Method, device and equipment for automatically generating esg recommendation report and storage medium
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- MACH INNOVATION TECHNOLOGY (SHENZHEN) CO LTD
- Filing Date
- 2025-03-19
- Publication Date
- 2026-06-02
Smart Images
Figure CN120218022B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, device, and storage medium for automatically generating ESG recommendation reports. Background Technology
[0002] In today's era of rapid information technology development, all industries face the need for processing and analyzing massive amounts of data. Whether it's internal corporate operations management and market research, or academic research and policy formulation in scientific research institutions, all rely on the collection, organization, and in-depth analysis of large amounts of documents and data. These processes often involve documents in various formats (such as documents, tables, and images) and complex and diverse data sources, requiring the extraction of key information for summarization, comparison, and analysis to support decision-making and report writing.
[0003] However, due to the high complexity of data processing technology, most existing solutions rely on highly manual processes to handle these tasks. Staff need to manually open and read documents, input useful information into specialized databases or spreadsheets, and then use various independent software tools to integrate, clean, and analyze the data. This series of steps is not only tedious but also heavily dependent on personal experience and judgment. Furthermore, at the end of data processing, report writing and related recommendations often rely on manual synthesis of all analytical results, which is not only time-consuming and labor-intensive but may also lead to biased conclusions due to subjective factors.
[0004] Therefore, how to improve the efficiency and accuracy of data processing, reduce human intervention, avoid bias caused by subjective human judgment, and provide efficient and intelligent data analysis and decision support for the industry has become an important issue that urgently needs to be addressed. Summary of the Invention
[0005] In view of this, embodiments of this application provide a method, apparatus, computer device, and storage medium for automatically generating ESG recommendation reports, which can effectively solve the core problems of the existing technology that adopts a hybrid mode of "manual + tool", such as high human resource consumption, high error rate, slow processing speed, and excessive reliance on manual labor.
[0006] In a first aspect, embodiments of this application provide a method for automatically generating an ESG recommendation report, including:
[0007] Process the company's raw data to generate an initial ESG report;
[0008] The initial ESG report is input into a generative intelligent model for analysis to obtain ESG recommendation topics.
[0009] The ESG recommendation topics are analyzed and processed to obtain search results;
[0010] The generative intelligent model is used to perform strategy analysis on the search results to obtain guiding strategies;
[0011] The guidance strategy is optimized using a pre-defined evaluation model to generate an ESG recommendation report.
[0012] In some embodiments, the processing of raw enterprise data and analysis in conjunction with a lifecycle assessment database to generate an initial ESG report includes:
[0013] Filter the raw data of the enterprises according to business type, geographical location and time range to obtain enterprise data that meets the criteria;
[0014] The qualified enterprise data is populated into the preset initial ESG report template, and relevant data from the life cycle assessment database is referenced during the population process;
[0015] The initial ESG report is generated by calculating the data of eligible enterprises using the preset formula in the initial ESG report template.
[0016] In some embodiments, the step of inputting the initial ESG report into a generative intelligent model for analysis to obtain ESG recommendation topics includes:
[0017] The step of inputting the initial ESG report into a generative intelligent model for analysis to obtain ESG recommendation topics includes:
[0018] The initial ESG report is input into a generative intelligent model for analysis, generating guidance and suggestions on the company's ESG status.
[0019] The guidance and recommendations on the enterprise's ESG status are processed to obtain ESG recommendation topics.
[0020] In some embodiments, the analysis and processing of the ESG suggestion topics to obtain search results includes:
[0021] The ESG recommendation topics are analyzed to generate topic embeddings corresponding to the ESG recommendation topics;
[0022] The topic is embedded as an index value and merged with the index values pre-generated in the lifecycle assessment database and the index values pre-generated in the enterprise's original data to form an embedding vector. This vector is then searched to obtain search results related to the ESG recommendation topic.
[0023] In some embodiments, the step of using the generative intelligent model to perform strategy analysis on the search results to obtain a guiding strategy includes:
[0024] Search results related to the ESG recommendation topics are input into the generative intelligent model for analysis to obtain the guidance strategies expressed through professional tone and vocabulary.
[0025] In some embodiments, optimizing the guidance strategy using a preset evaluation model to generate an ESG recommendation report includes:
[0026] The guidance strategy, expressed through professional tone and vocabulary, is optimized using a pre-defined evaluation model to obtain a quality score for the guidance strategy.
[0027] When the quality score is lower than a preset threshold, the parameters of the generative intelligent model are adjusted to regenerate the guidance strategy until the quality score of the guidance strategy is higher than or equal to the preset threshold.
