CN117350253A - Financial document automatic generation method and system of GPT model architecture - Google Patents
Financial document automatic generation method and system of GPT model architecture Download PDFInfo
- Publication number
- CN117350253A CN117350253A CN202311388756.3A CN202311388756A CN117350253A CN 117350253 A CN117350253 A CN 117350253A CN 202311388756 A CN202311388756 A CN 202311388756A CN 117350253 A CN117350253 A CN 117350253A
- Authority
- CN
- China
- Prior art keywords
- model
- text
- training
- document
- business
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
- G06N3/092—Reinforcement 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
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (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 provides a financial document automatic generation method and system based on a GPT model architecture, and relates to the field of computer systems. Inputting characters, forms, image character descriptions and characters in the image in a financial document to a pre-training model, and performing feature processing through an embedded network and a self-attention network in the pre-training model to obtain visual representation of an image area in the document image and text representation of the characters; determining a loss value according to the visual representation and/or the text representation by using a supervision strategy; the invention not only can fully cover business banking business scenes, but also can automatically adapt generated contract clauses according to specific businesses. In addition, on the basis, the invention can automatically generate documents commonly used by commercial banks such as books, industry analysis reports and the like, thereby greatly reducing the repeated workload of first-line business personnel of the commercial banks.
Description
Technical Field
The invention relates to the field of computer systems, in particular to a financial document automatic generation method and a financial document automatic generation system for a GPT model architecture.
Background
The general large language model (GPT-3, LLaMA and the like) is an NLP pre-training model based on a Transformer model architecture, can greatly improve the accuracy of text generation tasks in NLP tasks, and is widely applied to chat machines (such as ChatGPT), intelligent customer service and other scenes at present. A set of document automatic generation system is constructed by using a pre-training model based on a GPT architecture, so that business banking first-line business personnel can input requirement profiles such as contract templates of certain financial products, recent industry analysis reports of certain industries, bidding documents of certain enterprise financial services and the like, and related documents are automatically generated through a large language model, so that the manual workload is greatly reduced, and meanwhile, the accuracy of the related documents is improved.
However, the current automatic document generation system is limited to a certain type of document, such as automatic generation of a contract document, not only the type of generated document is single, but also the generated document is mostly a fixed text pattern generated based on fixed rules, and is not applicable any more if a new scene or content item is encountered. The documents generated according to the rules do not accord with the common language habits of human beings, and the documents can be used only after being generated by manual adjustment, so that a larger gap exists between the text effects generated by the intelligent language model.
Disclosure of Invention
Aiming at the requirements set forth in the background technology, the embodiment of the invention provides a financial document automatic generation method and a financial document automatic generation system of a GPT model framework, a set of document generation system capable of understanding Chinese requirements is built on the basis of a general large language model, generation and inspection of multiple types of documents are automatically completed, and finally documents with directly available scenes are generated.
A financial document automatic generation method of GPT model architecture includes the steps:
firstly, building a pre-training model based on a GPT-3 architecture or a LLaMA architecture (opened source), and obtaining a document pre-training model by pre-training an existing document;
specifically, the related documents of the business are arranged, the text, the table text, the text in the image and the like in the documents are classified by using an NLP text classification model according to the business types, the classification result is adjusted by using manpower, the classified text is input into a pre-training model, and the characteristic processing is carried out through an embedded network and a self-attention network in the pre-training model, so that the text representation of the text of the documents is obtained, and further, a financial text pre-training model is obtained; .
Step two, performing fine adjustment of a specific scene text generation model by using respective type distinguishing data such as contracts, identifications, industry analysis reports and the like;
and thirdly, automatically improving the model effect in the follow-up task by using the reinforcement learning model.
And fourthly, constructing a result auditing and feedback system.
