WO2023273295A1

WO2023273295A1 - Enterprise esg index determination method based on clustering technology, and related product

Info

Publication number: WO2023273295A1
Application number: PCT/CN2022/071253
Authority: WO
Inventors: 诸世卓; 邵熹; 胡逸群
Original assignee: 平安科技（深圳）有限公司
Priority date: 2021-07-02
Filing date: 2022-01-11
Publication date: 2023-01-05
Also published as: CN113190683A; CN113190683B

Abstract

Provided are an enterprise ESG index determination method based on a clustering technology, and a related product. The method comprises: acquiring M pieces of news, which are within a preset time period, of an enterprise to be evaluated; clustering the M pieces of news to obtain K first news groups; clustering one or more pieces of news in each first news group, so as to obtain L second news groups corresponding to each first news group; according to a piece of original news and H pieces of reprinted news, which are comprised in each second news group of the L second news groups, determining a target public opinion score of a news event corresponding to each first news group, wherein the target public opinion score is used for representing the influence of the news event corresponding to each first news group on said enterprise (104); and performing ESG evaluation on said enterprise according to the target public opinion score of the news event corresponding to each first news group, so as to obtain an ESG index of said enterprise (105).

Description

Enterprise ESG Index Determination Method and Related Products Based on Clustering Technology

This application claims the priority of the Chinese patent application with the application number 202110748931.X filed on July 2, 2021, and the title of the invention is "Method for Determining Enterprise ESG Index Based on Clustering Technology and Related Products", all of which The contents are incorporated by reference in this application.

technical field

This application relates to the technical field of data processing, in particular to a method for determining an enterprise ESG index based on clustering technology and related products.

Background technique

A company's ESG score is a comprehensive score for the company's environment (E), society (S), and governance (G). International and domestic companies have accumulated some successful experience in scoring ESG performance. Internationally renowned rating agencies such as MSCI and FTSE have established their own scoring standards and conducted ESG evaluations on internationally renowned companies. As the international community, such as various investment institutions and governments, attach importance to corporate responsibility, especially with the recent progress of global climate change cooperation, and China's international commitment to achieve carbon peaking by 2030 and carbon emissions by 2060 China's domestic investors and the government also have a lot of demand for companies' ESG scores.

The inventor realized that when evaluating the ESG performance of an enterprise, it is necessary to use news related to the enterprise to score the enterprise on ESG. At present, ESG scores for companies are mainly based on the number of news related to the company. However, there is a lot of noise in the news obtained. The accuracy of ESG scores is low, resulting in low accuracy of investment decisions based on ESG scores.

Contents of the invention

The embodiment of this application provides a method for determining an enterprise ESG index based on clustering technology and related products, distinguishing original news and reprinted news in multiple news articles of a news event, improving the accuracy of ESG scoring by enterprises, and formulating Accuracy of investment decisions.

In the first aspect, the embodiment of the present application provides a method for determining an enterprise ESG index based on clustering technology, including:

Obtain M pieces of news about the enterprise to be evaluated within a preset time period, where M is an integer greater than or equal to 1;

Clustering the M pieces of news to obtain K first newsgroups, wherein each first newsgroup in the K first newsgroups corresponds to a news event, and each first newsgroup includes One or more news articles in the M news articles, K is an integer greater than or equal to 1;

Clustering one or more news articles in each of the first newsgroups to obtain L second newsgroups corresponding to each of the first newsgroups, wherein each of the second newsgroups includes One piece of original news and H pieces of reprinted news corresponding to the original news, where H is an integer greater than or equal to 0;

According to an original news and H pieces of reposted news included in each of the L second newsgroups, determine the target public opinion score of the news event corresponding to each of the first newsgroups, the target The public opinion score is used to characterize the influence of the news event corresponding to each first news group on the enterprise to be evaluated;

According to the target public opinion score of the news event corresponding to each first news group, ESG evaluation is performed on the enterprise to be evaluated, and the ESG index of the enterprise to be evaluated is obtained;

Sending the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.

In the second aspect, the embodiment of the present application provides an enterprise ESG index determination device, including:

The obtaining unit is used to obtain M pieces of news of the enterprise to be evaluated within a preset time period, where M is an integer greater than or equal to 1;

A processing unit, configured to cluster the M pieces of news to obtain K first news groups, wherein each first news group in the K first news groups corresponds to a news event, and each of the K first news groups corresponds to a news event. The first news group includes one or more news articles in the M news articles, K is an integer greater than or equal to 1;

The sending unit is configured to send the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.

In a third aspect, an embodiment of the present application provides an electronic device, including: a processor, the processor is connected to a memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory , so that the electronic device executes the method as described in the first aspect, the method includes:

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program causes a computer to execute the method as described in the first aspect, the method comprising:

In the implementation manner of the present application, the ESG index with high precision can be sent to the target device, so as to improve the accuracy of the investment decision made.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings that need to be used in the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present application. For Those of ordinary skill in the art can also obtain other drawings based on these drawings without making creative efforts.

FIG. 1 is a schematic flowchart of a method for determining an enterprise ESG index based on clustering technology provided by an embodiment of the present application;

Fig. 2 is a block diagram of functional units of an enterprise ESG index determination device provided by the embodiment of the present application;

FIG. 3 is a schematic structural diagram of a device for determining an enterprise ESG index provided by an embodiment of the present application.

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of this application.

The terms "first", "second", "third" and "fourth" in the specification and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order . Furthermore, the terms "include" and "have", as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but optionally also includes unlisted steps or units, or optionally further includes For other steps or units inherent in these processes, methods, products or apparatuses.

Reference herein to an "embodiment" means that a particular feature, result, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The occurrences of this phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is understood explicitly and implicitly by those skilled in the art that the embodiments described herein can be combined with other embodiments.

