Automated ESG Evaluation Using News Clustering and Vectorization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Companies face challenges in evaluating risks associated with environmental, social, and governance (ESG) issues using non-financial data, as existing methods are subjective and slow, and automated classification and evaluation of news articles are difficult due to the lack of formalized data.
Innovation Solution
A device and method for collecting and analyzing news articles on the Internet, using morpheme analysis, vectorization, similarity analysis, and machine learning algorithms to classify and evaluate ESG risks, categorize news articles into environment, social, and governance issues, and calculate an ESG company evaluation score.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual ESG evaluation reports are written by companies, then evaluation depth and context understanding are improved, but evaluation speed and productivity deteriorate
Solution Approach 1:
The patent replaces manual mechanical analysis with an automated system combining NLP, machine learning, and knowledge graphs. The system automatically extracts ESG-related information from news articles and reports, classifies them into categories (environmental, social, governance), and generates evaluation scores without human intervention, thus maintaining accuracy while dramatically improving evaluation speed
Solution Approach 2:
The patent introduces an intermediary automated processing system that acts as a bridge between raw news data and final ESG evaluations. This intermediary system uses natural language processing to understand and structure unstructured data, then feeds processed information to the evaluation module, resolving the contradiction between maintaining quality and improving efficiency
2Productivity
If automated classification of news articles is implemented, then evaluation speed is improved, but classification accuracy deteriorates due to lack of formalized data
Solution Approach 1:
The patent applies preliminary action by pre-processing news articles through morphological analysis, vectorization, and similarity analysis before classification. The system pre-structures unstructured data into standardized formats, creating a foundation that enables accurate automated classification. This preliminary structuring resolves the accuracy issue while maintaining automation speed
Solution Approach 2:
The patent transforms the parameters of news data from unstructured text to structured vectors through morphological analysis and similarity calculations. By changing the data representation parameters, the system enables machine learning algorithms to process and classify information accurately, overcoming the limitation of non-formalized data while maintaining high automation speed
3Reliability
If comprehensive ESG analysis is performed on all news articles, then evaluation reliability is improved, but processing time and resource consumption increase
Solution Approach 1:
The patent extracts only ESG-relevant information from news articles using keyword matching, entity recognition, and topic modeling. Instead of analyzing entire articles, the system extracts specific ESG-related segments (environmental impacts, social responsibilities, governance practices), maintaining comprehensive evaluation reliability while significantly reducing processing time through selective extraction
Solution Approach 2:
The patent applies partial action by focusing analysis on the most critical ESG factors and high-impact news articles. The system uses similarity analysis to identify and prioritize relevant articles, performing detailed analysis only on those that significantly impact ESG scores. This selective approach maintains evaluation reliability while reducing overall processing time and resource consumption
Data Source
AI summary
Provided is an Environment, Social, Governance (ESG) company evaluation device for evaluating a company from the viewpoint of ESG to calculate a score. The device includes: a news collection unit configured to collect a plurality of news articles on Internet and classify them by date or company, and perform clustering on news articles whose similarity is higher than a reference value through a similarity analysis between the news articles; a news classification unit configured to classify each of the news articles related to an issue of environment, society, or governance; and an evaluation result deriving unit configured to calculate an ESG risk for a corresponding cluster in cluster units and calculate an ESG company evaluation score based on the calculated value.