CN116432625A - Enterprise annual report learning and generating system and method based on deep learning - Google Patents
Enterprise annual report learning and generating system and method based on deep learning Download PDFInfo
- Publication number
- CN116432625A CN116432625A CN202310187469.XA CN202310187469A CN116432625A CN 116432625 A CN116432625 A CN 116432625A CN 202310187469 A CN202310187469 A CN 202310187469A CN 116432625 A CN116432625 A CN 116432625A
- Authority
- CN
- China
- Prior art keywords
- annual
- learning
- report
- enterprise
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013135 deep learning Methods 0.000 title claims abstract description 23
- 238000003860 storage Methods 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 12
- 102100033814 Alanine aminotransferase 2 Human genes 0.000 claims description 17
- 101710096000 Alanine aminotransferase 2 Proteins 0.000 claims description 15
- 238000004519 manufacturing process Methods 0.000 claims description 6
- 230000006399 behavior Effects 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 5
- 230000010485 coping Effects 0.000 claims description 5
- 238000002560 therapeutic procedure Methods 0.000 claims description 5
- 238000013138 pruning Methods 0.000 claims description 4
- 238000009877 rendering Methods 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 4
- 230000008569 process Effects 0.000 description 7
- 230000001276 controlling effect Effects 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 4
- 101000779415 Homo sapiens Alanine aminotransferase 2 Proteins 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a deep learning-based enterprise annual report learning and generating system and method, wherein the method comprises the following steps of S1, constructing a code module of a crawler and a local file storage and analysis module; s2, inputting keywords into a local file storage and analysis module to obtain annual reports of a specified enterprise and a specified year; s3, inputting keywords into the crawler code module, and acquiring annual reports of specified enterprises and specified years from the Internet; s4, extracting text information in the annual messages obtained in the step S2 or the step S3, and constructing a JSON database; s5, training a large language character model by adopting a JSON database; s6, inputting enterprise brief introduction keywords, and automatically generating new year reports meeting the requirements by using a large language character model; the invention provides a more intelligent and practical enterprise annual report learning and generating system and method based on deep learning.
Description
Technical Field
The invention relates to the technical field of intelligent systems, in particular to an enterprise annual report learning and generating system and method based on deep learning.
Background
ESG, environmental, social and corporate governments (environmental, social Responsibility, corporate Governance), including information disclosure, assessment ratings and investment guidelines, is the basis for social liability investments and is an important component of green melt systems. Through ESG, investors can pay attention to the investment concepts and enterprise evaluation standards of the three non-financial indexes, and the core is that investment and business actions should not only consider financial indexes, but also consider social responsibility factors of the investors for environment, society, management and the like, so that enterprises can obtain sustainable development instead of short-term benefits.
GPT (generated Pre-trained Transformer) is a text generation model based on probabilistic modeling, which builds probabilistic descriptions of languages through extensive corpus training, and thus can be used for natural language generation. The ability of the language model to predict the next word depends on two factors: whether all the history context information can be fully utilized; second, whether more distant semantic information can be collected. The second question is whether the training corpus is large enough to provide a model with enough historical context for learning. Since the language model is an integral part of self-supervised learning, the optimization objective is to increase the likelihood that the language model will be seen in the text.
Internationally, the international financial reporting guidelines foundation is researching the ESG international standard, which will combine and introduce the existing plurality of ESG standards and specifications, reducing the disclosure and data costs. And many important economics in the world have established ESG policies and regulatory bodies from the point of view of regulations and policies. For example, the European Union has implemented the European Union taxonomies and the sustainable financial disclosure regulations, establishing a comprehensive policy hierarchy from taxonomies to disclosures, standardizing the entire supply chain of the investment industry.
In China, with the rise of the ESG field, some companies have also begun to issue their ESG reports, which provides a rich corpus for training of our models.
However, at the same time, we also see that in the market of China, ESG reports are used as information of each company, and a unified management mechanism is lacked; much of the content of an ESG report is more templated, but it still takes much time to manually write an EGS report. Thus, it would be desirable to automatically collect the annual reports and ESG reports from companies on the Internet, train a GPT model with the data, and automatically generate the required annual reports using the GPT model.
Disclosure of Invention
The method of the invention provides a more intelligent and practical enterprise annual report learning and generating system and method based on deep learning.