[0028] In some embodiments, the step of adjusting the parameters of the generative intelligent model to regenerate the guidance policy when the quality score is lower than a preset threshold, until the quality score of the guidance policy is higher than or equal to the preset threshold, further includes:
[0029] The guidance strategy that is higher than or equal to the preset threshold is converted according to a preset format to generate an ESG recommendation report.
[0030] Secondly, embodiments of this application provide an automatic ESG recommendation report generation apparatus, comprising:
[0031] The initial report generation module is used to process the company's raw data and analyze it in conjunction with data from the life cycle assessment database to generate an initial ESG report;
[0032] The topic generation module is used to input the initial ESG report into the generative intelligent model for analysis and obtain ESG suggested topics.
[0033] The search result acquisition module is used to analyze and process the ESG suggestion topics to obtain search results;
[0034] The guidance strategy generation module is used to perform strategy analysis on the search results using the generative intelligent model to obtain guidance strategies.
[0035] The recommendation report generation module is used to optimize the guidance strategy using a preset evaluation model and generate an ESG recommendation report.
[0036] Thirdly, embodiments of this application provide a computer device, the computer device including a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the automatic generation method of the ESG recommendation report described in the first aspect above.
[0037] Fourthly, embodiments of this application provide a computer-readable storage medium, wherein when the computer program is executed on a processor, it implements the method for automatically generating ESG recommendation reports as described in the first aspect.
[0038] The embodiments of this application have the following beneficial effects:
[0039] This application discloses an automated method, apparatus, computer device, and storage medium for generating ESG recommendation reports, significantly improving enterprises' management efficiency and decision-making quality in environmental, social, and governance (ESG) aspects. First, the company's raw data is efficiently processed and combined with data from a lifecycle assessment database to generate a detailed initial ESG report. This process not only reduces the time and manpower costs of manual data processing but also ensures the comprehensiveness and accuracy of the data. Next, the initial ESG report is input into a generative intelligent model for in-depth analysis, extracting key ESG recommendation themes. This step utilizes advanced natural language processing technology to quickly identify and focus on the company's core improvement points in the ESG field. Subsequently, the extracted ESG recommendation themes are further analyzed and processed to generate relevant search results, providing abundant reference materials and industry best practices. The generative intelligent model uses these search results for strategy analysis to generate practical guidance strategies, thus providing strong support for enterprises to formulate scientific ESG improvement plans. Finally, a pre-set evaluation model optimizes the guidance strategies, ensuring that the proposed strategies are not only actionable but also maximize the improvement of the company's ESG performance, ultimately generating a comprehensive and targeted ESG recommendation report. The method described in this application not only improves the efficiency and accuracy of data processing, reduces human intervention, and avoids bias caused by subjective judgment, but also provides enterprises with efficient and intelligent ESG data analysis and decision support, helping them achieve greater breakthroughs and progress on the path of sustainable development. Attached Figure Description
[0040] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0041] Figure 1 This illustration shows an application scenario of an automatic ESG recommendation report generation method according to an embodiment of this application;
[0042] Figure 2 A flowchart illustrating an embodiment of the present application of an automatic ESG recommendation report generation method is shown.
[0043] Figure 3 Another flowchart of an automatic generation method for an ESG recommendation report according to an embodiment of this application is shown;
[0044] Figure 4 The diagram illustrates an initial ESG report from an embodiment of the present application of an automatic ESG recommendation report generation method.
[0045] Figure 5 Another flowchart of an automatic generation method for an ESG recommendation report according to an embodiment of this application is shown;
[0046] Figure 6 Another flowchart of an automatic generation method for an ESG recommendation report according to an embodiment of this application is shown;
[0047] Figure 7 Another flowchart of an automatic generation method for an ESG recommendation report according to an embodiment of this application is shown;
[0048] Figure 8 This illustration shows a diagram illustrating the power consumption of a building material supplier in an automatic generation method for an ESG recommendation report according to an embodiment of this application.
[0049] Figure 9 A schematic diagram of an automatic ESG recommendation report generation device according to an embodiment of this application is shown. Detailed Implementation
[0050] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.
[0051] The components of the embodiments of this application described and illustrated in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0052] In the following text, the terms "comprising," "having," and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more combinations thereof. Furthermore, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.
[0053] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.
[0054] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0055] Considering that existing technologies employ a hybrid "manual + tool" approach, which suffers from core problems such as high human resource consumption, high error rates, slow processing speed, and over-reliance on manual labor, this paper proposes an automated method for generating ESG recommendation reports. This method automatically filters raw enterprise data to obtain an initial ESG report. Subsequently, the initial ESG report is input into a generative intelligent model to extract ESG recommendation topics. Next, the ESG recommendation topics are analyzed to obtain search results. The generative intelligent model then conducts in-depth analysis of the search results to derive guidance strategies. Finally, an evaluation model is used to optimize the guidance strategies, ultimately generating a professional and authoritative ESG recommendation report. This method not only significantly improves the efficiency of initial ESG report analysis and processing but also helps enterprises gain a deeper understanding of their ESG performance, providing strong support for the formulation and implementation of sustainable development strategies and helping enterprises better address ESG challenges and achieve sustainable development goals.
