Method, device, equipment, medium and program product for determining news impact

Abstract

The present application discloses a method, apparatus, device, medium and program product for determining the degree of news impact; wherein, the method comprises: obtaining news data to be processed; determining a target subject set associated with the news data; wherein, the target subject set comprises a target company subject, a target product subject and a target industry subject; determining public opinion impact parameters of the news data; and determining the degree of impact of the news data on subjects in the target subject set based on the public opinion impact parameters.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence and natural language processing technology, and in particular to a method, device, equipment, medium and program product for determining the degree of news impact. Background Art

[0002] In practical applications, the impact of news reports on the subject of the report is usually evaluated based on the positive or negative nature of the public opinion expressed in the news reports. However, the impact determined by the above method is not comprehensive enough to meet the actual analysis needs of news reports. Summary of the invention

[0003] Based on the above technical problems, the embodiments of the present application provide a method, device, equipment, medium and program product for determining the degree of news impact.

[0004] The technical solution provided by the embodiment of the present application is as follows:

[0005] The present application embodiment first provides a method for determining the impact degree of news, the method comprising:

[0006] Get the news data to be processed;

[0007] Determine a target subject set associated with the news data; wherein the target subject set includes a target company subject, a target product subject, and a target industry subject;

[0008] Determining public opinion impact parameters of the news data;

[0009] The influence degree of the news data on the subjects in the target subject set is determined based on the public opinion influence parameter.

[0010] In some embodiments, determining the public opinion impact parameter of the news data includes:

[0011] Performing sentiment analysis on the news data by using a first artificial intelligence model to determine the sentiment category of the news data;

[0012] Performing event analysis on the news data by using a second artificial intelligence model to determine the event type associated with the news data;

[0013] Determining event association strength based on the news data and the event type;

[0014] The public opinion impact parameter is determined based on the emotion category and the event association strength.

[0015] In some embodiments, determining the public opinion impact parameter based on the emotion category and the event association strength includes:

[0016] Normalizing the emotion categories to obtain emotion scores;

[0017] Normalizing the event association strength to obtain a strength score;

[0018] Determine the association weight;

[0019] The intensity score and the sentiment score are processed based on the association weight to obtain the public opinion impact parameter.

[0020] In some embodiments, determining the influence degree of the news data on the subjects in the target subject set based on the public opinion influence parameter includes:

[0021] Obtain industry revenue data of enterprises and / or organizations associated with the target industry entity;

[0022] Based on the industry revenue data and the public opinion impact parameter, determine the impact degree of the news data on the target industry entity.

[0023] In some embodiments, determining the influence degree of the news data on the subjects in the target subject set based on the public opinion influence parameter includes:

[0024] Determining the influence degree of the news data on the target company entity as the public opinion influence parameter;

[0025] The method further comprises:

[0026] Determining the company revenue ratio of the first affiliated enterprise of the target company entity; wherein the first affiliated enterprise includes a supply company or a purchasing company associated with the target company entity;

[0027] Based on the company's revenue ratio and the public opinion impact parameter, the impact degree of the news data on the first related enterprise is determined.

[0028] In some embodiments, determining the influence degree of the news data on the subjects in the target subject set based on the public opinion influence parameter includes:

[0029] Determine first proportion data of the target product entity in the products of the target company entity;

[0030] Based on the first proportion data and the public opinion influence parameter, the influence degree of the news data on the target company entity in the target product entity dimension is determined.

[0031] In some embodiments, the method further comprises:

[0032] Determine a second associated enterprise associated with the target product entity;

[0033] Determine second ratio data of the second associated enterprise being associated with the target product entity;

[0034] Based on the second proportion data and the public opinion influence parameter, the influence degree of the news data on the second related enterprise in the target product main dimension is determined.

[0035] In some embodiments, determining the target subject set associated with the news data includes:

[0036] Constructing a subject knowledge graph; wherein the subject knowledge graph includes the interrelated relationships between various types of company subjects, product subjects, and industry subjects;

[0037] Determine a subject type set associated with the news data; wherein the subject type set includes company type, product type and industry type;

[0038] Based on the matching distance between the subject name associated with the news data and the subject in the subject knowledge graph, the target subject set is determined from the subject knowledge graph; wherein the target subject set is associated with the subject type set.

[0039] The present application also provides a device for determining the degree of news impact, the device comprising:

[0040] An acquisition module is used to acquire news data to be processed;

[0041] A determination module, used to determine a target subject set associated with the news data; wherein the target subject set includes a target company subject, a target product subject, and a target industry subject;

[0042] The determination module is further used to determine a public opinion impact parameter of the news data; and determine the degree of impact of the news data on subjects in the target subject set based on the public opinion impact parameter.

[0043] An embodiment of the present application also provides a device for determining the degree of impact of news, the device comprising a processor and a memory; wherein a computer program is stored in the memory; when the computer program is executed by the processor, it can implement any of the methods for determining the degree of impact of news described above.

[0044] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored; when the computer program is executed by a processor of an electronic device, it can implement any of the methods for determining the degree of news impact as described above.

[0045] An embodiment of the present application also provides a computer program product, which includes a computer program; when the computer program is executed by a processor of an electronic device, it can implement any of the above methods for determining the degree of news impact.

[0046] In the method for determining the degree of news influence provided in the embodiment of the present application, after obtaining the news data to be processed, the target subject set associated with the news data is determined, and the target subject set includes the target company subject, the target product subject and the target industry subject. In this way, through the above method, a comprehensive analysis of the subjects associated with the news data is achieved; and by determining the public opinion influence parameters of the news data, the concrete and data-based representation of the influence of the news subjects in the public opinion dimension can be achieved; on this basis, the degree of influence of the news data on the subjects in the target subject set is determined based on the public opinion influence parameters, which not only achieves the accuracy of the degree of influence of the news data on the subjects, but also improves the comprehensiveness of the degree of influence of the news data on the subjects in the target subject set, thereby being able to meet the analysis needs of the degree of influence of news data in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A flowchart of a method for determining the degree of news impact provided in an embodiment of the present application;

[0048] Figure 2A This is a schematic diagram of the structure of the Hybrid Attention Networks (HAN);

[0049] Figure 2B is another structural schematic diagram of HAN;

[0050] Figure 3 A schematic diagram of the calculation process of the public opinion influence parameter provided in the embodiment of the present application;

[0051] Figure 4A A schematic diagram of the structure of the first knowledge graph provided in an embodiment of the present application;

[0052] Figure 4B A schematic diagram of the structure of the second knowledge graph provided in an embodiment of the present application;

[0053] Figure 4C A schematic diagram of the structure of the third knowledge graph provided in the embodiment of the present application;

[0054] Figure 4D A schematic diagram of the principle of industry matching provided in an embodiment of the present application;

[0055] Figure 5A A schematic diagram of a process for determining the impact of news provided in an embodiment of the present application;

[0056] Figure 5B A schematic diagram of a structure for determining the news impact level provided in an embodiment of the present application;

[0057] Figure 5C Another structural diagram of determining the news impact level provided in an embodiment of the present application;

[0058] Figure 6 A schematic diagram of the structure of a device for determining the degree of news impact provided in an embodiment of the present application;

[0059] Figure 7 A schematic diagram of the structure of a device for determining the degree of news impact provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0061] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] For financial investment fields including securities trading, public opinion and public sentiment, as non-traditional financial factors, have a significant and difficult-to-quantify impact on the price and market trend of securities trading. A positive public opinion atmosphere can inspire market confidence and drive up the price of securities trading, while a negative public opinion environment may cause market panic and lead to a drop in the price of securities trading. Especially when it comes to hot news and emergencies involving important listed companies or industries, the public's public opinion reaction and emotional tendency have become key factors in predicting the subsequent trend of securities trading.