To achieve the purpose, the invention provides the following technical scheme:
the invention provides a deep learning-based enterprise annual report learning and generating method, which comprises the following steps:
s1, constructing a code module and a local file storage and analysis module of a crawler;
s2, inputting keywords into a local file storage and analysis module to obtain annual reports of a specified enterprise and a specified year; if the local file storage and analysis module has no relevant data information, the step S3 is carried out, and if the local file storage and analysis module has relevant data information, the step S4 is directly carried out;
s3, inputting keywords into the crawler code module, and acquiring annual reports of specified enterprises and specified years from the Internet;
s4, extracting text information in the annual messages obtained in the step S2 or the step S3, and constructing a JSON database;
s5, training a large language character model by adopting a JSON database;
s6, inputting enterprise brief introduction keywords, and automatically generating new annual reports meeting the requirements by using the large language word model.
In the present invention, in step S3, the annual report crawler obtains the annual report in PDF format of the designated enterprise and the designated year from the internet, for example, the "huge tide information network". The crawler of the enterprise annual newspaper is mainly from the information disclosure platform of the marketing companies such as 'huge tide information network', and the requirement of the invention on the website selection is that a large-scale securities professional website which can cover the advertising information and market data of 2500 marketing companies in the deep Shanghai is required. Because only sufficient corpus of each industry can be used for subsequent training, the model can obtain context understanding.
Preferably, step S4 includes: the annual newspaper is split into modules, and the module information is classified by an unsupervised chemical learning method.
Preferably, the large language text model comprises a Bert or GPT-2 model.
In the present invention, the Bert model captures word and sentence level descriptions using two methods, masked LM and Next Sentence Prediction, respectively. The GPT-2 model is a text language learning model for artificial intelligence, and has the main function of generating a continuous writing text according to the content input by a user.
Preferably, the annual report obtained in the step S3 is in PDF or HTML format, and the following steps are further included between the step S3 and the step S4: and analyzing and converting the annual report in the PDF or HTML format into a text format by adopting a PDF or HTML analyzer.
Preferably, step S5 includes constructing an ESG report infrastructure in a large language text model.
Preferably, constructing the ESG report infrastructure includes the steps of:
r1, extracting common sentences and indexes from the ESG reports disclosed by the existing marketing companies based on the related information of the ESG reports, and performing color rendering to a certain extent, wherein the common sentences and indexes comprise, but are not limited to, coping with climate change in the environment field, reasonably utilizing natural resources, strictly controlling pollution in green production, human capital in the social field, safety management, social opportunity and company management and company behaviors in the management field, and constructing a basic and universal framework of the ESG report by using the sentences;
r2, on the basis of the basic and general framework constructed in the step R1, comparing the differential sentences and indexes of different industries, and pruning the basic report according to the future market prospect and the knowledge of different industries to form different ESG report frameworks with industry characteristics, wherein the differential sentences and indexes comprise but are not limited to: subject safety in the pharmaceutical industry, innovative therapy in the sanitary industry, green transportation in the logistics industry, student care in the educational industry, anti-fraud and information security in the telecommunications industry.
Preferably, the large language word model can learn multiple related tasks together.
In the present invention, the large language literal model uses more network parameters and a larger data set. For example, the GPT-2 model learns multiple related tasks together while GPT2 learns multiple tasks during the process. The largest model in GPT-2 has 48 layers and 15 billions of references, and the learning objective uses an unsupervised pre-training model for the supervising task.
In a second aspect of the invention, a code module comprising a crawler, a local file storage and analysis module, a JSON database, and a large language text model are provided.
Preferably, the method of the system is the enterprise annual report learning and generating method based on deep learning.
Preferably, the local file storage and analysis module may put a plurality of related tasks together for learning.
Preferably, the method for learning and generating the annual newspaper of the enterprise based on the deep learning comprises the following steps:
s1, constructing a code module and a local file storage and analysis module of a crawler;
s2, inputting keywords into a local file storage and analysis module to obtain annual reports of a specified enterprise and a specified year; if the local file storage and analysis module has no relevant data information, the step S3 is carried out, and if the local file storage and analysis module has relevant data information, the step S4 is directly carried out;
s3, inputting keywords into the crawler code module, and acquiring annual reports of specified enterprises and specified years from the Internet;
s4, extracting text information in the annual messages obtained in the step S2 or the step S3, and constructing a JSON database;
s5, training a large language character model by adopting a JSON database;
s6, inputting enterprise brief introduction keywords, and automatically generating new annual reports meeting the requirements by using the large language word model.