[0056] This application provides an embodiment of an automatic ESG recommendation report generation method, which can be applied to applications such as... Figure 1 In the application environment, specifically, the method for automatically generating ESG recommendation reports of this application is applied in a computer system, which includes, for example, Figure 1The diagram illustrates a client and server. The client communicates with the server over a network. The client, also known as the user terminal, is the program that provides local services to the client, corresponding to the server. Clients can be installed on, but are not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be a standalone server or a server cluster consisting of multiple servers. Specifically, the server first processes the enterprise's raw data and analyzes it in conjunction with data from the lifecycle assessment database to generate an initial ESG report. This initial ESG report is then transmitted to a generative intelligent model on the server for in-depth analysis to extract ESG recommendation topics. The extracted ESG recommendation topics are then further analyzed on the server to obtain relevant search results. These search results are again analyzed by the generative intelligent model to generate guidance strategies. Finally, a pre-defined assessment model optimizes these guidance strategies to generate the final ESG recommendation report. Throughout this process, the client is primarily responsible for user interaction, such as data uploading, report viewing, and strategy feedback, while complex data processing, intelligent analysis, and report generation are efficiently handled by the server. By rationally integrating the functions of the client and server, not only is the efficiency and accuracy of data processing improved, and human intervention reduced, avoiding biases caused by subjective human judgment, but enterprises are also provided with efficient and intelligent ESG data analysis and decision support, significantly enhancing their management level and sustainable development capabilities in environmental, social and governance (ESG) aspects.
[0057] Figure 2 A flowchart illustrating an automatic ESG recommendation report generation method according to an embodiment of this application is shown. The automatic ESG recommendation report generation method includes the following steps:
[0058] Step S100: Process the enterprise's raw data and analyze it in conjunction with the life cycle assessment database to generate an initial ESG report.
[0059] For example, raw enterprise data refers to datasets that have undergone preliminary screening and organization, and conform to industry standards and internal enterprise standards. These datasets encompass various types of enterprise information, such as production activity records, energy consumption details, waste emission statistics, employee welfare status, and corporate governance structure. After obtaining the raw enterprise data, a series of processing steps are required: first, data cleaning to remove errors and redundant information; second, data transformation to ensure data format consistency and standardization; and finally, data analysis to uncover the deeper meaning and value behind the data. After this series of processing steps, standardized enterprise data is obtained, and this standardized enterprise data is then used to populate a pre-designed initial ESG report template to generate an initial ESG report.
[0060] The Life Cycle Assessment (LCA) database aims to comprehensively assess the environmental impact of products throughout their entire lifecycle, from raw material acquisition and manufacturing to use and disposal. The LCA database covers key information such as environmental impact assessment data for various raw materials and energy consumption and emissions in different production processes. By combining and applying data from the LCA database, companies can more deeply assess the comprehensive environmental impact of their operations, thereby further improving the comprehensiveness and accuracy of their initial ESG reports.
[0061] An initial ESG report is an important document for assessing a company's performance across the three dimensions of Environmental, Social, and Governance. Through an initial ESG report, companies can clearly demonstrate their positive actions in environmental protection, fulfilling social responsibility, and improving corporate governance structures, thereby enhancing investor confidence and improving the company's social image and brand value.
[0062] In one alternative embodiment, such as Figure 3 As shown, in step S100, the enterprise's original data is processed, and the enterprise's original data is analyzed in conjunction with the life cycle assessment database to generate an initial ESG report, including:
[0063] Step S110: Filter the original enterprise data according to business type, geographical location and time range to obtain enterprise data that meets the criteria.
[0064] For example, raw enterprise data refers to data extracted from documents such as financial reports, invoices, and receipts. Data filtering involves several dimensions: First, filtering is based on the enterprise's main business or industry classification, such as manufacturing, services, new energy, and environmental protection. Second, filtering is based on the enterprise's geographical location, including country, region, city, or more specific geographical locations (e.g., industrial parks, commercial districts). Finally, filtering is based on the data's time attributes, covering enterprise data from the past year, the past quarter, or a specific time period. In other words, raw enterprise data is filtered based on three dimensions: business type, geographical location, and time range, to obtain data rows, columns, and specific values that are directly relevant to the initial ESG report and meet specific conditions.
[0065] Step S120: Fill the qualified enterprise data into the preset initial ESG report template, and refer to the relevant data in the life cycle assessment database during the filling process.