This application may relate to the field of artificial intelligence technology, such as acquiring and processing relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .

First of all, the corporate ESG index or ESG index involved in the embodiment of this application is the corporate ESG score, which can also be called ESG score. They are essentially the same and will not be distinguished later. Correspondingly, the ESG evaluation of enterprises, or the ESG scoring of enterprises, are essentially the same, and they are all used to determine the ESG index of enterprises.

Referring to Fig. 1, Fig. 1 is a clustering technology-based ESG index method for enterprises provided by the embodiment of the present application. This method is applied to the enterprise ESG index determination device. The method includes the following steps:

101: Obtain M pieces of news about the enterprise to be evaluated within a preset period of time, wherein the M pieces of news are used for ESG evaluation, and M is an integer greater than or equal to 1.

Exemplarily, the enterprise ESG index determination device can obtain W articles of news within a preset time period from multiple news media platforms through crawler technology, and identify the W articles of news (for example, text classification can be performed on W articles of news, ) Select N news related to ESG evaluation from W news. For example, if a certain news discloses the energy consumption of a certain company, this news is regarded as news related to ESG index.

Wherein, the preset time period may be any historical time period, for example, it may be the last ten days, last month, or last year, and so on. This application does not limit the preset time period. That is to say, the enterprise ESG index determination method of this application can evaluate the ESG of the enterprise based on the news in the most recent time period, and obtain the ESG index of the enterprise in the most recent time period; News conducts ESG evaluations on companies and obtains the ESG index of companies in a certain historical period.

Exemplarily, entity recognition can be performed on each of the N news articles to obtain the entity words in each news article, and determine the company involved in each news article according to the entity words in each news article, wherein the entity words can be is the name of the enterprise (for example, Chinese name, English name), the products of the enterprise, the location of the enterprise, the ranking of the enterprise, etc.; All companies involved in this news. Correspondingly, according to the companies involved in each news, the number of news related to each company in all companies involved in N news can be obtained, that is, M news related to the company to be evaluated is obtained, and M is less than or equal to N, wherein, the enterprise to be evaluated is any enterprise among all enterprises.

102: Clustering the M news pieces to obtain K first news groups, wherein each first news group in the K first news groups corresponds to a news event, and each first news group A group includes one or more news articles in the M news articles, K is an integer greater than or equal to 1.

Exemplarily, the device for determining the ESG index of an enterprise extracts semantic information from each of the M pieces of news to obtain a semantic vector corresponding to each piece of news, wherein the semantic vector corresponding to each piece of news is used to represent the information contained in each piece of news Describe the news event. Cluster M news based on the semantic vector of each news to obtain K first news groups, that is, determine the Euclidean distance between the semantic vectors of any two news in M news, and combine the semantic vectors of any two news The Euclidean distance between them is used as the first similarity between any two news articles; then, the two news articles with the first similarity greater than the first threshold are classified into one category, that is, they are classified into a first news group, Get the K first newsgroups. Since the first similarity between the semantic vectors of the two news is greater than the first threshold, it means that the two news describe a news event, so one or more news descriptions included in each first news group It's the same news event.

103: Clustering one or more news articles in each first news group to obtain L second news groups corresponding to each first news group, wherein each second news group The group includes one original news and H reprinted news corresponding to the original news, where H is an integer greater than or equal to 0.

Exemplarily, the enterprise ESG index determination device determines the second similarity between any two news articles in each first news group, and classifies the news with the second similarity greater than the second threshold into one category, that is, classifies them into one In the second newsgroup, get L second newsgroups. Specifically, the second similarity between any two news articles in each first news group can be the similarity between the news content of the two news articles, for example, the headline and summary of the two news articles can be calculated. The similarity between the title and the abstract is used as the similarity between the two news, if the similarity is greater than the second threshold, it means that the two news are the same news (that is, one of the two news One article is the original news, and the other is the news that reprinted the original news; or, both news are reprinted news). After the above clustering, a piece of original news and all reprinted news that reprinted this original news can be clustered into a second news group. Also, because the contents of the two original news articles are different, the two original news articles will be clustered into two different second newsgroups. Therefore, through the above-mentioned clustering method, all original news can be separated, and reprinted news corresponding to each original news can be clustered together to obtain L second news groups.

In addition, for each second news group, it must contain an original news and H reprinted news. When H is greater than or equal to 1, if you want to determine the original news in each second news group, you can Obtain the publication time of each news item in each second newsgroup. For original news, the publication time must be the earliest, so the news with the earliest publication time in each second news group can be used as the original news in each second news group, and the rest of the news is the first news of each second news group. Reprinted news in the group.

104: According to an original news and H pieces of reprinted news included in each of the L second news groups, determine the target public opinion score of the news event corresponding to each of the first news groups, so The target public opinion score is used to represent the influence of the news event corresponding to each first news group on the enterprise to be evaluated.

Exemplary, first determine the scaling ratio of each first news group from the news quantity, then determine the original public opinion score of each first news group from the news content; finally, the scaling ratio of each first news group and The original public opinion scores are fused to obtain the target public opinion scores of each first news group. Among them, the scaling ratio of each first news group represents the social attention to the news event corresponding to each first news group, and the target public opinion score is used to represent the news event corresponding to each first news group. Influence refers to the degree of influence on the enterprise being evaluated.

Specifically, a preset ratio of original news in each second news group is acquired, wherein the preset ratio represents the contribution of original news to social attention to news events corresponding to each first news group. It should be understood that original news is independently created by journalists and has nothing to do with troll hype. Therefore, all original news under a news event should attract the same attention to this news event, that is to say, every The original news is made in terms of the social attention received by the news event. Therefore, a ratio is preset for each original news event, so that the contribution of original news in terms of attention is not affected by the number of reprinted news; according to the preset ratio of original news in each second news group and each The number H of reposted news contained in the second newsgroup determines the scaling ratio of each second newsgroup.