Preferably, step S4 includes: the annual newspaper is split into modules, and the module information is classified by an unsupervised chemical learning method.
Preferably, the large language text model comprises a Bert or GPT-2 model.
Preferably, the annual report obtained in the step S3 is in PDF or HTML format, and the following steps are further included between the step S3 and the step S4: and analyzing and converting the annual report in the PDF or HTML format into a text format by adopting a PDF or HTML analyzer.
Preferably, step S5 includes constructing an ESG report infrastructure in a large language text model.
Preferably, constructing the ESG report infrastructure includes the steps of:
r1, extracting common sentences and indexes from the ESG reports disclosed by the existing marketing companies based on the related information of the ESG reports, and performing color rendering to a certain extent, wherein the common sentences and indexes comprise, but are not limited to, coping with climate change in the environment field, reasonably utilizing natural resources, strictly controlling pollution in green production, human capital in the social field, safety management, social opportunity and company management and company behaviors in the management field, and constructing a basic and universal framework of the ESG report by using the sentences;
r2, on the basis of the basic and general framework constructed in the step R1, comparing the differential sentences and indexes of different industries, and pruning the basic report according to the future market prospect and the knowledge of different industries to form different ESG report frameworks with industry characteristics, wherein the differential sentences and indexes comprise but are not limited to: subject safety in the pharmaceutical industry, innovative therapy in the sanitary industry, green transportation in the logistics industry, student care in the educational industry, anti-fraud and information security in the telecommunications industry.
Compared with the prior art, the invention has the beneficial effects and remarkable progress that: the enterprise annual newspaper learning and generating system and method based on deep learning provided by the invention solve the problem that a large number of annual newspaper are difficult to manually analyze, and solve the requirement of automatically generating high-quality annual newspaper according to prompts.
Drawings
In order to more clearly illustrate the technical solution of the present invention, a brief description will be given below of the drawings that are required to be used for the embodiments of the present invention.
It is obvious that the drawings in the following description are only drawings of some embodiments of the present invention, and that other drawings may be obtained from these drawings without inventive faculty for a person skilled in the art, but these other drawings also fall within the drawings required for the embodiments of the present invention.
Fig. 1 is a flowchart of an enterprise annual report learning and generating method based on deep learning in embodiment 1 of the present invention.
FIG. 2 is a schematic diagram of a structure in which a plurality of related tasks of GPT-2 of embodiment 2 of the invention are learned together;
fig. 3 is a flow chart of the annual newspaper production process of embodiment 3 of the present invention.
Detailed Description
In order to make the objects, technical solutions, beneficial effects and significant improvements of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the drawings provided in the embodiments of the present invention.
It is apparent that all of the described embodiments are only some, but not all, embodiments of the invention; all other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that the terms "first," "second," and "third" (if any) in the description and claims of the present invention and the drawings of the embodiments of the present invention are used merely for distinguishing between different objects and not for describing a particular sequential order. Furthermore, the terms "comprise," "include," and any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those listed steps or elements but may include other steps or elements not listed or inherent to such process, method, article, or apparatus.
It is to be understood that:
in the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "connected," "secured," and the like are to be construed broadly, and may be, for example, fixedly connected, detachably connected or movably connected, or integrally formed; either directly, indirectly, through intermediaries, or through an intangible signal connection, or even optically, in communication with one another, or in interaction with one another, unless expressly defined otherwise.
The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.
It should also be noted that the following embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
The following describes the technical scheme of the present invention in detail by using specific examples.
Example 1
As shown in FIG. 1, the method for learning and generating the annual newspaper of the enterprise based on the deep learning comprises the following steps:
s1, constructing a code module and a local file storage and analysis module of a crawler;
s2, inputting keywords into a local file storage and analysis module to obtain annual reports of a specified enterprise and a specified year; if the local file storage and analysis module has no relevant data information, the step S3 is carried out, and if the local file storage and analysis module has relevant data information, the step S4 is directly carried out;
s3, inputting keywords into the crawler code module, and acquiring annual reports of specified enterprises and specified years from the Internet;
s4, extracting text information in the annual messages obtained in the step S2 or the step S3, and constructing a JSON database;
s5, training a large language character model by adopting a JSON database;
s6, inputting enterprise brief introduction keywords, and automatically generating new annual reports meeting the requirements by using the large language word model.