Further: the reinforcement learning model adopts a neural network to calculate a plurality of loss values corresponding to service actions in the current state, and selects the optional action with the largest corresponding loss value from the optional action set as the service action of the current round; applying the business action of the present round to the business environment to obtain the feedback of the present round and the next state of the business environment, which are made by the business environment; calculating a plurality of profit label values corresponding to a plurality of alternative service actions in the current state, wherein the profit label values corresponding to the service actions of the current round are calculated based on the feedback of the current round, the next state and the neural network; for any other action in the plurality of alternative service actions, if the action belongs to the alternative action set, determining a benefit label value as a larger value in a loss value and a first threshold value; if the value belongs to the forbidden action set, determining the profit label value as a smaller value of the loss value and the second threshold value; the first threshold value and the second threshold value are smaller than the income label value corresponding to the business action of the round; the reinforcement learning model is trained based on the plurality of loss values and a plurality of benefit tag values.
Further: the financial document automatic generation system of the GPT model framework comprises a large language model data storage unit, a digital demand management unit, a Chinese vector generation unit, a large language model training unit, a reinforcement learning unit and a data feedback system;
the large language model data storage unit is used for storing business requirement documents, financial documents, text vectors, model scripts, model parameters, model intermediate variables and model effect evaluation values;
the digital demand management unit is used for providing an application platform for compiling, auditing and submitting the type and content requirements of the generated financial document;
the Chinese vector generation unit is used for reading the Chinese demand text from the large language model data storage unit, converting the Chinese demand text into digital vectors and then storing the digital vectors into the large language model data storage unit;
the large language model training unit is deployed on the model service cluster server, a pre-training model of a common document is trained by using a large language pre-training model, after the model training is completed, fine adjustment of a targeted model is carried out according to different types of documents, and the model parameters are stored in the large language model data storage unit and related parameters are called when the model is executed;
the reinforcement learning unit is deployed on the model service cluster server, evaluates and promotes the document effect automatically generated by the model, and meanwhile, the reinforcement learning model also needs to weight the model evaluation value fed back manually, and finally, the correction and promotion of the model effect are completed;
and the data feedback system is responsible for feeding back the generated financial document result to the demand applicant, and the system has functions of auditing, downloading, browsing and the like.
Further: the terminal device may include: the system comprises a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, when the terminal device is running, the processor communicates with the storage medium through the bus, and the processor executes the machine-readable instructions to execute the steps of the deep learning model training method as described in the previous embodiment.
Further: a storage medium storing a computer program which, when executed by a processor, performs the steps of the method described above.
Further: a computer program product comprising a computer program which, when executed by a processor, performs the method described above.
The invention has the beneficial effects that: the invention not only can fully cover business banking business scenes, but also can automatically adapt generated contract clauses according to specific businesses. In addition, on the basis, the invention can automatically generate documents commonly used by commercial banks such as books, industry analysis reports and the like, thereby greatly reducing the repeated workload of first-line business personnel of the commercial banks.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 shows a different entity internet docking scenario.
Fig. 2 shows a schematic diagram of the composition of the system of the present invention.
Fig. 3 shows a schematic diagram of the composition of the terminal device of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present invention, and it should be understood that the drawings in the present invention are for the purpose of illustration and description only and are not intended to limit the scope of the present invention. In addition, it should be understood that the schematic drawings are not drawn to scale. A flowchart, as used in this disclosure, illustrates operations implemented according to some embodiments of the present invention. It should be understood that the operations of the flow diagrams may be implemented out of order and that steps without logical context may be performed in reverse order or concurrently. Moreover, one or more other operations may be added to or removed from the flow diagrams by those skilled in the art under the direction of the present disclosure.
In addition, the described embodiments of the invention are only some, but not all, embodiments of the invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present invention.
It should be noted that the term "comprising" will be used in embodiments of the invention to indicate the presence of the features stated hereafter, but not to exclude the addition of other features. It should also be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. In the description of the present invention, it should also be noted that the terms "first," "second," "third," and the like are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance.