[0063] When evaluating the impact of market sentiment on the financial industry, including individual stocks, related technologies are usually based on the positive or negative nature of news reports, which makes it difficult for the solution to comprehensively and accurately measure the comprehensiveness of the impact of different news reports. As a result, it is impossible to provide a more detailed and refined analysis perspective on the breadth and depth of the impact of news reports, and it is impossible to accurately evaluate the possible consequences of news reports, thereby reducing the accuracy of investment decisions. At the same time, news reports are important data that can guide social public opinion and influence public sentiment. Therefore, a comprehensive and detailed analysis of news reports is an important part of predicting their impact on economic development, including the financial investment field.

[0064] On the other hand, the analysis scheme of the impact of news reports by related technologies only focuses on a single news subject and fails to examine the risk of news reports in a broader supply chain context. This isolated analysis method makes it difficult to reveal the transmission and diffusion status of the possible impact of news reports between upstream and downstream enterprises in the supply chain, and thus it is impossible to accurately assess its potential threat to the entire supply chain system; moreover, the current analysis method of news reports emphasizes the direct impact of a single event on a single subject, while ignoring the indirect impact of news reports in the industrial chain and supply chain. Public opinion events often have the characteristics of chain reactions, and one event may trigger a chain of turmoil in the entire supply chain system. Therefore, the above analysis method cannot fully capture the systematic impact of public opinion events on the entire market environment. Because the above analysis method fails to fully consider the transmission effect of public opinion risks in the supply chain, it is difficult to accurately judge the spread and potential impact of public opinion risks, so that effective countermeasures cannot be taken in a timely manner, and it may also lead to blind spots in the prediction of complex risks.

[0065] Therefore, the analysis of the impact of news reports in related technologies is not comprehensive enough and cannot meet the actual analysis needs of news reports.

[0066] Based on the above technical problems, the embodiments of the present application provide a method, device, equipment, medium and program product for determining the degree of news impact.

[0067] Figure 1 A flowchart of a method for determining the impact of news provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method may include the following steps:

[0068] Step 101: Obtain news data to be processed.

[0069] In one implementation, the news data may include news reports; illustratively, the news reports may be in at least one form of text, pictures, and videos.

[0070] Step 102: Determine a target subject set associated with the news data.

[0071] Among them, the target entity set includes target company entities, target product entities and target industry entities.

[0072] In one embodiment, the subject may include the core content around or associated with the news data; illustratively, in this solution, the news data may only report one subject.

[0073] In one embodiment, the target company entity may include at least one company directly or indirectly associated with the news data; illustratively, the target company entity may have at least one function including manufacturing, construction, logistics and transportation, education and medical care, and financial investment.

[0074] In one embodiment, the target product subject may include at least one or at least one product directly or indirectly associated with the news data; illustratively, the target product subject may include economic products, wherein the economic products may include securities including stocks, etc.

[0075] In one implementation, the target industry entity may include at least one industry directly or indirectly related to the news data; illustratively, the target industry entity may include industries such as food manufacturing, chemical production, and finance.

[0076] In one implementation, the target subject set may be determined by:

[0077] The news data is analyzed through a large language model (LLM) to extract the industries, companies and products associated with the news data, and the above industries, companies and products are respectively identified as target industry entities, target company entities and target product entities.

[0078] Through the artificial intelligence (AI) model, the news data is deeply analyzed to determine the target industry entities, target company entities and target product entities associated with the news data.

[0079] Step 103: Determine the public opinion impact parameter of the news data.

[0080] In one embodiment, the public opinion influence parameter may include a public opinion factor corresponding to the news data; illustratively, the public opinion factor may characterize the influence and / or degree of influence of the news data in the formation and evolution of public opinion, wherein the above-mentioned influence and / or degree of influence may be reflected in the form of at least two parameters or factors.

[0081] In one implementation, the public opinion influence parameter may have a type feature; illustratively, the public opinion influence parameter may include five types: strongly positive, strongly negative, neutral, weakly positive, and weakly negative.

[0082] In one implementation, the public opinion impact parameter may be determined in the following manner:

[0083] The news background corresponding to the news data is obtained, and the news data and the news background are analyzed through LLM to determine the degree parameter of the news data's influence on at least one public opinion under the economic environment represented by the news background, and then the degree parameter of at least one public opinion influence is statistically analyzed to obtain the public opinion influence parameter.

[0084] Step 104: Determine the degree of influence of the news data on the subjects in the target subject set based on the public opinion influence parameter.

[0085] In one embodiment, the degree of influence may include the type of negative influence, positive influence or neutral influence, and may also include the degree of the above various influence types.

[0086] In one implementation, the above-mentioned impact degree can be determined by:

[0087] Determine the subject weight factors associated with the subjects in the target subject set respectively, and then weight the public opinion impact parameters based on the company weight factors corresponding to the target company subject to obtain the degree of influence of the news data on the target company subject; weight the public opinion impact parameters based on the product weight factors corresponding to the target product subject to obtain the degree of influence of the news data on the target product subject; weight the public opinion impact parameters based on the industry weight factors corresponding to the target industry subject to obtain the degree of influence of the news data on the target industry subject; among them, the company weight factor, product weight factor and industry weight factor can respectively represent the component factors of the public opinion impact parameters on the target company subject, the target product subject and the target industry subject.

[0088] From the above, it can be seen that in the method for determining the degree of news influence provided in the embodiment of the present application, after obtaining the news data to be processed, the target subject set associated with the news data is determined, and the target subject set includes the target company subject, the target product subject and the target industry subject. In this way, through the above method, a comprehensive analysis of the subjects associated with the news data is achieved; and by determining the public opinion influence parameters of the news data, the concrete and data-based representation of the influence of the news subject in the public opinion dimension can be achieved; on this basis, the degree of influence of the news data on the subjects in the target subject set is determined based on the public opinion influence parameters, which not only achieves the accuracy of the degree of influence of the news data on the subjects, but also improves the comprehensiveness of the degree of influence of the news data on the subjects in the target subject set, thereby being able to meet the analysis needs of the degree of influence of news data in practical applications.