In this embodiment, step S4 includes: the annual newspaper is split into modules, and the module information is classified by an unsupervised chemical learning method.
In this embodiment, the large language text model includes a Bert or GPT-2 model.
In this embodiment, the annual report obtained in step S3 is in PDF or HTML format, and the following steps are further included between step S3 and step S4: and analyzing and converting the annual report in the PDF or HTML format into a text format by adopting a PDF or HTML analyzer.
In this embodiment, step S5 includes constructing an ESG report infrastructure in a large language text model.
In this embodiment, constructing the ESG report infrastructure includes the steps of:
r1, extracting common sentences and indexes from the ESG reports disclosed by the existing marketing companies based on the related information of the ESG reports, and performing color rendering to a certain extent, wherein the common sentences and indexes comprise, but are not limited to, coping with climate change in the environment field, reasonably utilizing natural resources, strictly controlling pollution in green production, human capital in the social field, safety management, social opportunity and company management and company behaviors in the management field, and constructing a basic and universal framework of the ESG report by using the sentences;
r2, on the basis of the basic and general framework constructed in the step R1, comparing the differential sentences and indexes of different industries, and pruning the basic report according to the future market prospect and the knowledge of different industries to form different ESG report frameworks with industry characteristics, wherein the differential sentences and indexes comprise but are not limited to: subject safety in the pharmaceutical industry, innovative therapy in the sanitary industry, green transportation in the logistics industry, student care in the educational industry, anti-fraud and information security in the telecommunications industry.
In this embodiment, the large language word model may put multiple related tasks together for learning.
Example 2
An enterprise annual report learning and generating system based on deep learning comprises a code module of a crawler, a local file storage and analysis module, a JSON database and a GPT-2 model module. As shown in fig. 2, GPT-2 uses more network parameters and a larger data set. In this process, GPT2 puts multiple related tasks together for learning, while learning multiple tasks. The largest model in GPT-2 has 48 layers and 15 billions of references, and the learning objective uses an unsupervised pre-training model for the supervising task.
In this embodiment, the system further includes a PDF or HTML parser, and the PDF or HTML parser is used to parse and convert the annual report in PDF or HTML format into text format.
In this embodiment, a text generation console is also included.
Example 3
In a specific embodiment, as shown in FIG. 3.
The first step: construction of an ESG report infrastructure.
Step 1: and extracting sentences and indexes to construct an ESG report basic framework. Based on related information such as ESG reports disclosed by the existing marketing companies, sentences and indexes common to different industry reports are extracted from the ESG reports, and the ESG reports are colored to a certain extent, wherein the sentences and indexes comprise, but are not limited to, coping with climate change in the environment field, reasonably utilizing natural resources, strictly controlling pollution by green production, human capital in the social field, safety management, social opportunity and company management and company behaviors in the management field, and a basic and universal framework of the ESG reports is constructed by the sentences.
Step 2: and comparing different industries, and perfecting an ESG report framework. Based on the basic framework, sentences and indexes appearing in different industries are compared, and according to the knowledge of an inventor on future market and different industries, basic reports are pruned to form different ESG report frameworks with industry characteristics, wherein the different sentences comprise, but are not limited to: subject safety in the pharmaceutical industry, innovative therapy in the sanitary industry, green transportation in the logistics industry, student care in the educational industry, anti-fraud and information security in the telecommunications industry.
And a second step of: learning process of GPT-2 model. Suppose we wish to train with the annual newspaper of all enterprises 2013-2021 of a-ply.
Step 1: we need to obtain a sufficiently rich historical context for model learning, so we build code modules for crawlers. In this module, we need to input two parameters, keyword and time interval, to download the training corpus meeting our requirements. The limited conditions "A strand" and year "2013-2021" of crawling are input into the crawler, and then the crawler automatically crawls all the annual reports meeting the conditions on the "huge tide information network" and stores the annual reports in PDF format.