Exemplarily, the scaling ratio of each second newsgroup can be expressed by formula (1):

heat _j ＝log2(H _j +1)+a formula (1)

Among them, heat _j is the zoom ratio of the jth second newsgroup in the L second newsgroups, H _j is the number of reprinted news contained in the jth second newsgroup, a is each second newsgroup The preset ratio of original news in , generally takes a value of 1.

It can be seen that when the above formula (1) calculates the scaling ratio of each second news group, as the number of reprinted news increases, the growth rate of the scaling ratio gradually weakens, so that the behavior of deliberate reprinting or hype will be weakened. That is to say, through the formula (1), the behavior of deliberate reposting or troll hype is weakened, so that the determined scaling ratio (attention) is more accurate, that is, it is not easily affected by troll hype.

Exemplarily, the scaling ratio of each second newsgroup can also be expressed by formula (2):

It can be seen that when H _j =0, that is, when the number of reprinted news in each second news group is zero, the scaling ratio of the second news group is a, that is, the scaling ratio column is only related to the number of original news ; When H _j is infinite, the scaling ratio of the second newsgroup is listed as 1+a, that is, there is an upper limit. In this way, even if there is speculation by the trolls, the behavior of the hype will be converged at 1+a, and it will not be endlessly affected by the hype behavior, and the society's attention to news events will not be infinitely enlarged. The degree of attention of the news event itself will not be infinitely magnified, so that the determined scaling ratio will be more accurate, and the final calculated target public opinion score will be more accurate.

Further, summing the scaling ratios of each second newsgroup to obtain a scaling ratio column of each first newsgroup. Therefore, the scaling ratio of each first newsgroup can be expressed by formula (3):

Wherein, heat is the zoom ratio of each first newsgroup.

In one embodiment of the present application, in determining the public opinion score of the news event corresponding to each first news group, a group of second news groups can be randomly selected from the L second news groups corresponding to each first news group , and carry out emotion recognition on any piece of news in the second newsgroup to obtain the emotion label corresponding to the news, and use the emotion label of the news as the emotion label corresponding to each first newsgroup; or, from A piece of news is randomly selected from all the news corresponding to each first news group for emotion recognition, and the emotional label of the news is obtained, and the emotional label of the news is used as the emotional label corresponding to each first news group. Because the news events described by all the news under each first news group are the same. Therefore, the mode of adding emotional tags for each first newsgroup is more flexible, and does not limit the mode of adding emotional tags for each first newsgroup. Wherein, the sentiment tag corresponding to each first news group is used to represent whether the news event corresponding to each first news group is a positive news event or a negative news event.

Further, when the emotion tag is used to represent that the news event described by each first news group is a negative news event, the specific content of the negative news event can be determined, for example, it can be the amount of fines, business interruption, high-level criminal detention or prohibition of government Purchasing, etc.; when the sentiment tag is used to represent that the news event described by each first news group is a positive news event, determine the specific content of the positive news event, for example, it can be effective for energy saving and emission reduction.

Further, determine the publishing media of all news in the L second newsgroups; for example, carry out text recognition to the news content of each news in the L second newsgroups, identify the content of each news from each news content Publishing media, for example, the publishing media of some news is located at the end of the news content, and the publishing media of some news is located at the very beginning of the news content. Therefore, carry out text recognition to the news content of each news, can obtain the publishing media of each news; Publish the media. It should be understood that the priority relationship of publishing media is preset in the enterprise ESG index determining device. Generally speaking, the priorities are as follows: national-level publishing media > provincial-level publishing media > municipal-level publishing media > self-media > entertainment media, and so on. Therefore, after identifying the publishing media of each news article in the L second newsgroups, the highest-level publishing media in each first newsgroup is determined according to the preset priority relationship.

Finally, according to the mapping relationship between the publishing media, emotional labels and public opinion scores, as well as the highest-level publishing media in each first news group and the corresponding emotional tags of each first news group, determine the The original public opinion score corresponding to the group.

Exemplarily, the mapping relationship between publishing media, emotional labels and public opinion scores can be as follows: if the news event is characterized as a neutral news event (that is, there is no criticism or praise), then set the public opinion score to 0; if the news event represents When it is a negative news event, points will be deducted from 0 to -10, and the degree of deduction is related to the specific content of the negative news event and the level of the publishing media. For example, the news event corresponding to each first news group is If executives are detained and the highest-level publishing media is a national-level publishing media, 10 points will be deducted, that is, the original public opinion score corresponding to each first news group is -10 points; if the news event is characterized as a positive news event, Add from 0 to 5. Similarly, the degree of deduction is related to the specific content of the negative news event and the level of the publishing media; 2 points are added to the publishing media of the first level, that is, the original public opinion score corresponding to the news event is 2 points. That is to say, the mapping relationship between various types of news events, various publishing levels and public opinion scores is pre-configured. After the highest-level publishing media and the type of news events of each first news group are determined, the original public opinion score corresponding to each first news group can be determined according to the mapping relationship.

In addition, because negative news is mostly initiated by other parties (such as the media, the government, etc.) to monitor public opinion, while positive news may be initiated by the company’s own public relations department, the range of deductions for negative news (-10) is lower than that of positive news. The range of extra points (+5) is wide, which makes the distribution of public opinion scores more reasonable.

Finally, the original public opinion score of each first news group is multiplied by the scaling ratio to obtain the target public opinion score corresponding to each first news group.

Therefore, the target public opinion score corresponding to each first news group can be expressed by formula (4):

adjscoreEvent＝heat*adjscore Formula (4)

Among them, adjscoreEvent is the target public opinion score corresponding to each first newsgroup, and adjscore is the original public opinion score of each first newsgroup.