[0066] As an example, when populating the initial ESG report template with data from eligible companies, relevant data from the Life Cycle Assessment (LCA) database is referenced. That is, when analyzing a company's environmental impact, not only is the data provided by the company relied upon, but background information or industry standards from external databases are also utilized. For example, information on product carbon footprint, resource usage, and life cycle costs helps to more comprehensively assess a company's ESG performance.
[0067] For example, an initial ESG report template is a pre-designed document or form containing multiple sections and subsections. Each section corresponds to one or more key aspects of the initial ESG report, used to organize and present its content. For instance, an initial ESG report template might include multiple sections, each with a dedicated area for data display and interpretation, covering information such as type, source, usage, emission indicators, emissions, and total emissions. It may also include comparisons with industry standards, competitors, or historical data, providing crucial decision-making support for investors, regulators, consumers, and other stakeholders. These templates help companies gain a more comprehensive understanding of their ESG performance and develop improvement strategies accordingly.
[0068] Step S130: Calculate the data of the eligible enterprises using the preset formula in the initial ESG report template to generate the initial ESG report.
[0069] For example, the initial ESG report template typically also includes a series of pre-set formulas and indicators. These pre-set formulas are used to calculate corporate data that has been screened and corrected for reference data. The calculation results are then summarized into the initial ESG report to form a complete report that provides a quantitative analysis of corporate sustainability and social responsibility.
[0070] For example, data such as electricity bills and raw material purchase orders submitted by building material suppliers are filtered and screened based on business type, geographical location, and time frame. Next, the filtered data is populated into the corresponding initial ESG report template. For instance, the electricity consumption data of a company headquarters for ovens and generators between 2023 and 2024 is entered into the template for fixed emission sources for that year; simultaneously, the raw material purchase quantity data for that year is entered into the template for indirect emission sources. During the data population process, relevant data from the LCA database is used for quantitative analysis, including greenhouse gas emission indicators for different appliances and raw materials throughout their lifecycles. Subsequently, the initial ESG report template, which integrates company data and LCA database information, is processed according to pre-set formulas within the template. For example, the electricity consumption of ovens is multiplied by the emission factors of various greenhouse gases based on the energy composition of electricity and its corresponding emission factors (e.g., coal, natural gas, hydropower, nuclear power, renewable energy, etc.), and then the emissions of each greenhouse gas are summed to obtain the total emissions. The total emissions data obtained from these calculations are then entered into the corresponding positions in the initial ESG report template to generate a complete initial ESG report, such as... Figure 4 As shown.
[0071] Step S200: Input the initial ESG report into the generative intelligent model for analysis to obtain ESG recommendation topics.
[0072] Exemplifying examples are generative AI models, such as the GPT series (e.g., GPT-3, GPT-4) and T5 (Text-To-Text Transfer Transformer). These models can understand and generate natural language text, utilizing machine learning and deep learning techniques to generate new content by analyzing historical data patterns. Generative AI models can learn from information in databases and generate new content based on user input prompts or conditions.
[0073] For example, before inputting the initial ESG report into the generative intelligent model for analysis, the collected initial ESG report is first cleaned to remove irrelevant information, noisy data, and duplicate content. Then, the text data in the report is converted into a format suitable for model processing, including word segmentation, stop word removal, and stemming. Subsequently, based on the analysis requirements and data processing capabilities, a suitable generative intelligent model is selected, such as the GPT series, T5, or other large-scale language models.
[0074] To improve the accuracy and relevance of ESG suggestion topics, this application optimizes the RAG (Retrieval-Augmented Generation) method. Specifically, during the segmentation and embedding process, background information is input along with each paragraph and considered before conversion to the embedding representation. In addition to using embedding search, TF-IDF (Term Frequency-Inverse Document Frequency) encoding is incorporated to enhance the keyword search capability of the RAG database. Before generating the final suggestion topics, a re-ranker network is introduced to optimize the ranking of search results. Subsequently, the model is further trained by adjusting parameters such as model parameters, optimizer, and learning rate to improve its overall performance and accuracy. To evaluate the stability and reliability of the model, multiple methods, including cross-validation, are employed to ensure consistent performance across different datasets.
[0075] After model training is complete, the initial ESG report is used as input data and fed into the trained generative intelligent model to extract features from the text data in the initial ESG report. These features include word frequency, TF-IDF values, and word vectors, which reflect key information and themes in the text. The model extracts key ESG-related thematic features from the report, which may cover aspects such as the company's environmental performance indicators, social responsibility projects, and governance practices.
[0076] In other implementations, the model can be further optimized based on the quality and accuracy of the generated topics. This can be achieved by adjusting the model structure, increasing training data, or using more advanced natural language processing techniques. To validate the model's performance, it can be evaluated using an independent initial ESG report dataset, calculating metrics such as accuracy, recall, and F1 score to assess its accuracy and reliability in different scenarios. Finally, the initial ESG reports are input into the deployed model to generate ESG recommendation topics, providing strong support for the formulation and improvement of enterprise ESG strategies.