Step 2: and analyzing each PDF annual report file by using a PDF analyzer, and extracting the text in the annual report to obtain JSON data.
Step 3: the GPT-2 model was trained using JSON data.
And a third step of: GPT-2 model generation process. Assuming that we wish to generate the "environmental governance" section of company a annual newspaper, then the variables we input need to include a section of company a's profile and the required section name.
Step 1: writing company a profile, placed in the first portion of the input; the profile cannot be automatically generated and must be manually written.
Step 2: keywords of the paragraph to be generated are given, here "environmental governance".
Step 3: in the visualization window, the GPT-2 model is controlled to generate paragraphs.
In the description of the above specification:
the terms "this embodiment," "an embodiment of the invention," "as shown in … …," "further improved embodiments," and the like, mean that a particular feature, structure, material, or characteristic described in the embodiment or example is included in at least one embodiment or example of the invention; in this specification, a schematic representation of the above terms is not necessarily directed to the same embodiment or example, and the particular features, structures, materials, or characteristics described, etc. may be combined or combined in any suitable manner in any one or more embodiments or examples; furthermore, various embodiments or examples, as well as features of various embodiments or examples, described in this specification may be combined or combined by one of ordinary skill in the art without undue experimentation.
Finally, it should be noted that:
the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting thereof;
although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will appreciate that modifications may be made to the technical solutions described in the foregoing embodiments, or equivalents may be substituted for some or all of the technical features thereof, without departing from the spirit of the technical solutions of the embodiments of the present invention, and that insubstantial improvements and modifications or substitutions by one skilled in the art from the disclosure herein are within the scope of the invention as claimed.
Claims (10)
1. The enterprise annual report learning and generating method based on deep learning is characterized by comprising the following steps of:
s1, constructing a code module and a local file storage and analysis module of a crawler;
s2, inputting keywords into a local file storage and analysis module to obtain annual reports of a specified enterprise and a specified year; if the local file storage and analysis module has no relevant data information, the step S3 is carried out, and if the local file storage and analysis module has relevant data information, the step S4 is directly carried out;
s3, inputting keywords into the crawler code module, and acquiring annual reports of specified enterprises and specified years from the Internet;
s4, extracting text information in the annual messages obtained in the step S2 or the step S3, and constructing a JSON database;
s5, training a large language character model by adopting a JSON database;
s6, inputting enterprise brief introduction keywords, and automatically generating new annual reports meeting the requirements by using the large language word model.
2. The method for learning and generating the annual newspaper of the enterprise based on the deep learning as claimed in claim 1, wherein the step S4 comprises: the annual newspaper is split into modules, and the module information is classified by an unsupervised chemical learning method.
3. The method for learning and generating annual messages for enterprises based on deep learning as claimed in claim 1, wherein said large language text model comprises Bert or GPT-2 model.
4. The method for learning and generating the annual report of the enterprise based on the deep learning as claimed in claim 1, wherein the annual report acquired in the step S3 is in PDF or HTML format, and the following steps are further included between the step S3 and the step S4: and analyzing and converting the annual report in the PDF or HTML format into a text format by adopting a PDF or HTML analyzer.
5. The method for learning and generating annual messages for enterprises based on deep learning as set forth in claim 1, wherein step S5 includes constructing an ESG report infrastructure in a large language text model.
6. The method for learning and generating the annual newspaper of the enterprise based on deep learning as recited in claim 5, wherein constructing the ESG report infrastructure comprises the steps of:
r1, extracting common sentences and indexes from the ESG reports disclosed by the existing marketing companies based on the related information of the ESG reports, and performing color rendering to a certain extent, wherein the common sentences and indexes comprise, but are not limited to, coping with climate change in the environment field, reasonably utilizing natural resources, strictly controlling pollution in green production, human capital in the social field, safety management, social opportunity and company management and company behaviors in the management field, and constructing a basic and universal framework of the ESG report by using the sentences;
r2, on the basis of the basic and general framework constructed in the step R1, comparing the differential sentences and indexes of different industries, and pruning the basic report according to the future market prospect and the knowledge of different industries to form different ESG report frameworks with industry characteristics, wherein the differential sentences and indexes comprise but are not limited to: subject safety in the pharmaceutical industry, innovative therapy in the sanitary industry, green transportation in the logistics industry, student care in the educational industry, anti-fraud and information security in the telecommunications industry.