Among them, the scaling ratio heat plays the role of an amplifier in formula (4), that is, the original public opinion score of each news event is enlarged or reduced. For example, if the scaling ratio is 1.5, the original public opinion score is enlarged by 1.5 times. Since the scaling ratio itself reflects the social attention to the news event, through the multiplicative fusion in formula (4), negative public opinion can be deducted more points, and positive public opinion can be added more points, so that the final calculated The target public opinion score is more reasonable and accurate.

105: Perform an ESG evaluation on the enterprise to be evaluated according to the target public opinion score of the news event corresponding to each first news group, and obtain an ESG index of the enterprise to be evaluated.

Exemplarily, the target public opinion score corresponding to each first news group may be used as an ESG index when ESG evaluation is performed on the enterprise to be evaluated from the news dimension.

106: Send the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.

It can be seen that in the implementation of the present application, after obtaining the news for ESG evaluation of the enterprise to be evaluated, instead of directly using the number of news to evaluate the ESG of the enterprise to be evaluated, the method of clustering is adopted to For the news groups belonging to each news event, ESG evaluation is carried out from the granularity of news events to achieve more accurate evaluation; then, for the news groups under each news event, ESG evaluation is not directly performed using the number of news in the news group, Instead, further clustering from the dimension of original news and reprinted news to evaluate companies for ESG evaluation, and different distinctions between original news and reprinted news can further improve the accuracy of ESG evaluation and make the obtained ESG index relatively high. In this way, ESG indexes with high precision can be sent to the target equipment, and the accuracy of the investment decision made can be improved.

In one embodiment of this application, in essence, according to the different roles of the target device, the decisions related to the enterprise to be evaluated are also different. The ESG index of this application has the following application scenarios:

Scenario 1: When the target equipment is the equipment of an investment institution, and the ESG index of the enterprise to be evaluated is sent to the investment institution, the investment institution can make a decision related to the enterprise to be evaluated: investment related to the enterprise to be evaluated decision making. For example, since the ESG index of an enterprise reflects the value and sustainable development capability of the enterprise, when the ESG index of the enterprise to be evaluated is high, the investment decision can be to increase the investment amount and investment cycle of the enterprise to be evaluated ; When the ESG index of the enterprise to be evaluated is low, the investment can definitely divest the investment in the enterprise to be evaluated or reduce the investment in the enterprise to be evaluated, and so on. In general, sending the ESG index of the enterprise to be evaluated to the investment institution can provide direction guidance for the investment decision-making of the investment institution and reduce investment risk.

Scenario 2: When the target equipment is the equipment of the enterprise to be evaluated, and the ESG index of the enterprise to be evaluated is sent to the enterprise to be evaluated, the enterprise to be evaluated can formulate a decision related to the enterprise to be evaluated as follows: management decisions. For example, since the ESG index of a company reflects the value and sustainable development capabilities of the company, as investors' acceptance of the ESG index deepens and they attach importance to corporate social responsibility, they are more willing to invest in higher ESG indexes. enterprise. Therefore, when the ESG index of the enterprise to be evaluated is high, the management decision makes a decision to strengthen corporate management for the enterprise to be evaluated, and continue to maintain excellent performance in ESG; when the ESG index of the enterprise to be evaluated is low, the management decision The decision is to adjust the company's development strategy, improve the company's sustainable development, and improve the ESG index. Generally speaking, sending the ESG index of the companies to be evaluated to the companies to be evaluated will help promote the companies to be evaluated to improve their own ESG evaluation and guide the healthy development of the companies to be evaluated.

Scenario 3: When the target equipment is the equipment of the government or social organization, the ESG index of the enterprise to be evaluated is sent to the government or social organization, then the government or social organization can make a decision related to the enterprise to be evaluated. Evaluate enterprise-related support decisions. Exemplarily, the ESG index of an enterprise reflects the value and sustainable development capability of the enterprise. Therefore, when the ESG index of the enterprise to be evaluated is high, it indicates that the development potential of the enterprise to be evaluated is relatively large, and the support decision can greatly promote the enterprise to be evaluated so as to provide more development opportunities for such enterprises; When the ESG index of the enterprise to be evaluated is low, it indicates that the development potential of the enterprise to be evaluated is low, and the support decision can be to order such an enterprise to adjust the company's development strategy, or reduce support to guide the enterprise to be evaluated to a healthy development direction Adjustment.

Referring to FIG. 2, FIG. 2 is a block diagram of functional units of an enterprise ESG index determination device provided in the embodiment of the present application. The enterprise ESG index determination device 200 includes: an acquisition unit 201, a processing unit 202 and a sending unit 203, wherein:

Acquisition unit 201 is used to obtain M pieces of news of the enterprise to be evaluated within the preset time period, wherein M is an integer greater than or equal to 1;

The processing unit 202 is configured to cluster the M pieces of news to obtain K first news groups, wherein each first news group in the K first news groups corresponds to a news event, and each of the K first news groups corresponds to a news event. A first news group includes one or more news articles in the M news articles, K is an integer greater than or equal to 1;

The sending unit 203 is configured to send the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.

In some possible implementations of the present application, according to an original news item and H reprinted news items included in each of the L second news groups, it is determined that each of the first news groups corresponds to In terms of the target public opinion scoring of news events, the processing unit 202 is specifically used for:

According to an original news included in each of the L second newsgroups, and the number H of reprinted news, determine the scaling ratio of the news event corresponding to each of the first newsgroups, the The scaling ratio of the news event corresponding to each first news group represents the degree of attention to the news event corresponding to each first news group;

According to the news content of an original news and the news content of H reprinted news included in each second news group in the L second news groups, determine the original news event corresponding to each of the first news groups Score of public opinion;

Perform product processing on the scaling ratio of the news event corresponding to each first news group and the original public opinion score to obtain the target public opinion score of the news event corresponding to each first news group.