[0077] In one alternative embodiment, such as Figure 5 As shown, in step S200, the initial ESG report is input into the generative intelligent model for analysis to obtain ESG recommendation topics, including:
[0078] Step S210: Input the initial ESG report into the generative intelligent model for analysis to generate guidance and suggestions on the company's ESG status;
[0079] For example, the guidance includes generalized directions and insights. Generalized directions refer to universal or overarching trends in a company's ESG performance and development, extracted from the initial ESG report. Examples include the company's progress in environmental protection, fulfillment of social responsibility, and optimization of corporate governance structures. Insights refer to profound understandings and insights derived from the analysis of data from the initial ESG report and generative intelligent models. These generalized directions and insights reveal key information, patterns, or trends from the initial ESG report and their potential impact on the company's future development.
[0080] Step S220: Process the guidance and suggestions information on the enterprise's ESG status to obtain ESG suggestion topics.
[0081] For example, after obtaining guidance and recommendations on the ESG status of enterprises, all relevant ESG recommendation topics are sought based on the guidance and recommendations in the seventeen UN Sustainable Development Goals, such as: (1) Environmental recommendations: improve energy efficiency and reduce carbon emissions; promote the use of renewable energy and reduce dependence on fossil fuels; implement green supply chain management to ensure the environmental compliance of suppliers. (2) Governance recommendations: improve corporate governance structure and enhance decision-making efficiency and transparency; strengthen risk management to ensure the sound operation of enterprises; promote the implementation of sustainable development strategies to ensure long-term value creation.
[0082] In other words, based on the guidance and recommendations derived from the initial ESG report, themes for improvement suggestions or strategies for a company's ESG status can be identified. This helps companies better identify their strengths and weaknesses in ESG performance and develop effective strategies to improve and enhance it.
[0083] Step S300: Analyze and process the ESG suggestion topics to obtain search results.
[0084] As an example, each ESG recommendation topic is analyzed independently to generate topic embeddings. These embedding vectors are used to build an index, which is then combined with pre-generated index vectors in the LCA database and raw data index vectors provided by the enterprise to perform embedding vector searches, thereby obtaining relevant search results.
[0085] In one alternative embodiment, such as Figure 6 As shown, in step S300, the ESG suggestion topics are analyzed and processed to obtain search results, including:
[0086] Step S310: Analyze the ESG recommendation topics and generate topic embeddings for the corresponding ESG recommendation topics.
[0087] Exemplary topic embedding is a technique that represents topics as vectors that capture semantic relationships and similarities between topics. This allows ESG recommendation topics to be transformed into a processable data format in subsequent steps. The generated topic embeddings are combined with index vectors from the LCA database and the company's original data, leveraging embedding vector search techniques for efficient searching and matching within the initial ESG report, database, or repository. This process helps users quickly find data and information related to ESG recommendation topics, improving the accuracy and efficiency of data retrieval.
[0088] Step S320: The topic embedding is used as an index value and merged with the index value pre-generated in the LCA database and the index value pre-generated in the enterprise's original data to form an embedding vector and perform a search to obtain search results related to the ESG suggestion topic.
[0089] As an example, the generated topic embeddings are used as keywords for search or indexing, relevant data from the LCA database and raw enterprise data are converted into index values. Then, the topic embeddings of the ESG recommendation topics, the index values from the LCA database, and the index values from the raw enterprise data are merged to form a larger set of embedding vectors. This merged set of embedding vectors is used for search tasks, specifically searching the embedding vector space for LCA data or raw enterprise data most relevant to a particular ESG recommendation topic. In other words, the search results include results directly related to the ESG recommendation topic.
[0090] In short, by leveraging data analytics and processing techniques to tightly integrate ESG recommendations, LCA databases, and raw corporate data, companies can gain strong support for developing sustainable development strategies. By searching and analyzing this relevant data, companies can better understand their ESG performance and formulate improvement strategies accordingly.
[0091] Step S400: Use a generative intelligent model to perform strategy analysis on the search results to obtain guidance strategies.
[0092] Demonstratively, data and information related to ESG recommendations obtained through embedded vector search are input into a generative intelligent model. This model then performs strategic analysis on the input search results, such as analyzing historical data and current trends to predict future development directions in the ESG field. It assesses the potential risks of different ESG issues to enterprises or industries, including changes in environmental regulations, public opinion pressure, and corporate governance issues. It also identifies potential opportunities in the ESG field, such as green investment opportunities and social responsibility projects. Based on this analysis, the generative intelligent model then generates specific ESG guidance strategies, such as improvement suggestions, action plans, goal setting, and potential risk warnings.