[0089] Based on the above embodiments, in the method for determining the news impact degree provided in the embodiments of the present application, determining the public opinion impact parameter of the news data can be achieved by the following steps:

[0090] Step A1: Perform sentiment analysis on news data through the first AI model to determine the sentiment category of the news data; perform event analysis on the news data through the second AI model to determine the event type associated with the news data.

[0091] In one embodiment, the first AI model and the second AI model may be the same or different.

[0092] In practical applications, the positive or negative emotions conveyed by news will have different degrees of impact on related industries, companies and products, among which the impact of negative news is particularly significant. In addition, the wording and expression of news can also reflect the severity of the event. Usually, news with fierce words or exaggerated expressions are considered to have a higher impact intensity. Secondly, relying solely on news text to judge the impact intensity may lead to deviations, because although the text description of some events is not fierce enough, their actual impact may be very serious. Therefore, it is crucial to perform sentiment analysis on news data.

[0093] In one embodiment, the first AI model can receive news data, identify and process the news data, and output sentiment categories including positive, negative and neutral; exemplarily, the first AI model can also quantify the sentiment categories to obtain sentiment scores, namely sentiment logit scores; exemplarily, the first AI model can analyze the news data with the help of the first prompt to obtain sentiment categories or sentiment logit scores, wherein the first prompt can be: predict sentiment categories ['positive', 'negative', 'neutral'] for the news data, and predict the logit scores of each category, and the data output by the first AI model can be in the form of: {'positive': 'first score', 'negative': 'second score', 'neutral': 'third score'}.

[0094] In one embodiment, under the condition that news data is associated with the financial field, event types may include bankruptcy liquidation, major asset losses, being interviewed, company listing, major safety accidents, equity pledge, corporate financing, corporate acquisition, and pledge release.

[0095] In one embodiment, the second AI model can analyze the news data based on the second prompt to obtain the event type associated with the news data; illustratively, the second prompt can be: predicting the event type of news data in the financial field, limiting the event type result to a specified word range, and outputting it in the form of {'event type': 'type identifier'}.

[0096] Step A2: determining event correlation strength based on news data and event type.

[0097] In one embodiment, the event correlation strength may represent the degree of influence of news data on various events represented by the event type; illustratively, the event correlation strength may include two types: strong correlation and weak correlation.

[0098] In one embodiment, the event correlation strength can be obtained by processing news data and event types through a third AI model; illustratively, the third AI model can be the same as or different from the first AI model and the second AI model.

[0099] In one embodiment, the third AI model can process news data and event types based on the third prompt; wherein the third prompt can be: predicting the impact intensity of news data on events represented by event types, and predicting the event logits scores corresponding to each event category, and the data output format is: {'strong correlation': 'fourth logit score', 'weak correlation': 'fifth logit score'}.

[0100] It should be noted that the first AI model to the third AI model may be of the same type.

[0101] Exemplarily, the first AI model to the third AI model may all have natural language processing functions, that is, they may all be natural language processing (NLP) models; wherein, NLP is a key branch in the field of artificial intelligence, which aims to enable computer devices to understand, analyze and generate human language; NLP can cover the processing and implementation of multiple tasks from basic text processing to complex semantic understanding.

[0102] Among them, sentiment analysis is an important branch of NLP, which focuses on automatically interpreting and classifying sentiment tendencies from text data. Sentiment tendencies can include positive, negative or neutral tendencies, as well as more fine-grained emotions such as pleasure, anger and sadness.

[0103] Sentiment analysis is particularly important in social media analysis, through which we can understand the public's emotional attitudes towards brands, products, and events. In addition, event classification is another important task of NLP, which aims to identify and classify specific types of events in text. Event classification plays a key role in scenarios such as news summarization and public opinion monitoring. It can extract key events and generate concise news summaries. At the same time, by analyzing news events reported on social media and news websites, it can timely obtain the public's attention and attitude towards specific topics.

[0104] Exemplarily, the first to third AI models can all be generative LLMs, wherein generative LLMs can capture the complex laws and patterns of human language from massive text data, thereby generating coherent and meaningful text. The first to third AI models of this solution can be trained based on massive sample data, so that the first AI model and the third AI model have cross-domain knowledge and language understanding capabilities, and can also cover multiple fields and scenarios.

[0105] Exemplarily, the first to third AI models can all be Chain of Thought (CoT), where CoT is a reasoning technology used in large model prompt word engineering. This technology can simulate the human thinking process, reveal the essence of news data layer by layer through step-by-step logical reasoning and analysis, and thus draw reliable conclusions.

[0106] In the embodiment of the present application, CoT technology can be used to determine the impact of news data on the subjects in the target subject set and its positive and negative impact. Among them, CoT can first decompose complex news data into multiple logical steps to form small problems that are easy to handle and understand; secondly, CoT can use known information and logical rules to reason and analyze each small problem and draw preliminary conclusions; then, CoT can combine the conclusions of all small problems to form a comprehensive understanding of the original problem; finally, CoT can continuously correct and optimize the reasoning process based on new information until the final conclusion is reached.

[0107] Step A3: Determine public opinion impact parameters based on sentiment categories and event correlation strength.

[0108] In one implementation, the public opinion impact parameter may be determined in the following manner:

[0109] The sentiment categories and event correlation strengths are statistically averaged, and the results of the statistical average are determined as public opinion impact parameters.

[0110] In related technologies, when conducting public opinion analysis on news reports, it is usually implemented through neural network structures including HAN. Among them, HAN converts news text into vector representation through word embedding, and uses a multi-level attention mechanism to integrate the views and comments of different analysts, and uses the attention mechanism to imitate the sequential content dependence and diversified influence of humans when capturing public opinion influence from news reports; at the same time, when HAN is applied to the field of securities trading, a TopK-Dropout strategy can be constructed so that its attention mechanism can focus on key public opinions in the field of securities trading.

[0111] In order to simulate the influence of diversity, HAN introduced the attention mechanism at the word and news level in the first half of the network structure, and assigned corresponding weights to different words and news. HAN automatically learned the weight distribution and extracted information from news texts more effectively according to the effectiveness of different news. In order to form sequential content dependency, HAN used a bidirectional recurrent neural network (BiGRU) in the second half of the network structure to process the time series data in the news data, and further assigned different weights to the news of each date through the attention mechanism at the time level, so as to achieve the integration of all news data and finally output the prediction of stock trends.

[0112] Figure 2A is a schematic diagram of the structure of HAN, Figure 2A As shown, HAN includes a word embedding layer (Word Embedding Layer), a temporal attention layer (Temporal Attention Layer), a news attention (News-level Attention Layer) and a bidirectional gated recurrent unit (BiGRU Layer), and may also include a multilayer perceptron (Multilayer Perceptron, MLP) corresponding to the Discriminative Network.

[0113] In actual applications, different words usually have different importance in the process of news reporting. Therefore, the relevant technology also provides Figure 2B Another structural diagram of HAN shown in FIG. Figure 2A The structure shown, Figure 2B A word attention layer (Attention Layer) at the word level corresponding to the WordEmbedding Layer is added.