7. The method for learning and generating annual messages for enterprises based on deep learning as set forth in claim 1, wherein the large language text model can learn a plurality of related tasks together.
8. The enterprise annual report learning and generating system based on deep learning is characterized by comprising a code module of a crawler, a local file storage and analysis module, a JSON database and a large language text model.
9. The deep learning-based business annual newspaper learning and generating system in accordance with claim 8, wherein the method of the system is the method in any one of claims 1-7.
10. The deep learning-based business annual newspaper learning and generating system as recited in claim 8 wherein the local file storage and analysis module learns a plurality of related tasks together.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310187469.XA CN116432625A (en) | 2023-03-02 | 2023-03-02 | Enterprise annual report learning and generating system and method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310187469.XA CN116432625A (en) | 2023-03-02 | 2023-03-02 | Enterprise annual report learning and generating system and method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116432625A true CN116432625A (en) | 2023-07-14 |
Family
ID=87083821
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310187469.XA Pending CN116432625A (en) | 2023-03-02 | 2023-03-02 | Enterprise annual report learning and generating system and method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116432625A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117252565A (en) * | 2023-11-10 | 2023-12-19 | 北京华品博睿网络技术有限公司 | Company bright spot generation method and system |
-
2023
- 2023-03-02 CN CN202310187469.XA patent/CN116432625A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117252565A (en) * | 2023-11-10 | 2023-12-19 | 北京华品博睿网络技术有限公司 | Company bright spot generation method and system |
| CN117252565B (en) * | 2023-11-10 | 2024-02-06 | 北京华品博睿网络技术有限公司 | Company bright spot generation method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111428053B (en) | Construction method of tax field-oriented knowledge graph | |
| Leopold et al. | Learning from quality issues of BPMN models from industry | |
| Dehler-Holland et al. | Assessing technology legitimacy with topic models and sentiment analysis–The case of wind power in Germany | |
| CN107908671A (en) | Knowledge mapping construction method and system based on law data | |
| Frantz et al. | Institutional grammar | |
| CN111581376A (en) | Automatic knowledge graph construction system and method | |
| Smeureanu et al. | Business ontology for evaluating corporate social responsibility | |
| CN116432625A (en) | Enterprise annual report learning and generating system and method based on deep learning | |
| Cheng et al. | A cognitive level evaluation method based on a deep neural network for online learning: from a bloom’s taxonomy of cognition objectives perspective | |
| CN113095050A (en) | Intelligent ticketing method, system, equipment and storage medium | |
| Ni et al. | Rapid generation of emergency response plans for unconventional emergencies | |
| CN117094390A (en) | Knowledge graph construction and intelligent search method oriented to ocean engineering field | |
| CN113240443B (en) | Entity attribute pair extraction method and system for power customer service question and answer | |
| Fu et al. | Research on Discrete Dynamic Modeling of Learner Behavior Analysis in English Teaching | |
| Fisher et al. | Using language processing to evaluate the equivalency of the FASB and IASB standards | |
| Papadakis et al. | Ontology-guided Knowledge Graph Construction to Support Scheduling in a Train Maintenance Depot. | |
| Ten | The digitalization of economy as the" response" to the global change. | |
| Panditharathna et al. | Question and answering system for investment promotion based on nlp | |
| Yan et al. | Construction of Dynamic Knowledge Graph for Grid day-ahead Scheduling Optimization Decision | |
| Atencio et al. | Developing a Tool for PBO Design and Analysis: An Integration Between Enterprise Architecture and Project Management Domains | |
| CN114996407B (en) | Remote supervision relation extraction method and system based on packet reconstruction | |
| Mu et al. | Construction of Knowledge Graph for Emergency Resources | |
| Wu et al. | Achieving Greater Educational Impact through Data Intelligence: Practice, Challenges and Expectations of Education | |
| Wong | THE IMPACT OF LARGE-SCALE LANGUAGE MODELS ON THE FUTURE DEVELOPMENT OFACCOUNTING | |
| CN108875938B (en) | Method for making implicit knowledge explicit in traditional Chinese medicine collaborative development process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |