. According to a specific machine learning model of the method and system according to the present invention, digital asset risk indicators with different update frequencies can be extracted from web2 data and web3 data sources, thereby improving the performance of the model in obtaining more comprehensive risk-related information. By combining prompt engineering and general large language models, complex web2 text data is converted into a data format that can be combined with web3 on-chain structured data in a relatively simple manner. The method and system according to the present invention can provide more comprehensive insights into digital asset risks.

[0021] According to one embodiment, as Figure 1 shown, the digital asset risk information processing method 100 of the present invention includes, in step 110, selecting a digital asset to be risk-evaluated. The digital asset can be a certain enterprise-level digital asset that a user is concerned about, or a certain digital asset itself, such as "the digital asset derivatives trading platform FTX", "Bored Ape Yacht Club", etc. In step 120, multi-mode / multi-modal web2 data related to the selected digital asset is extracted from web2 data sources. This extraction step 120 can be carried out by means such as web crawlers, application programming interfaces (APIs), or database direct connectors, based on information such as keywords and digital asset names, to obtain relevant information and data from web2 data sources.

[0022] Web2 data sources can include but are not limited to, for example, social media platforms such as Twitter, WeChat, Facebook, Reddit, etc., public websites such as www.bbc.com, www.coindesk.com etc., search engines such as Google, Baidu, Bing, etc., public chains such as Bitcoin, Ethereum, BSC, Solana, etc., and / or marketplaces such as OpenSea, Rarible, SuperRare, etc., and so on.

[0023] The extracted web2 data can be structured data, such as tabular data, or unstructured data, such as text, images, videos, audio, and other types of data. Unstructured data can be processed through artificial intelligence technologies such as Natural Language Processing (NLP) and combined with structured data for analysis.

[0024] In step 130, method 100 converts the extracted web2 non-text source data into web2 text processing data and filters the web2 source data and the web2 text processing data to obtain web2-related risk information regarding the selected digital asset. The web2-related risk information can include negative tweets, negative comments, and / or negative reports regarding the digital asset. This filtering step can be performed using at least one of the following filters: sentiment score negativity, popularity index, authenticity index, time range, etc.

[0025] In step 140, method 100 uses natural language processing techniques to perform feature engineering and obtain web2 statistical information from the web2 source data and the web2 text processing data, such as obtaining statistical information regarding negative tweets.

[0026] In step 150, method 100 extracts web3-related risk information regarding the selected digital asset from the web3 source data, such as on-chain transaction records. Specifically, information extraction step 150 includes a primary feature extraction step 152 and a secondary feature extraction step 154. For information that can be directly obtained from on-chain transaction records, such as information related to market volatility risk, performing the primary feature extraction step 152 can achieve the extraction of web3-related risk information for the digital asset. For more complex features, such as information related to wash trading (or virtual trading), information extraction step 150 further performs the secondary feature extraction step 154 to obtain information that cannot be directly obtained from on-chain transaction records, such as deep information like the transaction quantity statistics of wash trading.

[0027] In step 170, method 100 inputs the web2 statistical information and the web3-related risk information into a first machine learning model, such as a dedicated machine learning model, to obtain a risk score regarding the digital asset and the corresponding risk information features.

[0028] Additionally, method 100 may further include, at step 180, generating corresponding visualization information, such as visualization information of types like statistical charts, based on the risk score and the corresponding risk information features, and displaying it on the screen.

[0029] Method 100 may further include, at step 190, determining whether the obtained risk score exceeds a preset threshold. If so, method 100 includes, at step 192, generating and sending out a risk warning message. If not, method 100 may stop or repeat the previous operation steps.

[0030] According to another embodiment, as Figure 2 shown, the digital asset risk information processing method 200 of the present invention includes, at step 210, selecting a digital asset to be risk - evaluated. The digital asset may be a certain digital asset that a user is concerned about, such as "FTX, a digital asset derivatives trading platform", "Bored Ape Yacht Club", etc. At step 220, multi - mode web2 data related to the selected digital asset is extracted from a web2 data source. This extraction step 220 can be carried out by means such as web crawlers, application programming interfaces (APIs), or database direct connectors, based on information such as keywords and digital asset names, to obtain relevant information and data from the web2 data source.

[0031] Non - limiting web2 data sources may include, for example, social media platforms such as Twitter, WeChat, Facebook, Reddit, public websites such as www.bbc.com, www.coindesk.com etc., search engines such as Google, Baidu, Bing, public blockchains such as Bitcoin, Ethereum, BSC, Solana, and / or marketplace venues such as OpenSea, Rarible, SuperRare, and so on.

[0032] The extracted web2 data may be structured data, such as tabular data, or unstructured data, such as data of types like text, images, videos, audio, etc. The unstructured data can be processed by artificial intelligence technologies such as Natural Language Processing (NLP) and combined with the structured data for analysis.

[0033] In step 230, method 200 converts the extracted web2 non-text source data into web2 text processing data, and filters the web2 text source data and the web2 text processing data to obtain web2-related risk information regarding the selected digital asset. The web2-related risk information may include negative tweets, negative comments, and / or negative reports regarding the digital asset, etc. This filtering step may be performed using at least one of the following filters: sentiment score negativity, popularity index, authenticity index, time range, etc.

[0034] In step 240, method 200 embeds the web2-related risk information into a prompt template, such as a customized prompt template, to generate first-level prompt information for inputting into a second machine learning model in subsequent steps.

[0035] In step 250, method 200 extracts web3-related risk information regarding the selected digital asset from the web3 source data, such as on-chain transaction records. Specifically, the information extraction step 250 includes a first-level feature extraction step 252 and a second-level feature extraction step 254. For information that can be directly obtained from on-chain transaction records, for example, for information related to market volatility risk, performing the first-level feature extraction step 252 can achieve the extraction of web3-related risk information of the digital asset. For more complex features, such as information related to wash trading (or virtual trading), the information extraction step 250 further performs the second-level feature extraction step 254 to obtain information that cannot be directly obtained from on-chain transaction records, such as deep information like the transaction quantity statistics of wash trading.

[0036] In step 270, method 200 inputs the first-level prompt information generated in step 240 into a second machine learning model to summarize key points. The second machine learning model may be, for example, a general large language model.

[0037] In step 280, method 200 combines the key points obtained in step 270 with the web3-related information obtained in step 260 to generate second machine learning model prompt information.

[0038] In step 290, method 200 generates risk insight information regarding the digital asset based on the second machine learning model prompt information obtained in step 280. The risk insight information may include, for example, at least one of the following: potential risk information, comparable historical event information, and risk prevention strategy or risk mitigation action strategy information.

[0039] On the other hand, the present invention provides a digital asset risk information processing system for extracting information related to digital asset risks from information sources and generating risk indicators for digital assets. In some embodiments, the risk indicators of digital assets may include the selected risk scores of the digital assets to be evaluated, risk insights regarding the digital assets, such as potential risks, potential consequences after the occurrence of risks, as well as risk prevention strategies or risk mitigation action strategies.

[0040] According to one embodiment, the digital asset risk information processing system 300 of the present invention may include Figure 3 all or some of the modules shown. As Figure 3 shown, the system 300 includes a data storage module 310, a data processing module 320, a model module 330, and an application module 340.

[0041] The data storage module 310 may include an SQL database 312 and a non - SQL database 314. The SQL database 312 may include Oracle databases, MySQL databases, etc., for storing structured tabular data. The non - SQL database 314 may include MangoDB, Couchbase, etc., for storing unstructured data of types such as text data, picture data, video data, audio data, etc.

[0042] The data processing module 320 may include multiple sub - modules. In some embodiments, the data processing module 320 may include a prompt engineering sub - module 322, an indicator formulation sub - module 323, a data collection sub - module 324, a multi - mode transformation sub - module 325, and a natural language processing sub - module 326.