[0114] HAN can be Figure 2B The three attention layers in the neural network are used to simulate the process of humans capturing the degree of public opinion influence from news reports. Among them, word attention refers to the focus of humans on certain key words and sentences when browsing text, and abstracting important information from them to form an understanding of the text; news attention refers to the different attention given to different news by humans when reading multiple news due to the differences in the information contained in the news; and time series attention refers to people allocating attention to different dates based on the daily differences in the importance and timeliness of the news. The three attention groups are all reflected in the form of neural network weights, and ultimately those news that have a greater impact on stock prices are given a higher weight coefficient, thereby determining the impact of public opinion news on individual stocks.

[0115] However, the sample data used for HAN training is small in size, and the sample data is determined through manual annotation. This results in a narrow coverage of sample data, which cannot cover the diverse news reports and their related subject ranges in a complex and changing market environment, thus limiting the generalization ability of HAN. On the other hand, the manual annotation of sample data can easily introduce the analyst's subjective judgment, thus affecting the accuracy of data processing during HAN training. Especially when HAN processes controversial news data, the above subjective judgment factors will directly lead to deviations in processing results. At the same time, based on the above two points, HAN is difficult to integrate and process news reports from different channels and in different formats, which leads to an increase in the one-sidedness and limitations of public opinion analysis results, and thus cannot fully reflect the true face of market public opinion.

[0116] In the embodiments of the present application, by analyzing news data through the first AI model to the third AI model, the negative effects of insufficient samples and insufficient accuracy of HAN can be overcome.

[0117] From the above, it can be seen that in the method for determining the degree of news impact provided in the embodiment of the present application, sentiment analysis is performed on news data through the first AI model to determine the sentiment category of the news data, and event analysis is performed on the news data through the second AI model to determine the event type associated with the news data. In this way, through the above operations, a multi-faceted and fine-grained analysis of the news data is achieved; on this basis, the event correlation strength is determined based on the news data and the event type. In this way, the strength of the correlation between the news data and the events represented by the event type is reflected through the event correlation strength; on this basis, based on the sentiment category and the event correlation strength, the public opinion impact parameters are determined, thereby improving the accuracy and comprehensiveness of the public opinion impact parameters.

[0118] Based on the above embodiments, in the method for determining the news impact degree provided in the embodiments of the present application, determining the public opinion impact parameter based on the emotion category and the event association strength can be achieved by the following steps:

[0119] Step B1: normalize the emotion categories to obtain emotion scores; normalize the event association strength to obtain strength scores.

[0120] In one implementation, the sentiment score may be obtained by any of the following methods:

[0121] The inverse of the sentiment logit score corresponding to the sentiment category is determined as the sentiment score.

[0122] The sentiment logit score is processed by softmax to obtain the sentiment score.

[0123] Exemplarily, the process of obtaining the intensity score may be the same as the process of obtaining the emotion score, which will not be described in detail here.

[0124] Step B2: Determine the association weight.

[0125] In one implementation, the association weight may include two weight values, corresponding to the sentiment score and the intensity score respectively.

[0126] In one implementation, the association weight may be predetermined or may be adjusted; illustratively, the size of the weight value in the association weight may represent the importance of the score corresponding to the weight value.

[0127] Step B3: Process the intensity score and sentiment score based on the association weight to obtain public opinion impact parameters.

[0128] Figure 3 The following is a schematic diagram of the calculation process of the public opinion influence parameter provided in the embodiment of the present application. Figure 3 As shown, the sentiment logit scores may include senti_pos, senti_neu and senti_neg, corresponding to positive, neutral and negative sentiment types respectively; the sentiment logit scores are normalized by softmax to obtain senti_pos*, senti_neu* and senti_neg*.

[0129] Exemplarily, the event association strength may include infl_str and infl_weak, which are used to represent strong association and weak association, respectively. The event association strength is normalized by softmax to obtain infl_str* and infl_weak*.

[0130] For example, senti_pos*, senti_neu*, senti_neg*, infl_str*, and infl_weak* can be processed through the dual-index fusion process shown in formula (1) to obtain the public opinion impact parameter score(x, y).

[0131]

[0132] Among them, β and 1 can be two weight values ​​in the associated weight, score(x) can include senti_pos*, senti_neu* and senti_neg*, and score(y) can include infl_str* and infl_weak*.

[0133] For example, according to the size of the public opinion impact parameter, the news data can be divided into seven categories: strongly positive, weakly positive, neutral, strongly negative, and weakly negative, so as to intuitively show the nature and influence of the News Bureau.

[0134] From the above, it can be seen that in the method for determining the degree of news impact provided in the embodiment of the present application, by normalizing the sentiment category and the event association strength respectively to obtain the sentiment score and the intensity score, the data category difference represented by the sentiment category and the event association strength can be reduced; and after determining the association weight, the intensity score and the sentiment score are processed based on the association weight to obtain the public opinion influence parameter, which can improve the comprehensiveness and accuracy of the public opinion influence parameter.

[0135] Based on the above embodiments, in the method for determining the news influence degree provided in the embodiments of the present application, determining the influence degree of news data on subjects in the target subject set based on the public opinion influence parameter can be achieved by the following steps:

[0136] Step C1: Obtain industry revenue data of enterprises and / or organizations associated with the target industry entity.

[0137] In one embodiment, the industry revenue data may include revenue data of enterprises and / or organizations associated with the target industry entity within a first time period; illustratively, the first time period may be monthly, quarterly or annual, and the first time period may be associated with a time period corresponding to the news data.

[0138] Step C2: Based on industry revenue data and public opinion impact parameters, determine the impact of news data on target industry entities.

[0139] In one implementation, the impact of news data on the target industry entity can be obtained in the following way:

[0140] The industry revenue data is sorted based on the enterprises and / or organizations associated with the target industry entities to obtain the industry revenue ranking results. The median or specified percentile in the industry revenue ranking results is then used as the industry weight. The public opinion influence parameters are weighted based on the industry weight to obtain the degree of influence of the news data on the target industry entities.

[0141] As can be seen from the above, in the method for determining the degree of news influence provided in the embodiment of the present application, after obtaining the industry revenue data of the enterprises and / or organizations associated with the target industry entity, the degree of influence of the news data on the target industry entity is determined based on the industry revenue data and the public opinion influence parameter. In this way, through the above steps, the correlation between the degree of influence of the news data on the target industry entity and the public opinion influence parameter and the industry revenue data is achieved, thereby improving the accuracy of the above-mentioned degree of influence.

[0142] Based on the above embodiments, in the method for determining the news influence degree provided in the embodiments of the present application, determining the influence degree of news data on subjects in the target subject set based on the public opinion influence parameter can also be achieved in the following ways:

[0143] The degree of influence of news data on the target company is determined as the public opinion influence parameter.

[0144] Exemplarily, the method for determining the news impact degree provided in the embodiment of the present application may further perform the following steps:

[0145] Step D1: Determine the corporate revenue ratio of the first affiliated enterprise of the target company entity.

[0146] Among them, the first affiliated enterprise includes the supply company or purchasing company associated with the target company.

[0147] In one embodiment, the supply company may include a company that provides services and / or materials to the target entity company; illustratively, the supply company may include a supplier of the target entity company.

[0148] In one embodiment, the purchasing company may include a company that purchases products produced or services provided by the target entity company; illustratively, the purchasing company may include a purchaser of the target entity company.

[0149] In one embodiment, the company revenue ratio associated with the supply company may include the ratio between the supplier's sales in the second time period and the annual sales; illustratively, the second time period may be monthly, quarterly or annual, and the second time period may be associated with the time period corresponding to the news data.

[0150] In one embodiment, the company revenue ratio associated with the purchasing company may include the ratio between the purchasing company's purchase amount in the second period and its annual purchase amount.

[0151] Step D2: Determine the impact of the news data on the first related enterprise based on the company's revenue ratio and the public opinion impact parameter.

[0152] In one implementation, the impact degree of the news data on the first related enterprise may be determined by any of the following methods:

[0153] If the first associated company is a supply company, the revenue ratio of the company associated with the supply company will be used as the weight to perform weighted processing on the public opinion impact parameter to obtain the impact of the news data on the supply company.

[0154] If the first associated company is a purchasing company, the revenue ratio of the companies associated with the purchasing company will be used as the weight to perform weighted processing on the public opinion impact parameter to obtain the impact of the news data on the purchasing company.

[0155] From the above, it can be seen that the method for determining the degree of news impact provided in the embodiment of the present application simplifies the process of determining the degree of impact of news data on the target company entity by determining the degree of impact of news data on the target company entity as the public opinion impact parameter; and after determining the company revenue ratio of the first associated enterprise associated with the target company entity, the degree of impact of the news data on the first associated enterprise is determined based on the company revenue ratio and the public opinion impact parameter, and the first associated enterprise includes a supply company or a purchasing company associated with the target company entity. In this way, through the above steps, the degree of impact of the news data on the upstream and downstream industrial chains of the target company entity represented by the first associated enterprise associated with the target company entity can be determined, thereby expanding the scope and breadth of the above-mentioned degree of impact in the company dimension.

[0156] Based on the above embodiments, in the method for determining the news influence degree provided in the embodiments of the present application, determining the influence degree of news data on subjects in the target subject set based on the public opinion influence parameter can also be achieved by the following steps:

[0157] Step E1: Determine first proportion data of the target product entity in the products of the target company entity.

[0158] In one embodiment, the first proportion data may include the revenue ratio of the revenue data corresponding to the target product entity to the revenue data of all products of the target company entity; accordingly, the first statistical result can be obtained by statistics on the revenue data corresponding to the target product entity, and the second statistical result can be obtained by statistics on the revenue data of all products of the target company entity, and then the ratio between the first statistical result and the second statistical result can be determined as the first proportion data.

[0159] In one implementation, the first proportion data may be associated with a time period in which the news data is located.

[0160] Step E2: Based on the first proportion data and the public opinion influence parameter, determine the influence of the news data on the target company subject in the target product subject dimension.

[0161] In one implementation, the influence of news data on the target company subject in the target product subject dimension can be achieved in the following ways:

[0162] The first proportion data is used as the weight to perform weighted calculation on the public opinion impact parameter, and the result of the weighted calculation is determined as the degree of influence of the news data on the target company entity in the target product dimension.

[0163] As can be seen from the above, in the method for determining the degree of news influence provided in the embodiment of the present application, after determining the first proportion data of the target product subject in the products of the target company subject, the degree of influence of the news data on the target company subject in the dimension of the target product subject is determined based on the first proportion data and the public opinion influence parameter. Through the above steps, the refinement and comprehensiveness of the above influence degree are improved, thereby improving the accuracy of the above influence degree.

[0164] Based on the above embodiments, the method for determining the news impact degree provided in the embodiments of the present application may further perform the following steps:

[0165] Step F1: Determine the second associated enterprise associated with the target product entity.

[0166] In one embodiment, the second associated enterprises may include upstream and downstream enterprises associated with the target product entity; illustratively, the upstream and downstream enterprises may provide services and support for the production, sales, promotion and maintenance of the target product entity.

[0167] In one implementation, the second affiliated enterprise may be determined in the following manner:

[0168] Based on the industrial chain in which the target product entity is located, the upstream and downstream products of the target product entity are determined, and then the enterprises corresponding to the upstream and downstream products are determined, and the enterprises corresponding to the upstream and downstream products are determined as the second related enterprises.

[0169] Step F2: Determine the second ratio data of the second associated enterprise and the target product entity.

[0170] In one implementation, the second proportion data may include the proportion of the revenue data of the target product entity to the revenue data of the second associated enterprise.

[0171] Step F3: Based on the second proportion data and the public opinion influence parameter, determine the influence degree of the news data on the second related enterprise in the target product main body dimension.

[0172] In one implementation, the influence of the news data on the second related enterprise in the target product main body dimension can be achieved in the following ways:

[0173] The second proportion data is used as weight data to perform weighted processing on the public opinion impact parameter, and the result of the weighted processing is determined as the above-mentioned impact degree.

[0174] As can be seen from the above, the news influence degree determination method provided in the embodiment of the present application, after determining the second associated enterprise associated with the target product body, determines the second ratio data of the second associated enterprise associated with the target product body, and determines the influence degree of the news data on the second associated enterprise in the target product body dimension based on the second ratio data and the public opinion influence parameter. In this way, through the above operation, the scope of the influence degree is expanded and the comprehensiveness of the influence degree is improved.

[0175] Based on the above embodiments, in the method for determining the news impact degree provided in the embodiments of the present application, determining the target subject set associated with the news data can be achieved by the following steps:

[0176] Step G1: Construct the subject knowledge graph.

[0177] Among them, the subject knowledge graph includes the interrelated relationships between various types of company subjects, product subjects and industry subjects.

[0178] Node Type Node Description Company Node Represents an independent corporate entity, including company code, stock code and name Product Node Represents a product entity, including product code and name Industry Node Indicates the industry branch of the market, including industry code and name Financial Data Node Indicates the company's financial information, such as main business income and income ratio Product Links Indicates the upstream and downstream relationship between products Industry chain nodes Represents the hierarchical structure between industries

[0179] Table 1

[0180] In one embodiment, the company subject, product subject, and industry subject in the subject knowledge graph may exist in the form of nodes; Table 1 is a statistical result of the types of subject nodes provided in the embodiment of the present application. Exemplarily, the relationship between the subject nodes in the subject knowledge graph may be as shown in Table 2:

[0181] Relationship Type Relationship Description Starting Node Termination Node Production relations Products produced by the company company product Belong to relationship The industry the product belongs to product industry Father-son relationship Hierarchical relationships between industries industry industry Product parent-child relationship The relationship between raw materials and finished products Product Chain Product Chain Industry relations stage Mapping relationships between industries industry industry

[0182] Table 2

[0183] In the above table, for example, the company's financial information can represent the company's financial status and operating performance, while the product chain and industry chain data show the intrinsic connection and upstream and downstream relationship between products and industries.

[0184] In one implementation, the subject knowledge graph may be constructed in any of the following ways:

[0185] Acquire various structured data on industry classifications, company financial information, product chains, and industry chains in advance, analyze the correlation between the above data, connect company nodes, product nodes, and industry nodes through specific relationship edges, and thus build a complex and orderly subject knowledge graph; the above structured data can cover multiple dimensions of the financial market, the levels and correlation information between multiple industries.

[0186] Neo4j technology is used to analyze various types of structured data to build a subject knowledge graph containing three types of entities: companies, products, and industries. Among them, Neo4j, as an advanced graph database technology, is designed for processing and analyzing complex relationship networks. Neo4j can intuitively display the key nodes in the industrial chain and the mutual influence between them, so as to effectively infer the specific impact of news reports on the industrial chain. Neo4j's graph construction capability provides strong support for in-depth understanding of the dynamics of the industrial chain and the impact of events.

[0187] Figure 4A A schematic diagram of the structure of the first knowledge graph provided in an embodiment of the present application. Figure 4B A schematic diagram of the structure of the second knowledge graph provided in an embodiment of the present application. Figure 4C A schematic diagram of the structure of the third knowledge graph provided in an embodiment of the present application.

[0188] The second knowledge graph may represent a set of two knowledge graphs constructed based on structured data from different sources, and the third knowledge graph may represent the mapping relationship between products and industries, and the mapping relationship between companies and products. The main knowledge graph may include the first knowledge graph to the third knowledge graph.

[0189] Step G2: determine the subject type set associated with the news data.

[0190] Among them, the subject type set includes company type, product type and industry type.

[0191] In one implementation, the subject type set may include types of subjects associated with or included in the news data.

[0192] In one implementation, the subject type set may be determined by:

[0193] AI LLM is used to process news data in combination with prompt words to determine the subject type set. Specifically, by inputting the preset fourth prompt into AI LLM: "Please determine whether the news is about a company, an industry, or an economic product. If it is about a company, return 0; if it is about an industry, return 1; if it is about an economic product, return 2", the AI ​​LLM model will return the corresponding digital code according to the content of the news data, thereby realizing a quick judgment of the subject type; wherein the prompt word may include the fourth prompt.

[0194] Step G3: Determine the target subject set from the subject knowledge graph based on the matching distance between the subject name associated with the news data and the subject in the subject knowledge graph.

[0195] Among them, the target subject set is associated with the subject type set.

[0196] In one implementation, the subject name may include a company name, an industry name, and a product name.

[0197] In one implementation, the process of determining the target subject set may be implemented with the aid of Retrieval-Augmented Generation (RAG) and HanLP.

[0198] Among them, RAG technology is an innovative information processing framework that combines the advantages of information retrieval and generation models. In its data processing flow, a retrieval model (such as ChromaDB) is first used to accurately find the most relevant documents to the input query from a predefined text database. Subsequently, the text data is deeply processed through a generation model including a generative pre-trained transformer (GPT) to generate an answer or summary that integrates the retrieval results. The introduction of RAG technology has significantly improved the accuracy and richness of information extraction, and provided a powerful tool for accurately understanding and responding to complex public opinions.

[0199] Hanlp is a powerful natural language processing toolkit with the characteristics of high precision, high performance, up-to-date corpus, clear architecture and customization. The embodiment of the present application uses Hanlp's Named Entity Recognition (NER) technology to accurately extract company names and other relevant information from news data. Hanlp provides a more efficient and accurate solution. In addition, the Hanlp version also supports multiple joint tasks and multiple single tasks in multiple languages ​​including simplified and traditional Chinese, English, Japanese, Russian, French and German, meeting the complex information processing needs across languages ​​and domains.

[0200] In one embodiment, the target product body can be determined by:

[0201] AI LLM is used to extract product names from news data, and match product names with products in the main knowledge graph. Distance calculation is used as the matching criterion, and automatic screening and sorting are performed. Finally, the top three product nodes that best match the product name are selected and determined as the target product subject. Through the above steps, specific product information can be provided for public opinion analysis. For example, the matching distance calculation between product names and knowledge graphs follows the squaredL2 norm of ChromaDB by default, as shown in formula (2):

[0202]

[0203] Among them, d is the matching distance, A i is the product name, B i The product node.

[0204] In one implementation, the target industry entity may be determined by:

[0205] RAG technology is used to match industries based on ChromaDB's retrieval model and AI big model. Figure 4D As shown, Figure 4D A schematic diagram of the industry matching principles provided for an embodiment of the present application.

[0206] First, we use ChromaDB to build a vector retrieval model for all nodes at all levels in all industries, so that each industry node can be efficiently stored and retrieved in the form of vectors.

[0207] Next, starting from the top-level industry, the industry names associated with the news data are matched with the vectors of all the top-level industry nodes. During the matching process, the squared L2 norm is used as a calculation method to measure the vector distance to evaluate the similarity between the industry name and the industry node. Through calculation, the top 5 industries with the closest distance are selected as candidate industry entities.

[0208] Then, the candidate industry entities and news data are input into the AI ​​big model to trigger the AI ​​big model to further determine which industry entity is closest to the news data based on the content and context information of the news data, and return the industry as the matching result of the current level. In the above process, the fifth prompt can also be combined to guide the AI ​​big model to select an industry; the fifth prompt can be: "Please determine which industry {industry_list} the following news is closest to? Return the elements in the list, no other explanation is required, no text can be included, and only one can be returned."

[0209] At the next industry level, starting from the selected top industry, find all nodes of the next industry level contained therein. Similarly, calculate the vector distance between the news data and these industry nodes, and find the sub-industry with the smallest distance. And so on, until you traverse to the bottom industry level and get a complete industry chain. Finally, select the industry node of the closest level on the industry chain as the target industry subject. Through the above steps, the accuracy of the target industry subject is improved, so that the target industry subject can accurately reflect the industry category to which the news text belongs.

[0210] In one embodiment, the target company entity may be determined by:

[0211] The Hanlp model is used to extract the target company entity. After preprocessing the news data, the entity recognition function of Hanlp is used to extract key information such as the company name from the news data. Then, the target company entity is determined from the subject knowledge graph by the distance between the company name and the company node in the subject knowledge graph through the method provided in the above embodiment. Among them, the Hanlp model can accurately identify and classify various components in the news, such as companies, organizations, names, etc., among which the company name is marked as a specific entity category and accurately extracted.

[0212] As can be seen from the above, in the method for determining the degree of news influence provided in the embodiment of the present application, a subject knowledge graph including various types of company subjects, product subjects, and industry subjects is constructed, and after determining the subject type set associated with the news data, a target subject set is determined from the subject knowledge graph based on the matching distance between the subject name associated with the news data and the subject in the subject knowledge graph, and the target subject set is associated with the subject type set. In this way, through the above operation, the accuracy and precision of the target subject set are improved.

[0213] Figure 5A A flowchart of determining the impact of news provided in the embodiment of the present application is shown in FIG. Figure 5A As shown, after inputting the public opinion influence factor, the impact degree of the news data can be determined from the product perspective, the enterprise perspective and the industry perspective through impact propagation reasoning; wherein the public opinion influence factor can be the public opinion influence parameter in the aforementioned embodiment.

[0214] Among them, from the product perspective, first, the impact on the enterprise can be deduced from the product, which can specifically include extracting economic products, matching economic products in the database, the business to which the economic products belong in the enterprise, and the proportion of the business in the company's revenue, and then determining pct*factor_a; wherein the economic product can correspond to the target product subject in the aforementioned embodiment, the enterprise can be the target company subject, and pct*factor_a can be the degree of influence of the news data on the target product subject of the target company subject.

[0215] Secondly, find the upstream and downstream products according to the industrial chain, find the corresponding enterprises involved, and deduce the impact, which may specifically include: extracting economic products, matching economic products in the database, upstream and downstream economic products and involved enterprises, the business to which the economic products belong in the enterprise, and the proportion of the business in the enterprise's revenue, so as to determine pct*factor_b; wherein, the upstream and downstream products can be the second related enterprise in the aforementioned embodiment, and pct*factor_b can be the impact of the news data on the second related enterprise.

[0216] From the perspective of enterprises, the impact on upstream and downstream enterprises and the extent of their impact can be determined based on the supply chain. Specifically, it can include: extracting enterprises, suppliers and customers, and the proportion of purchases / sales to the annual purchases / sales, and then determining pct*factor_c; wherein the enterprise can be the target company in the aforementioned embodiment, and the supplier and customer can be the first associated enterprise, and pct*factor_c can be the impact of the news data on the first associated enterprise.

[0217] From an industry perspective, we can locate leading companies based on the companies with the largest industry revenue and deduce the size of the impact. Specifically, we can include: extracting industries, the revenue rankings of companies in the industry, and the company's share of the total industry revenue, and finally obtain pct*factor_d; among which, pct*factor_d can be the degree of impact of news data on the target industry entities.

[0218] Figure 5B A schematic diagram of a structure for determining the news impact level provided in an embodiment of the present application, such as Figure 5B As shown, this structure can be implemented by the following modules:

[0219] The subject association module is used to obtain news content, and obtain the subject product, subject company and subject industry from the news data through RAG technology and Hanlp technology, and send them to the public opinion analysis module respectively; among them, the subject product, subject company and subject industry can be the target product subject, target company subject and target industry subject respectively, and the news content can be news data.

[0220] Exemplarily, the public opinion analysis module is used to analyze news data, main products, main companies and main industries through CoT technology to determine the public opinion impact parameters; then analyze the impact degree of each direct subject to determine the impact degree of news data on the main products, main companies and main industries.

[0221] Exemplarily, the impact propagation reasoning module can also be used to analyze the impact degree on other indirect entities.

[0222] Figure 5C Another structural diagram of determining the news impact degree provided in the embodiment of the present application is as follows: Figure 5CAs shown, the structure can include three parts: subject association, public opinion analysis, and impact propagation reasoning. Among them, subject association is used to build a knowledge graph based on industrial chain supply chain data, and analyze the news public opinion data based on the knowledge graph, so as to perform news subject type judgment, and further perform product matching, company extraction and industry matching, so as to determine the target industry subject, target product subject and target company subject; for public opinion analysis, it can perform sentiment analysis, event type prediction and event impact intensity prediction; for impact propagation reasoning, it can quantify the influence of upstream and downstream products in the industrial chain, upstream and downstream companies in the supply chain and related companies in the industry from the perspective of economic products, enterprises and industries based on public opinion influencing factors.

[0223] Specifically, the above structure can use a large model to perform entity extraction, sentiment analysis, event type prediction and event impact intensity prediction on news data, and combine the prediction results to obtain the public opinion impact types including strong positive, strong negative, medium, weak positive and weak negative. Combined with the transmission method of the industrial chain and supply chain, the impact degree of the news public opinion is output.

[0224] Through the above-mentioned process and structure, a multi-dimensional and comprehensive prediction and determination of the impact degree of news data is achieved, thereby improving the comprehensiveness and accuracy of the impact degree of news data.

[0225] In summary, the technical solution provided by the embodiment of the present application realizes the fusion of multi-source heterogeneous data and constructs a detailed industrial supply chain knowledge map. It not only integrates data from multiple dimensions such as industry, company finance and economic products, but also reveals the complex industrial chain and supply chain dynamics. By integrating these data, a solid foundation is provided for public opinion impact analysis, and a more comprehensive and accurate market dynamic view can be obtained, which helps to accurately reflect market dynamics, timely discover potential risks, and provide strong support for subsequent regulatory decisions; secondly, by using AI big models to predict news sentiment, event types, and impact intensity, compared with traditional public opinion analysis methods, the inefficient methods of manual labeling and rule matching are abandoned, and the powerful semantic understanding and analysis capabilities of AI big models are relied on, which not only improves the efficiency of information processing, but also improves the accuracy of prediction; by automatically extracting and analyzing a large amount of public opinion information, it can respond to changes in market public opinion more quickly, and provide data basis for timely determination of regulatory decisions, thereby providing conditions for maintaining market order and investor interests; finally, through impact propagation reasoning, an effective impact transmission model is constructed, and the risk signals hidden behind public opinion are identified through the production and supply chain.

[0226] At the same time, this plan can also explain implicit and indirect public opinion risks, thereby improving the intuitiveness and real-time nature of public opinion risks, providing data support for the designation and implementation of risk management strategies, improving the scientific nature and effectiveness of market supervision, and improving the targeted nature of market risk resolution.

[0227] Moreover, in the field of natural language processing, large models such as GPT-3 have demonstrated strong capabilities in complex tasks such as zero-sample learning, laying a solid foundation for using large models for text analysis. These large language models can deeply understand the emotional tendencies of texts at the semantic level by fitting the statistical laws of language. This company applies LLM to the field of public opinion analysis. By combining the analytical capabilities of LLM with the characteristics of public opinion data, it can make rapid and accurate judgments on market changes, thereby providing more powerful support for decision-making.

[0228] Based on the above embodiments, the present application also provides a device for determining the impact degree of news. Figure 6 A schematic diagram of the structure of a device for determining the news impact degree provided in an embodiment of the present application is shown in FIG. Figure 6 As shown, the news impact degree determining device 6 may include:

[0229] An acquisition module 601 is used to acquire news data to be processed;

[0230] Determination module 602, used to determine the target subject set associated with the news data; wherein the target subject set includes a target company subject, a target product subject, and a target industry subject;

[0231] The determination module 602 is also used to determine the public opinion impact parameter of the news data; based on the public opinion impact parameter, determine the impact degree of the news data on the subjects in the target subject set.

[0232] In some embodiments, the determination module 602 is used to perform sentiment analysis on news data through a first AI model to determine the sentiment category of the news data; perform event analysis on the news data through a second AI model to determine the event type associated with the news data; determine the event association strength based on the news data and the event type; and determine the public opinion impact parameter based on the sentiment category and the event association strength.

[0233] In some embodiments, the determination module 602 is used to normalize the sentiment categories to obtain sentiment scores; normalize the event association strength to obtain strength scores; determine association weights; and process the strength scores and sentiment scores based on the association weights to obtain public opinion impact parameters.

[0234] In some embodiments, the acquisition module 601 is used to acquire industry revenue data of enterprises and / or organizations associated with the target industry entity;

[0235] The determination module 602 is used to determine the impact of the news data on the target industry entity based on the industry revenue data and the public opinion impact parameter.

[0236] In some embodiments, determination module 602 is used to determine the impact of news data on the target company entity as a public opinion impact parameter; determine the company revenue ratio of the first affiliated enterprise of the target company entity; based on the company revenue ratio and the public opinion impact parameter, determine the impact of the news data on the first affiliated enterprise; wherein the first affiliated enterprise includes a supply company or a purchasing company associated with the target company entity.

[0237] In some embodiments, the determination module 602 is used to determine the first proportion data of the target product entity in the products of the target company entity; based on the first proportion data and the public opinion influence parameter, determine the influence degree of the news data on the target company entity in the target product entity dimension.

[0238] In some embodiments, determination module 602 is used to determine a second associated enterprise associated with the target product entity; determine second proportion data associated with the second associated enterprise and the target product entity; and determine the degree of influence of news data on the second associated enterprise in the target product entity dimension based on the second proportion data and public opinion influence parameters.

[0239] In some embodiments, the determination module 602 is used to construct a subject knowledge graph; wherein the subject knowledge graph includes the interrelated relationships between various types of company subjects, product subjects, and industry subjects;

[0240] The determination module 602 is further used to determine a subject type set associated with the news data; wherein the subject type set includes company type, product type and industry type;

[0241] The determination module 602 is also used to determine the target subject set from the subject knowledge graph based on the matching distance between the subject name associated with the news data and the subject in the subject knowledge graph; wherein the target subject set is associated with the subject type set.

[0242] The present application also provides a device for determining the impact of news. Figure 7 A schematic diagram of the structure of a device for determining the news impact level provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the news impact degree determination device 7 includes a processor 701 and a memory 702; wherein the memory 702 stores a computer program; when the computer program is executed by the processor 701, it can implement the news impact degree determination method as described above.

[0243] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored; when the computer program is executed by a processor of an electronic device, it can implement the method for determining the degree of news impact as described in any of the above.

[0244] An embodiment of the present application also provides a computer program product, which includes a computer program; when the computer program is executed by a processor of an electronic device, it can implement the method for determining the degree of news impact as described in any of the above.

[0245] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0246] The methods disclosed in the various method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0247] The features disclosed in the various product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0248] The features disclosed in the various method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0249] It should be noted that the above-mentioned computer-readable storage medium can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic random access memory (FRAM), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM) and other memories; it can also be various electronic devices including one or any combination of the above-mentioned memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc.

[0250] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0251] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0252] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus necessary general hardware nodes, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0253] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0254] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0255] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0256] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for determining the degree of news impact, characterized in that: The method comprises: Get the news data to be processed; Determine a target subject set associated with the news data; wherein the target subject set includes a target company subject, a target product subject, and a target industry subject; Determining public opinion impact parameters of the news data; The influence degree of the news data on the subjects in the target subject set is determined based on the public opinion influence parameter.

2. The method according to claim 1, characterized in that Determining the public opinion impact parameter of the news data includes: Performing sentiment analysis on the news data by using a first artificial intelligence model to determine the sentiment category of the news data; Performing event analysis on the news data by using a second artificial intelligence model to determine the event type associated with the news data; Determining event association strength based on the news data and the event type; The public opinion impact parameter is determined based on the emotion category and the event association strength.

3. The method according to claim 2, characterized in that The determining the public opinion impact parameter based on the emotion category and the event association strength includes: Normalizing the emotion categories to obtain emotion scores; Normalizing the event association strength to obtain a strength score; Determine the association weight; The intensity score and the sentiment score are processed based on the association weight to obtain the public opinion impact parameter.

4. The method according to claim 1, characterized in that: The determining, based on the public opinion influence parameter, the influence degree of the news data on the subjects in the target subject set includes: Obtain industry revenue data of enterprises and / or organizations associated with the target industry entity; Based on the industry revenue data and the public opinion impact parameter, determine the impact degree of the news data on the target industry entity.

5. The method according to claim 1, characterized in that The determining, based on the public opinion influence parameter, the influence degree of the news data on the subjects in the target subject set includes: Determining the influence degree of the news data on the target company entity as the public opinion influence parameter; The method further comprises: Determining the company revenue ratio of the first affiliated enterprise of the target company entity; wherein the first affiliated enterprise includes a supply company or a purchasing company associated with the target company entity; Based on the company's revenue ratio and the public opinion impact parameter, the impact degree of the news data on the first related enterprise is determined.

6. The method according to claim 1, characterized in that The determining, based on the public opinion influence parameter, the influence degree of the news data on the subjects in the target subject set includes: Determine first proportion data of the target product entity in the products of the target company entity; Based on the first proportion data and the public opinion influence parameter, the influence degree of the news data on the target company entity in the target product entity dimension is determined.

7. The method according to claim 1 or 6, characterized in that: The method further comprises: Determine a second associated enterprise associated with the target product entity; Determine second ratio data of the second associated enterprise being associated with the target product entity; Based on the second proportion data and the public opinion influence parameter, the influence degree of the news data on the second related enterprise in the target product main dimension is determined.

8. The method according to claim 1, characterized in that The step of determining a target subject set associated with the news data includes: Constructing a subject knowledge graph; wherein the subject knowledge graph includes the interrelated relationships between various types of company subjects, product subjects, and industry subjects; Determine a subject type set associated with the news data; wherein the subject type set includes company type, product type and industry type; Based on the matching distance between the subject name associated with the news data and the subject in the subject knowledge graph, the target subject set is determined from the subject knowledge graph; wherein the target subject set is associated with the subject type set.

9. A device for determining the degree of news impact, characterized in that: The news impact degree determination device comprises: An acquisition module is used to acquire news data to be processed; A determination module, used to determine a target subject set associated with the news data; wherein the target subject set includes a target company subject, a target product subject, and a target industry subject; The determination module is further used to determine a public opinion impact parameter of the news data; and determine the degree of impact of the news data on subjects in the target subject set based on the public opinion impact parameter.

10. A device for determining the impact of news, characterized in that: The news impact determination device includes a processor and a memory; wherein a computer program is stored in the memory; when the computer program is executed by the processor, it can implement the news impact determination method as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The storage medium stores a computer program; when the computer program is executed by a processor of an electronic device, the method for determining the degree of news impact as described in any one of claims 1 to 8 can be implemented.

12. A computer program product, characterized in that The program product includes a computer program; when the computer program is executed by a processor of an electronic device, it can implement the method for determining the degree of news impact as described in any one of claims 1 to 8.

Images