In some possible implementations of the present application, according to an original news included in each second news group in the L second news groups, and the number H of reprinted news, determine the number of each first news In terms of the scaling ratio of the news event corresponding to the group, the processing unit 202 is specifically used for:

Obtaining a preset ratio of original news in each second news group, where the preset ratio represents the contribution of the original news to social attention to news events corresponding to each first news group Spend;

Determine the scaling ratio of each second news group according to the preset ratio of original news in each second news group and the number H of reprinted news included in each second news group;

The scaling ratios of the L second newsgroups are summed to obtain the scaling ratios of the news events corresponding to each of the first newsgroups.

In some possible implementations of the present application, according to the news content of an original news and the news content of H reprinted news included in each of the L second news groups, it is determined that each In terms of the original public opinion score of the news event corresponding to the first news group, the processing unit 202 is specifically used for:

Perform emotion recognition on the news in each first news group to obtain an emotion tag corresponding to each first news group, wherein the emotion tag corresponding to each first news group is used to represent each The news event corresponding to the first news group is a positive news event or a negative news event;

Acquiring the publishing media of each news item in each of the first newsgroups;

determining the highest-level publishing media in each of the first newsgroups according to the publishing media of each news in each of the first newsgroups;

According to the mapping relationship between published media, emotional tags and public opinion scores, and the highest-level published media in each first news group and the corresponding emotional tags of each first news group, determine each The original public opinion score of the news event corresponding to the first news group.

In some possible implementations of the present application, in terms of performing emotion recognition on the news in each of the first newsgroups to obtain the emotion tags corresponding to each of the first newsgroups, the processing unit 202 is specifically configured to:

Randomly select a second news group from the L second news groups, and carry out emotional recognition to any news in the second news group, obtain the emotional label with the news, and use the emotional label of the news The label is used as the sentiment label corresponding to each of the first newsgroups;

or,

Randomly select a piece of news from one or more pieces of news included in each of the first newsgroups, and carry out emotion recognition to the piece of news, obtain the emotion label corresponding to the piece of news, and the emotion corresponding to the piece of news The label serves as the sentiment label corresponding to each first newsgroup.

In some possible implementation manners of the present application, in terms of clustering the M pieces of news to obtain K first newsgroups, the processing unit 202 is specifically configured to:

Performing semantic information extraction on each of the M pieces of news to obtain a semantic vector of each of the news, wherein the semantic vector of each of the news is used to represent the news events described in each of the news;

determining the first similarity between the semantic vectors of any two news in the M news;

According to the first similarity between the semantic vectors of any two news in the M news, the M news are clustered to obtain K first news groups, wherein, among the K first news groups The first similarity between the semantic vectors of any two news articles in each first news group is greater than the first threshold.

In some possible implementation manners of the present application, in terms of clustering one or more news articles in each first news group to obtain L second news groups corresponding to each first news group , the processing unit 202 is specifically used for:

determining a second similarity between any two news articles in each first news group, wherein the second similarity between any two news articles in each first news group is used to characterize each The similarity between the news content of any two news articles in the first news group;

According to the second similarity between any two news articles in each first news group, one or more news articles included in each first news group are clustered to obtain L second news groups, Wherein, the second similarity between any two news articles in each second news group is greater than a second threshold.

Referring to FIG. 3 , FIG. 3 is a schematic structural diagram of an electronic device provided in an embodiment of the present application. As shown in FIG. 3 , an electronic device 300 includes a transceiver 301 , a processor 302 and a memory 303 . They are connected through a bus 304 . The memory 303 is used to store computer programs and data, and can transmit the data stored in the memory 303 to the processor 302 .

The processor 302 is used to read the computer program in the memory 303 to perform the following operations:

Controlling the transceiver 301 to obtain M pieces of news about the enterprise to be evaluated within a preset time period, where M is an integer greater than or equal to 1;

The control transceiver 301 sends the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device can make an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.

In some possible implementations of the present application, according to an original news item and H reprinted news items included in each of the L second news groups, it is determined that each of the first news groups corresponds to In terms of the target public opinion scoring of news events, the processor 302 is specifically used for:

In some possible implementations of the present application, according to an original news included in each second news group in the L second news groups, and the number H of reprinted news, determine the number of each first news Regarding the scaling ratio of the news event corresponding to the group, the processor 302 is specifically configured to perform the following operations:

In some possible implementations of the present application, according to the news content of an original news and the news content of H reprinted news included in each of the L second news groups, it is determined that each In terms of the original public opinion score of the news event corresponding to the first news group, the processor 302 is specifically configured to perform the following operations:

In some possible implementations of the present application, in terms of performing emotion recognition on the news in each of the first newsgroups to obtain the emotion tags corresponding to each of the first newsgroups, the processor 302 is specifically configured to perform the following operate:

or,

Randomly select a piece of news from one or more pieces of news included in each of the first newsgroups, and carry out emotion recognition to the piece of news, obtain the emotion label corresponding to the piece of news, and the emotion corresponding to the piece of news The label serves as the sentiment label corresponding to each first news group.

In some possible implementations of the present application, in terms of clustering the M pieces of news to obtain K first newsgroups, the processor 302 is specifically configured to perform the following operations:

Carry out semantic information extraction to each piece of news in the M pieces of news, obtain the semantic vector of each piece of news, wherein, the semantic vector of each piece of news is used to characterize the news event described in each piece of news;

In some possible implementation manners of the present application, in terms of clustering one or more news articles in each first news group to obtain L second news groups corresponding to each first news group , the processor 302 is specifically configured to perform the following operations:

Specifically, the above-mentioned transceiver 301 can be the acquisition unit 201 and the sending unit 203 of the enterprise ESG index determination device 200 of the embodiment shown in FIG. 2 , and the above-mentioned processor 302 can be the enterprise ESG index determination The processing unit 202 of the device 200 .

The embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement any clustering-based Some or all of the steps in the corporate ESG index determination methodology for technology.

Optionally, the storage medium involved in this application, such as a computer-readable storage medium, may be non-volatile or volatile.

The embodiment of the present application also provides a computer program product, the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable the computer to execute the method described in the above method embodiments Part or all of the steps of any method for determining an enterprise ESG index based on clustering technology.

It should be noted that for the foregoing method embodiments, for the sake of simple description, they are expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the described action sequence. Depending on the application, certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the application.

In the foregoing embodiments, the descriptions of each embodiment have their own emphases, and for parts not described in detail in a certain embodiment, reference may be made to relevant descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or can be Integrate into another system, or some features may be ignored, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented not only in the form of hardware, but also in the form of software program modules.

The integrated units may be stored in a computer-readable memory if implemented in the form of a software program module and sold or used as an independent product. Based on this understanding, the technical solution of the present application is essentially or part of the contribution to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory. Several instructions are included to make a computer device (which may be a personal computer, server or network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes such as U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk.

Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above-mentioned embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, and the memory can include: a flash disk , Read-only memory (English: Read-Only Memory, referred to as: ROM), random access device (English: Random Access Memory, referred to as: RAM), magnetic disk or optical disc, etc.

The embodiments of the present application have been introduced in detail above, and specific examples have been used in this paper to illustrate the principles and implementation methods of the present application. The descriptions of the above embodiments are only used to help understand the methods and core ideas of the present application; meanwhile, for Those skilled in the art will have changes in specific implementation methods and application scopes based on the ideas of the present application. In summary, the contents of this specification should not be construed as limiting the present application.

Claims

A method for determining an enterprise ESG index based on clustering technology, including:

Obtain M pieces of news about the enterprise to be evaluated within a preset time period, where M is an integer greater than or equal to 1;

Clustering the M pieces of news to obtain K first newsgroups, wherein each first newsgroup in the K first newsgroups corresponds to a news event, and each first newsgroup includes One or more news articles in the M news articles, K is an integer greater than or equal to 1;

Clustering one or more news articles in each of the first newsgroups to obtain L second newsgroups corresponding to each of the first newsgroups, wherein each of the second newsgroups includes One piece of original news and H pieces of reprinted news corresponding to the original news, where H is an integer greater than or equal to 0;

According to an original news and H pieces of reposted news included in each of the L second newsgroups, determine the target public opinion score of the news event corresponding to each of the first newsgroups, the target The public opinion score is used to characterize the influence of the news event corresponding to each first news group on the enterprise to be evaluated;

According to the target public opinion score of the news event corresponding to each first news group, ESG evaluation is performed on the enterprise to be evaluated, and the ESG index of the enterprise to be evaluated is obtained;

Sending the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.
The method according to claim 1, wherein, according to an original news and H reprinted news included in each second news group in the L second news groups, it is determined that each first news The target public opinion score of the news event corresponding to the group, including:

According to an original news included in each of the L second newsgroups, and the number H of reprinted news, determine the scaling ratio of the news event corresponding to each of the first newsgroups, the The scaling ratio of the news event corresponding to each first news group represents the degree of social attention to the news event corresponding to each first news group;

According to the news content of an original news and the news content of H reprinted news included in each second news group in the L second news groups, determine the original news event corresponding to each of the first news groups Score of public opinion;

Perform product processing on the scaling ratio of the news event corresponding to each first news group and the original public opinion score to obtain the target public opinion score of the news event corresponding to each first news group.
The method according to claim 2, wherein, according to an original news included in each second news group in the L second news groups, and the number H of reprinted news, it is determined that each of the second news groups The scaling ratio of news events corresponding to a newsgroup, including:

Obtaining a preset ratio of original news in each second news group, where the preset ratio represents the contribution of the original news to social attention to news events corresponding to each first news group Spend;

Determine the scaling ratio of each second news group according to the preset ratio of original news in each second news group and the number H of reprinted news included in each second news group;

The scaling ratios of the L second newsgroups are summed to obtain the scaling ratios of the news events corresponding to each of the first newsgroups.
The method according to claim 2 or 3, wherein, according to the news content of an original news and the news content of H reprinted news included in each second news group in the L second news groups, Determining the original public opinion score of the news event corresponding to each first news group includes:

Perform emotion recognition on the news in each first news group to obtain an emotion tag corresponding to each first news group, wherein the emotion tag corresponding to each first news group is used to represent each The news event corresponding to the first news group is a positive news event or a negative news event;

Acquiring the publishing media of each news item in each of the first newsgroups;

determining the highest-level publishing media in each of the first newsgroups according to the publishing media of each news in each of the first newsgroups;

According to the mapping relationship between published media, emotional tags and public opinion scores, and the highest-level published media in each first news group and the corresponding emotional tags of each first news group, determine each The original public opinion score of the news event corresponding to the first news group.
The method according to claim 4, wherein said carrying out emotion recognition to the news in each of the first newsgroups to obtain the corresponding emotion tags of each of the first newsgroups comprises:

Randomly select a second news group from the L second news groups, and carry out emotional recognition to any news in the second news group, obtain the emotional label with the news, and use the emotional label of the news The label is used as the sentiment label corresponding to each of the first newsgroups;

or,

Randomly select a piece of news from one or more pieces of news included in each of the first newsgroups, and carry out emotion recognition to the piece of news, obtain the emotion label corresponding to the piece of news, and the emotion corresponding to the piece of news The label serves as the sentiment label corresponding to each first news group.
The method according to claim 5, wherein said clustering said M pieces of news to obtain K first newsgroups comprises:

Performing semantic information extraction on each of the M pieces of news to obtain a semantic vector of each of the news, wherein the semantic vector of each of the news is used to represent the news events described in each of the news;

determining the first similarity between the semantic vectors of any two news in the M news;

According to the first similarity between the semantic vectors of any two news in the M news, the M news are clustered to obtain K first news groups, wherein, among the K first news groups The first similarity between the semantic vectors of any two news articles in each first news group is greater than the first threshold.
The method according to claim 6, wherein said clustering one or more news articles in each first news group obtains L second news articles corresponding to each first news group group, including:

determining a second similarity between any two news articles in each first news group, wherein the second similarity between any two news articles in each first news group is used to characterize each The similarity between the news content of any two news articles in the first news group;

According to the second similarity between any two news articles in each first news group, one or more news articles included in each first news group are clustered to obtain L second news groups, Wherein, the second similarity between any two news articles in each second news group is greater than a second threshold.
A device for determining an enterprise ESG index, comprising:

The obtaining unit is used to obtain M pieces of news of the enterprise to be evaluated within a preset time period, where M is an integer greater than or equal to 1;

A processing unit, configured to cluster the M pieces of news to obtain K first news groups, wherein each first news group in the K first news groups corresponds to a news event, and each of the K first news groups corresponds to a news event. The first news group includes one or more news articles in the M news articles, K is an integer greater than or equal to 1;

Clustering one or more news articles in each of the first newsgroups to obtain L second newsgroups corresponding to each of the first newsgroups, wherein each of the second newsgroups includes One piece of original news and H pieces of reprinted news corresponding to the original news, where H is an integer greater than or equal to 0;

According to an original news and H pieces of reposted news included in each of the L second newsgroups, determine the target public opinion score of the news event corresponding to each of the first newsgroups, the target The public opinion score is used to characterize the influence of the news event corresponding to each first news group on the enterprise to be evaluated;

According to the target public opinion score of the news event corresponding to each first news group, ESG evaluation is performed on the enterprise to be evaluated, and the ESG index of the enterprise to be evaluated is obtained;

The sending unit is configured to send the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.
An electronic device, comprising: a processor and a memory, the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the The electronic device implements the following methods:

Obtain M pieces of news about the enterprise to be evaluated within a preset time period, where M is an integer greater than or equal to 1;

Clustering the M pieces of news to obtain K first newsgroups, wherein each first newsgroup in the K first newsgroups corresponds to a news event, and each first newsgroup includes One or more news articles in the M news articles, K is an integer greater than or equal to 1;

Clustering one or more news articles in each of the first newsgroups to obtain L second newsgroups corresponding to each of the first newsgroups, wherein each of the second newsgroups includes One piece of original news and H pieces of reprinted news corresponding to the original news, where H is an integer greater than or equal to 0;

According to an original news and H pieces of reposted news included in each of the L second newsgroups, determine the target public opinion score of the news event corresponding to each of the first newsgroups, the target The public opinion score is used to characterize the influence of the news event corresponding to each first news group on the enterprise to be evaluated;

According to the target public opinion score of the news event corresponding to each first news group, ESG evaluation is performed on the enterprise to be evaluated, and the ESG index of the enterprise to be evaluated is obtained;

Sending the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.
The electronic device according to claim 9, wherein, performing said one piece of original news and H pieces of reposted news included in each second newsgroup in said L second newsgroups, determine said each first newsgroup The target public opinion score of a news event corresponding to a news group includes:

According to an original news included in each of the L second newsgroups, and the number H of reprinted news, determine the scaling ratio of the news event corresponding to each of the first newsgroups, the The scaling ratio of the news event corresponding to each first news group represents the degree of social attention to the news event corresponding to each first news group;

According to the news content of an original news and the news content of H reprinted news included in each second news group in the L second news groups, determine the original news event corresponding to each of the first news groups Score of public opinion;

Perform product processing on the scaling ratio of the news event corresponding to each first news group and the original public opinion score to obtain the target public opinion score of the news event corresponding to each first news group.
The electronic device according to claim 10, wherein performing said determining said each news according to an original news included in each of said L second newsgroups and the number H of reprinted news The zoom ratio of the news event corresponding to the first newsgroup, including:

Obtaining a preset ratio of original news in each second news group, where the preset ratio represents the contribution of the original news to social attention to news events corresponding to each first news group Spend;

Determine the scaling ratio of each second news group according to the preset ratio of original news in each second news group and the number H of reprinted news included in each second news group;

The scaling ratios of the L second newsgroups are summed to obtain the scaling ratios of the news events corresponding to each of the first newsgroups.
The electronic device according to claim 10 or 11, wherein, executing the news content of an original news and the news content of H reprinted news according to each second news group in the L second news groups. Content, determining the original public opinion score of the news event corresponding to each first news group, including:

Perform emotion recognition on the news in each first news group to obtain an emotion tag corresponding to each first news group, wherein the emotion tag corresponding to each first news group is used to represent each The news event corresponding to the first news group is a positive news event or a negative news event;

Acquiring the publishing media of each news item in each of the first newsgroups;

According to the publishing media of each piece of news in each of the first newsgroups, determine the highest-level publishing media in each of the first newsgroups;

According to the mapping relationship between published media, emotional tags and public opinion scores, and the highest-level published media in each first news group and the corresponding emotional tags of each first news group, determine each The original public opinion score of the news event corresponding to the first news group.
The electronic device according to claim 12, wherein performing the emotion recognition on the news in each of the first newsgroups to obtain the emotion tags corresponding to each of the first newsgroups includes:

Randomly select a second news group from the L second news groups, and carry out emotional recognition to any news in the second news group, obtain the emotional label with the news, and use the emotional label of the news The label is used as the sentiment label corresponding to each of the first newsgroups;

or,

Randomly select a piece of news from one or more pieces of news included in each of the first newsgroups, and carry out emotion recognition to the piece of news, obtain the emotion label corresponding to the piece of news, and the emotion corresponding to the piece of news The label serves as the sentiment label corresponding to each first news group.
The electronic device according to claim 13, wherein performing the clustering of the M pieces of news to obtain K first newsgroups, comprising:

Performing semantic information extraction on each of the M pieces of news to obtain a semantic vector of each of the news, wherein the semantic vector of each of the news is used to represent the news events described in each of the news;

determining the first similarity between the semantic vectors of any two news in the M news;

According to the first similarity between the semantic vectors of any two news in the M news, the M news are clustered to obtain K first news groups, wherein, among the K first news groups The first similarity between the semantic vectors of any two news articles in each first news group is greater than the first threshold.
A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the following method:

Obtain M pieces of news about the enterprise to be evaluated within a preset time period, where M is an integer greater than or equal to 1;

Clustering the M pieces of news to obtain K first newsgroups, wherein each first newsgroup in the K first newsgroups corresponds to a news event, and each first newsgroup includes One or more news articles in the M news articles, K is an integer greater than or equal to 1;

Clustering one or more news articles in each of the first newsgroups to obtain L second newsgroups corresponding to each of the first newsgroups, wherein each of the second newsgroups includes One piece of original news and H pieces of reprinted news corresponding to the original news, where H is an integer greater than or equal to 0;

According to an original news and H pieces of reposted news included in each of the L second newsgroups, determine the target public opinion score of the news event corresponding to each of the first newsgroups, the target The public opinion score is used to characterize the influence of the news event corresponding to each first news group on the enterprise to be evaluated;

According to the target public opinion score of the news event corresponding to each first news group, ESG evaluation is performed on the enterprise to be evaluated, and the ESG index of the enterprise to be evaluated is obtained;

Sending the ESG index of the enterprise to be evaluated to the target device, so that the user of the target device makes an investment decision related to the enterprise to be evaluated according to the ESG index of the enterprise to be evaluated.
The computer-readable storage medium according to claim 15, wherein, performing the said one piece of original news and H pieces of reprinted news included in each of said L second newsgroups, determining said The target public opinion score of the news event corresponding to each first news group, including:

According to an original news included in each of the L second newsgroups, and the number H of reprinted news, determine the scaling ratio of the news event corresponding to each of the first newsgroups, the The scaling ratio of the news event corresponding to each first news group represents the degree of social attention to the news event corresponding to each first news group;

According to the news content of an original news and the news content of H reposted news included in each second news group in the L second news groups, determine the original news event corresponding to each first news group Score of public opinion;

Perform product processing on the scaling ratio of the news event corresponding to each first news group and the original public opinion score to obtain the target public opinion score of the news event corresponding to each first news group.
The computer-readable storage medium according to claim 16, wherein, performing said determining according to an original news included in each of the L second newsgroups and the number H of reprinted news The scaling ratio of the news event corresponding to each first news group includes:

Obtaining a preset ratio of original news in each second news group, where the preset ratio represents the contribution of the original news to social attention to news events corresponding to each first news group Spend;

Determine the scaling ratio of each second news group according to the preset ratio of original news in each second news group and the number H of reprinted news included in each second news group;

The scaling ratios of the L second newsgroups are summed to obtain the scaling ratios of the news events corresponding to each of the first newsgroups.
The computer-readable storage medium according to claim 16 or 17, wherein, performing the news content of an original news and the reprinting of H articles according to each second newsgroup in the L second newsgroups The news content of the news determines the original public opinion score of the news event corresponding to each first news group, including:

Perform emotion recognition on the news in each first news group to obtain an emotion tag corresponding to each first news group, wherein the emotion tag corresponding to each first news group is used to represent each The news event corresponding to the first news group is a positive news event or a negative news event;

Acquiring the publishing media of each news item in each of the first newsgroups;

determining the highest-level publishing media in each of the first newsgroups according to the publishing media of each news in each of the first newsgroups;

According to the mapping relationship between published media, emotional tags and public opinion scores, and the highest-level published media in each first news group and the corresponding emotional tags of each first news group, determine each The original public opinion score of the news event corresponding to the first news group.
The computer-readable storage medium according to claim 18, wherein performing the emotion recognition on the news in each of the first newsgroups to obtain the emotion tags corresponding to each of the first newsgroups includes:

Randomly select a second news group from the L second news groups, and carry out emotional recognition to any news in the second news group, obtain the emotional label with the news, and use the emotional label of the news The label is used as the sentiment label corresponding to each of the first newsgroups;

or,

Randomly select a piece of news from one or more pieces of news included in each of the first newsgroups, and carry out emotion recognition to the piece of news, obtain the emotion label corresponding to the piece of news, and the emotion corresponding to the piece of news The label serves as the sentiment label corresponding to each first newsgroup.
The computer-readable storage medium according to claim 19, wherein performing the clustering of the M articles of news obtains K first newsgroups, comprising:

Performing semantic information extraction on each of the M pieces of news to obtain a semantic vector of each of the news, wherein the semantic vector of each of the news is used to represent the news events described in each of the news;

determining the first similarity between the semantic vectors of any two news in the M news;

According to the first similarity between the semantic vectors of any two news in the M news, the M news are clustered to obtain K first news groups, wherein, among the K first news groups The first similarity between the semantic vectors of any two news articles in each first news group is greater than the first threshold.