[0093] In an optional embodiment, in step S400, a generative intelligent model is used to perform strategy analysis on the search results to obtain a guiding strategy, including:
[0094] Search results related to ESG recommendations are input into a generative intelligent model for analysis to obtain ESG guidance strategies expressed in professional tone and terminology.
[0095] As an example, data and information related to ESG recommendations, obtained through embedding vector search, are input into a generative intelligent model for in-depth analysis. Based on the analysis of the relevant data and information, the generative intelligent model provides guidance strategies using professional terminology and tone. Professional terminology includes relevant field-specific jargon, industry terms, and related keywords; while professional tone refers to a formal, objective, and rigorous expression.
[0096] It's worth noting that generative intelligent models are AI models trained on a large number of professional suggestion documents. Therefore, they can use professional tone and vocabulary when generating suggestions, providing in-depth, specific, and practical guidance strategies. For example, in practical applications, feedback provided by enterprises can be used to further optimize and improve the generative intelligent model, thereby enhancing the quality and accuracy of the guidance strategies.
[0097] Step S500: Optimize the guidance strategy using a preset evaluation model to generate an ESG recommendation report.
[0098] Exemplary, the evaluation model can be a machine learning-based classifier or regressor, or a deep learning model such as BERT or GPT. Evaluation criteria are set according to the characteristics of the guidance strategy, such as accuracy, relevance, completeness, readability, and professionalism, with weights assigned to each criterion to reflect its importance in the overall evaluation. Next, a preset threshold for quality scoring is set based on the evaluation criteria and weights. The guidance strategy is input into the preset evaluation model, which scores the guidance strategy according to the preset evaluation criteria and weights. Based on the scoring results, the generative intelligent model is adjusted to generate an ESG recommendation report.
[0099] In one alternative embodiment, such as Figure 7 As shown, in step S500, the guidance strategy is optimized using a preset evaluation model to generate an ESG recommendation report, including:
[0100] Step S510: Optimize the guidance strategy represented by professional tone and words using a preset evaluation model to obtain a quality score for the guidance strategy.
[0101] For example, the preset evaluation model is the BERT evaluation model. BERT (Bidirectional Encoder Representations from Transformers) is an advanced natural language processing (NLP) technology that can deeply understand the contextual information of text, thereby enabling more accurate analysis and evaluation. Using the BERT evaluation model, the tone and wording in the guidance strategy can be adjusted and optimized. For example, 'immoral' can be modified to the more accurate 'violates ethical standards'; 'wasteful' can be modified to the more professional 'low resource utilization', etc.
[0102] Subsequently, the BERT evaluation model scores the guidance strategies based on preset evaluation criteria and weights. These criteria typically cover multiple dimensions, including accuracy, relevance, completeness, readability, and professionalism, to ensure the comprehensiveness and objectivity of the evaluation. Each criterion is assigned a corresponding weight to reflect its importance in the overall evaluation.
[0103] Ultimately, the evaluation results will include scores for each criterion and an overall quality score. These scores will serve as an important basis for us to adjust and optimize the generative intelligent model in order to further improve the quality and accuracy of the guidance strategy.
[0104] Step S520: When the quality score is lower than the preset threshold, adjust the parameters of the generative intelligent model to regenerate the guidance strategy until the quality score of the guidance strategy is higher than or equal to the preset threshold.
[0105] As an example, the preset thresholds are set based on a variety of metrics, including but not limited to readability and tone. For instance, readability is assessed using Flesh Reading Comfort, reading time, vocabulary size, sentence size, and number of difficult words. Tone, on the other hand, is analyzed using tone analysis tools to assess the article's confidence and analytical quality. These metrics are taken into account to generate a quality score.
[0106] To ensure high standards, the preset threshold for the quality score is set at the upper quartile of historical report scores. If the quality score is below the preset threshold, it indicates that the quality of the guidance strategy has room for improvement and requires further adjustments and optimizations, such as adjusting the temperature parameter, until the quality score of the guidance strategy equals or exceeds the preset threshold. It is worth noting that adjusting the temperature parameter can make the generative intelligent model pay more attention to diversity or accuracy when generating suggestions, thereby improving the quality of the suggestions.
[0107] Simultaneously, the generative intelligent model extracts words that can be improved and places them into a context window. The context window allows the model to consider the specific meaning and usage of these words in the context when regenerating suggestions, thereby generating more accurate and coherent guidance policies. Through iterative generation and evaluation, the high quality and accuracy of the final generated guidance policy can be ensured.
[0108] In an optional embodiment, in step S520, when the quality score is lower than a preset threshold, the parameters of the generative intelligent model are adjusted to regenerate the guidance policy, until the quality score of the guidance policy is higher than or equal to the preset threshold, and the process further includes:
[0109] Guidance strategies that are higher than or equal to a preset threshold are converted into an ESG recommendation report according to a preset format.
[0110] As an example, the generated guidance strategies are arranged in a logical order to form coherent paragraphs. The content within each paragraph is then optimized and adjusted to ensure accuracy and completeness. All paragraphs are then merged into a final ESG recommendation report. This report should include an introduction, ESG recommendation classification, recommendation details, and assessment results to comprehensively reflect its content and quality. For example, it might recommend that the building materials supplier refer to best practices from its peers, prioritize the procurement and use of environmentally friendly building materials with lower greenhouse gas emissions, and select efficient and energy-saving production equipment and machinery to improve electricity efficiency. For instance, an ESG recommendation report involving a building materials supplier could include a line graph showing the supplier's monthly electricity consumption changes over a year. Figure 8As shown; and the following specific recommendations can also be presented to help building material suppliers reduce electricity usage and improve overall operational efficiency: (1) Optimize production processes: Conduct in-depth analysis of existing production processes to identify and eliminate unnecessary energy waste. Significantly improve the energy efficiency of production lines by introducing energy-saving equipment and advanced technologies. (2) Implement energy consumption monitoring systems: Install energy consumption monitoring equipment to track electricity usage in real time. This will help identify high-energy-consuming equipment and their usage periods, thereby developing targeted energy-saving measures. (3) Invest in high-efficiency equipment: Prioritize investment in high-efficiency machinery and equipment, such as variable frequency drives and energy-saving motors. These devices can significantly reduce electricity consumption while maintaining production efficiency during operation. (4) Employee training and awareness enhancement: Conduct regular energy-saving training to improve employees' awareness of energy use. Encourage employees to make energy-saving suggestions and actively take energy-saving measures in their daily work. (5) Strengthen equipment maintenance: Regularly maintain equipment to ensure it operates in optimal condition. Good maintenance can not only reduce failure rates but also effectively reduce energy waste. (6) Utilize renewable energy: Consider installing solar panels or other renewable energy systems on facilities to reduce dependence on external electricity. This can not only reduce electricity bills, but also enhance the company's environmental image. (7) Make good use of natural light: When there is sufficient natural light near the window, indoor lighting should be turned off. Consider installing light sensors to reduce energy waste caused by human error. (8) Manage the use of electrical appliances: For appliances that will not be used in the next hour, they should be turned off or set to standby mode in time to avoid unnecessary power consumption. (9) Optimize the use of air conditioning: Set the air conditioning temperature between 24°C and 26°C to ensure a balance between comfort and energy efficiency. (10) Implement a separate switch system: Use a separate switch system in areas such as corridors, and flexibly adjust the lighting according to the actual use to improve energy efficiency.
[0111] By implementing the above recommendations, building material suppliers can effectively reduce electricity consumption, improve operational efficiency, and make a positive contribution to sustainable development.
[0112] The method for automatically generating ESG recommendation reports according to embodiments of this application first processes the enterprise's raw data to generate an initial ESG report. This initial ESG report is then input into a generative intelligent model for analysis to obtain ESG recommendation topics. Next, the ESG recommendation topics are processed to obtain search results, which are then input into the generative intelligent model for analysis to generate guidance strategies. Finally, a preset evaluation model is used to optimize the guidance strategies, generating the ESG recommendation report. This method not only improves the efficiency of analyzing and processing initial ESG reports but also provides enterprises with a comprehensive, accurate, and standardized initial ESG report, helping them better understand their ESG performance and formulate corresponding improvement strategies.
[0113] This application also proposes an automatic generation device for ESG recommendation reports, such as... Figure 9 As shown, the automatic generation device for this ESG recommendation report includes:
[0114] The initial report generation module 91 is used to process the enterprise's raw data and generate an initial ESG report;
[0115] The topic generation module 92 is used to input the initial ESG report into the generative intelligent model for analysis and obtain ESG suggested topics.
[0116] The search result acquisition module 93 is used to analyze and process the ESG suggestion topics to obtain search results;
[0117] The guidance strategy generation module 94 is used to perform strategy analysis on the search results using the generative intelligent model to obtain a guidance strategy.
[0118] The recommendation report generation module 95 is used to optimize the guidance strategy using a preset evaluation model and generate an ESG recommendation report.
[0119] It is understood that the apparatus in this embodiment corresponds to an automatic generation method for an ESG recommendation report in the above embodiments. The options in the above embodiments are also applicable to this embodiment, so they will not be described again here.
[0120] This application also provides a computer device, exemplary of which includes a processor and a memory, wherein the memory stores a computer program, and the processor, by running the computer program, causes the computer device to perform the functions of the various modules in the above-described method for automatically generating ESG recommendation reports or the above-described apparatus for automatically generating ESG recommendation reports.
[0121] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Network Processor (NP), Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.
[0122] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory is used to store computer programs, and the processor can execute the computer programs accordingly after receiving execution instructions.
[0123] This application also provides a computer-readable storage medium for storing the computer program used in the aforementioned computer device. For example, the computer-readable storage medium may include, but is not limited to, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0124] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, in alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0125] In addition, the functional modules or units in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0126] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.
[0127] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.
Claims
1. A method for automatically generating ESG recommendation reports, characterized in that, The method includes: The process involves processing raw enterprise data and analyzing it in conjunction with data from a lifecycle assessment database to generate an initial ESG report. This includes: filtering the raw enterprise data based on business type, geographical location, and time range to obtain eligible enterprise data; filling the eligible enterprise data into a preset initial ESG report template, referencing relevant data from the lifecycle assessment database during the filling process; and using preset formulas in the initial ESG report template to calculate the eligible enterprise data and generate the initial ESG report. The initial ESG report is input into a generative intelligent model for analysis to obtain ESG recommendation topics. The ESG recommendation topics are analyzed and processed to obtain search results, including: analyzing the ESG recommendation topics and generating topic embeddings corresponding to the ESG recommendation topics; using the topic embeddings as index values, merging them with index values pre-generated in the life cycle assessment database and index values pre-generated in the enterprise's original data to form an embedding vector and performing a search to obtain search results related to the ESG recommendation topics; The generative intelligent model is used to perform strategy analysis on the search results to obtain guiding strategies; The guidance strategy is optimized using a pre-defined evaluation model to generate an ESG recommendation report.
2. The method for automatically generating ESG recommendation reports according to claim 1, characterized in that, The step of inputting the initial ESG report into a generative intelligent model for analysis to obtain ESG recommendation topics includes: The initial ESG report is input into a generative intelligent model for analysis, generating guidance and suggestions on the company's ESG status. The guidance and recommendations on the enterprise's ESG status are processed to obtain ESG recommendation topics.
3. The method for automatically generating ESG recommendation reports according to claim 1, characterized in that, The step of using the generative intelligent model to perform strategy analysis on the search results to obtain guiding strategies includes: Search results related to the ESG recommendation topics are input into the generative intelligent model for analysis to obtain the guidance strategies expressed through professional tone and vocabulary.
4. The method for automatically generating ESG recommendation reports according to claim 3, characterized in that, The step of optimizing the guidance strategy using a preset evaluation model to generate an ESG recommendation report includes: The guidance strategy, expressed through professional tone and vocabulary, is optimized using a pre-defined evaluation model to obtain a quality score for the guidance strategy. When the quality score is lower than a preset threshold, the parameters of the generative intelligent model are adjusted to regenerate the guidance strategy until the quality score of the guidance strategy is higher than or equal to the preset threshold.
5. The method for automatically generating ESG recommendation reports according to claim 4, characterized in that, The step of adjusting the parameters of the generative intelligent model to regenerate the guidance policy when the quality score is lower than a preset threshold, until the quality score of the guidance policy is higher than or equal to the preset threshold, further includes: The guidance strategy that is higher than or equal to the preset threshold is converted according to a preset format to generate an ESG recommendation report.
6. An automatic generation device for ESG recommendation reports, characterized in that, The device includes: The initial report generation module processes raw enterprise data and analyzes it in conjunction with data from the life cycle assessment database to generate an initial ESG report. This includes: filtering the raw enterprise data based on business type, geographical location, and time range to obtain eligible enterprise data; filling the eligible enterprise data into a preset initial ESG report template, referencing relevant data from the life cycle assessment database during the filling process; and calculating the eligible enterprise data using preset formulas in the initial ESG report template to generate the initial ESG report. The topic generation module is used to input the initial ESG report into the generative intelligent model for analysis and obtain ESG suggested topics. The search result acquisition module is used to analyze and process the ESG recommendation topics to obtain search results, including: analyzing the ESG recommendation topics and generating topic embeddings corresponding to the ESG recommendation topics; using the topic embeddings as index values, merging them with the index values pre-generated by the life cycle assessment database and the index values pre-generated by the enterprise's original data to form an embedding vector and performing a search to obtain search results related to the ESG recommendation topics; The guidance strategy generation module is used to perform strategy analysis on the search results using the generative intelligent model to obtain guidance strategies. The recommendation report generation module is used to optimize the guidance strategy using a preset evaluation model and generate an ESG recommendation report.
7. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the automatic generation method of the ESG recommendation report according to any one of claims 1-5.
8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed on a processor, implements the method for automatically generating ESG recommendation reports according to any one of claims 1-5.