[0043] The data collection sub - module 324 can obtain data from a stream processing platform, such as Apache Kafka or Amazon Kinesis, etc., and can be managed using a preset data pre - processing workflow script and / or program. The script can clean and perform exploratory data analysis, and conduct quality checks on each data type. Subsequently, the pre - processed data can be processed through SQL streams / non - SQL streams and stored in the database. For each stream processing platform, separate corresponding pre - processing scripts and database storages can be provided for each data type (text, image, audio - video content, etc.).

[0044] The multi - mode conversion sub - module 325 can convert each data type in the multimedia data into text format. For example, an audio file can be converted into plain text format for further processing.

[0045] The natural language processing sub-module 326 is used to analyze text data and formulate indexes to filter out the information feeds that are most relevant to risks. The natural language processing methods adopted may include methods such as sentiment analysis, topic models, named entity recognition, tokenization, stemming, lemmatization, bag-of-words, etc. Formulating indexes may include scoring negative emotions, popularity indexes, authenticity indexes, etc.

[0046] The prompt engineering sub-module 322 can embed the filtered information feeds into a preset prompt template to generate first-level prompt information for input into the large language model. In one example, the first-level prompt information can be "Provide the key meaning in the following tweet: (specific content of the tweet)".

[0047] The metric formulation sub-module 323 can be a machine learning model for generating and / or calculating statistical features. For web2 data, features (such as statistical features) are mainly obtained by the natural language processing sub-module 326 through additional processing, such as through aggregation processing. For web3 data, steps including first-level feature extraction and second-level feature extraction can be executed. For example, after detecting risky activities in on-chain transactions by performing first-level feature extraction using a machine learning model such as a graph neural network, the final feature information can be obtained by performing second-level feature extraction.

[0048] The model module 330 may include two sub-modules, namely the general large language model sub-module 332 and the dedicated machine learning model sub-module 334. In some embodiments, the general large language model sub-module 332 may receive the prompt information from the prompt engineering sub-module 322, i.e., the web2 key points, as well as the web3 statistical features, and generate risk insights. In some instances, the risk insights may include potential risks, potential outcomes (by referring to comparable historical events), and / or risk mitigation action strategies. In some embodiments, the web2 key points and web3 statistical features may be stored in a general database and used as inputs to the general large language model sub-module 332. In some embodiments, risk insights may be generated by leveraging commercial APIs (such as OpenAI GPT4) or running a self-built large language model locally. In some embodiments, the dedicated machine learning model sub-module 334 may be customized according to the required use cases to evaluate the overall risk level, e.g., presented by a calculated risk score. In some embodiments, the inputs to the dedicated machine learning model sub-module 334 may be sourced from web2 statistical features and web3 statistical features. In some embodiments, the output of the dedicated machine learning model sub-module 334 may be a risk score and corresponding key features.

[0049] The application module 340 may include multiple sub-modules. In some embodiments, the application module 340 includes a notification sub-module 342, a visualization sub-module 344, and an understanding sub-module 346. The notification sub-module 342 may monitor the risk score of the dedicated machine learning model and issue a risk warning message when the risk score is greater than a preset threshold. In some embodiments, the visualization sub-module 344 may generate corresponding visual risk diagrams, such as charts and tables, based on the statistical functions / outputs of the dedicated machine learning model sub-module 334. In some instances, the visualization sub-module 104 may generate relevant risk charts and / or risk tables based on the risk score and / or corresponding key features. In some embodiments, the understanding sub-module 346 may present the output of the general large language model sub-module 332, such as risk insights, in the form of natural language.

[0050] Figure 4 Shown corresponding to the foregoing reference Figure 1 and Figure 2 shown embodiments of, based on Figure 3 the system shown performs the digital asset risk information processing method of the process 400. As Figure 4 shown, corresponding to Figure 1 and Figure 2Steps 120 and 220 of the method shown. Process 400, through data collection sub-module 324, uses methods such as web crawlers, application programming interfaces, or database direct connectors, etc., based on keywords (e.g., the selected data asset names to be risk-assessed), to collect relevant data from a variety of different data sources, and continuously tracks information and status updates. The data sources can include web2 data sources and web3 data sources. The data sources may include but are not limited to the following web2 data sources: social media platforms 410, such as Twitter, WeChat, Facebook, Reddit, etc.; search engines 420, such as Google, Baidu, Bing, etc.; public websites 430, such as www.bbc.com, www.coindesk.com, etc. The data sources may also include but are not limited to the following web3 data sources: public blockchains 440, such as Bitcoin, Ethereum, BSC, Solana, etc.; trading markets 450, such as OpenSea, Rarible, SuperRare, etc.; and other data sources 460. The data can be multi-modal data. The data can include structured data, such as tabular data, etc., and also include unstructured data, such as text, pictures, videos, audio, etc. Machine learning techniques such as natural language processing sub-module 325, image processing, or video analysis are used to analyze the unstructured data and combine the unstructured data with the structured data.

[0051] Corresponding to Figure 1 and Figure 2 Steps 130 and 230 of the method shown. Process 400 uses the multi-mode transformation sub-module 325 to convert multi-modal data, such as pictures, videos, audio, etc., into text data, and then filters a large amount of web2 data to obtain web2-related risk information related to the selected digital assets. The filters can include a variety of pre-set filters, including but not limited to, sentiment score negativity, popularity index, authenticity index, time range, etc. Corresponding to Figure 1 and Figure 2 Steps 140 and 240 of the method shown. Process 400 uses natural language processing techniques to perform feature engineering and formulate statistical features (e.g., the number of negative tweets, etc.) based on the text data. These statistical features can be used in subsequent machine learning models.

[0052] Corresponding to Figure 1 and Figure 2For steps 150 and 250 of the method shown, process 400, through the metric formulation sub-module 323, uses a feature engineering module (such as formulating feature engineering) and a deep neural network method to perform first-level feature extraction and second-level feature extraction on web3 data and obtain web3 statistical features. In addition, the steps for web2 data and the steps for web3 data can be executed in parallel, serially, and / or independently.

[0053] Corresponding to Figure 1 For step 170 of the method shown, process 400 inputs the extracted statistical features into the dedicated machine learning model 334. The model outputs a risk score and the corresponding risk information features. Then, corresponding to Figure 1 For step 180 of the method shown, process 400, through the visualization sub-module 344, generates risk-related charts and tables according to a predefined template and presents them on the display screen. If the risk score is greater than a preset threshold, the notification sub-module 342 can issue a risk notification and / or a risk warning.

[0054] Independently of or in addition to the foregoing process, corresponding to Figure 2 For step 270 of the method shown, process 400 inputs the first-level prompt information generated in step 240 into the general large language model 332 to summarize the key points and outputs the web2 key points. Then, the web2 key points are merged with the identified risk data of the web3 statistical features (for example, the number of wash trading activities of the project). In some embodiments, the merged risk data can be stored in a general database, and then, using large language model prompts, risk insights in the form of text, charts, and other suitable representations can be generated. The risk insights can include one or more of potential risks, historically similar events, mitigation strategies, etc.

[0055] The present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or limiting. The example embodiments are selected and described to explain the technical solutions and practical applications of the present invention and enable those of ordinary skill in the art to understand the various embodiments of the present invention, which may include various modifications suitable for the intended specific purposes.

[0056] Some functional units in this document are exemplified by modules as the execution units of the corresponding technical features to describe the relevant technical solutions and technical features. Those skilled in the art will understand that a module can be implemented as a circuit, a logic chip, or any kind of discrete component. A module should be understood as not being limited to a physical form or a hardware form. A module can also be implemented in a software form and / or a functional form, and this software or function can be executed or implemented by different types of processor architectures. In the embodiments of the present invention, a module can also include computer instructions or executable code, and can instruct a computer processor to perform a series of operations according to the received instructions. Those skilled in the art can select specific implementation modules according to the technical solutions provided by the present invention. The execution units and modules described in the present invention are exemplary rather than restrictive or exhaustive. Therefore, the protection scope defined by the claims of the present invention should not be understood as being limited only to the technical features and technical solutions presented in the embodiments.

Claims

1. A method for processing digital asset risk information, characterized in that: The method comprises: Obtaining web2 multimodal source data and web3 source data, wherein the web2 multimodal source data includes web2 non-text source data and web2 text source data; Converting the web2 non-text source data into web2 text processed data; Filtering the web2 text source data and the web2 text processed data to obtain web2 related risk information involving the digital asset; Obtaining web2 statistical information from the web2 text source data and the web2 text processed data; Extracting web3-related risk information involving the digital asset from the web3 source data; The web2 statistical information and the web3 related risk information are input into a first machine learning model to obtain a risk score involving the digital asset and a corresponding risk information feature.

2. A method for processing digital asset risk information, characterized in that: The method comprises: Obtaining web2 multimodal source data and web3 source data, wherein the web2 multimodal source data includes web2 non-text source data and web2 text source data; Converting the web2 non-text source data into web2 text processed data; Filtering the web2 text source data and the web2 text processed data to obtain web2 related risk information involving the digital asset; Embedding the web2-related risk information into a prompt template to generate first-level prompt information; Extracting web3-related risk information involving the digital asset from the web3 source data; Inputting the primary prompt information into a second machine learning model to summarize key points; Merging the web3 related information with the key point to generate second machine learning model prompt information; Based on the second machine learning model prompt information, risk insight information related to the digital assets is generated.

3. The method according to claim 1 or 2, characterized in that: The web2-related risk information includes at least one of the following: negative tweets, negative comments, and negative reports involving the digital asset.

4. The method according to claim 1 or 2, characterized in that: The filtering of the web2 text source data and the web2 text processed data includes using at least one of the following filters: sentiment score negativity, popularity index, authenticity index, and time range.

5. The method according to claim 1 or 2, characterized in that: The web3-related risk information includes on-chain transaction records.

6. The method according to claim 5, characterized in that The extraction of web3-related risk information involving the digital assets from the web3 source data includes primary feature extraction and secondary feature extraction. The primary feature extraction includes directly generating web3-related risk information from the on-chain transaction records. The secondary feature extraction includes detecting abnormal transaction information from the on-chain transaction records, as well as statistical information based on the abnormal transaction information.

7. The method according to claim 1, characterized in that The method further includes, after obtaining the risk score related to the digital asset, converting the risk score into a chart, and displaying the chart.

8. The method according to claim 1 or 7, characterized in that: If the risk score exceeds a preset threshold, the method further includes generating and issuing risk warning information.

9. A digital asset risk information processing system, characterized in that: The system comprises: information processing unit; and A non-volatile information storage medium coupled to the information processing unit, wherein the non-volatile information storage medium stores instructions, wherein the instructions cause the information processing unit to perform the following steps: Obtaining web2 multimodal source data and web3 source data, wherein the web2 multimodal source data includes web2 non-text source data and web2 text source data; Converting the web2 non-text source data into web2 text processed data; Filtering the web2 text source data and the web2 text processed data to obtain web2 related risk information involving the digital asset; Obtaining web2 statistical information from the web2 text source data and the web2 text processed data; Extract web3 related risk information involving the digital asset from the web3 source data The web2 statistical information and the web3 related risk information are input into a first machine learning model to obtain a risk score involving the digital asset.

10. A digital asset risk information processing system, characterized in that: The system comprises: information processing unit; and A non-volatile information storage medium coupled to the information processing unit, wherein the non-volatile information storage medium stores instructions, wherein the instructions cause the information processing unit to perform the following steps: Obtaining web2 multimodal source data and web3 source data, wherein the web2 multimodal source data includes web2 non-text source data and web2 text source data; Converting the web2 non-text source data into web2 text processed data; Filtering the web2 text source data and the web2 text processed data to obtain web2 related risk information involving the digital asset; Embedding the web2-related risk information into a prompt template to obtain first prompt information; Extracting web3 related risk information involving the digital asset from the web3 source data, Inputting the first prompt information into a second machine learning model to summarize key points; Merging the web3 related information with the key point to generate second machine learning model prompt information; Based on the second machine learning model prompt information, risk insight information related to the digital assets is generated.

Images