The present invention is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the method for automatically generating financial documents of GPT model architecture according to the present invention includes the steps of:
firstly, building a pre-training model based on a GPT-3 architecture or a LLaMA architecture (opened source), and obtaining a document pre-training model by pre-training an existing document;
specifically, the related documents of the business are arranged, the text, the table text, the text in the image and the like in the documents are classified by using an NLP text classification model according to the business types, the classification result is adjusted by using manpower, the classified text is input into a pre-training model, and the characteristic processing is carried out through an embedded network and a self-attention network in the pre-training model, so that the text representation of the text of the documents is obtained, and further, a financial text pre-training model is obtained;
step two, performing fine adjustment of a specific scene text generation model by using respective type distinguishing data such as contracts, identifications, industry analysis reports and the like;
aiming at the document pre-training model, fine adjustment under a specific scene is carried out, the document pre-training model is specifically obtained firstly, and an enhancement model is constructed based on the document pre-training model, wherein the enhancement model is obtained by adding a multi-view compression representation module between at least two hidden layers of the document pre-training model, the multi-view compression representation model comprises N layers of self-encoders, and N is a positive integer; performing fine tuning training on the enhancement model by using training data to obtain a target model, wherein the target model comprises the enhancement model and a downstream prediction model; updating parameters of the document pre-training model, the multi-view compression representation module and the downstream prediction model in the fine-tuning training process, wherein the training target is to minimize the difference between the output result of the downstream prediction module and an expected value; after the fine tuning training is finished, removing the multi-view compression representation module from the target model obtained by training to obtain a document pre-training model generated by the specific scene text;
thirdly, using a reinforcement learning model to realize that the model automatically improves the model effect in a subsequent task;
the reinforcement learning model adopts a neural network to calculate a plurality of loss values corresponding to service actions in the current state, and selects the optional action with the largest corresponding loss value from the optional action set as the service action of the current round; applying the business action of the present round to the business environment to obtain the feedback of the present round and the next state of the business environment, which are made by the business environment; calculating a plurality of profit label values corresponding to a plurality of alternative service actions in the current state, wherein the profit label values corresponding to the service actions of the current round are calculated based on the feedback of the current round, the next state and the neural network; for any other action in the plurality of alternative service actions, if the action belongs to the alternative action set, determining a benefit label value as a larger value in a loss value and a first threshold value; if the value belongs to the forbidden action set, determining the profit label value as a smaller value of the loss value and the second threshold value; the first threshold value and the second threshold value are smaller than the income label value corresponding to the business action of the round; the reinforcement learning model is trained based on the plurality of loss values and a plurality of benefit tag values.
And fourthly, constructing a result auditing and feedback system.
As shown in fig. 2, the system is divided into six units, including a large language model data storage unit, a number demand management unit, a Chinese vector generation unit, a large language model training unit, a reinforcement learning unit and a data feedback system. I will develop a detailed description of:
and a large language model data storage unit. And storing business requirement documents, financial documents, text vectors, model scripts, model parameters, model intermediate variables and model effect evaluation values.
And a digital demand management unit. The system aims to provide an application platform for compiling, auditing and submitting the requirements of the types and contents of financial documents for first-line business personnel of commercial banks, wherein the business personnel compiles detailed descriptions of the required documents according to business requirements, and after auditing by business director, the business personnel transfer to a follow-up model execution flow, and relevant required texts are saved in a large language model data storage unit.
And a Chinese vector generation unit. And a Word2Vec compressing module is deployed and is responsible for reading the Chinese demand text from the large language model data storage unit, converting the Chinese demand text into a digital vector and then storing the digital vector into the large language model data storage unit.
And a large language model training unit. The unit is deployed on a model service cluster server, and a large language pre-training model based on a transducer model, such as GPT-3 or LLaMA, is used for training a pre-training model aiming at common documents in the financial industry. After the model training is completed, fine adjustment of the targeted model is carried out according to different types of financial documents, the model parameters are stored in a large language model data storage unit, and related parameters are called when the model is executed.
And a reinforcement learning unit. The unit is deployed on a model service cluster server, the effect of the model automatically generated by using human feedback Reinforcement Learning (RLHF) is evaluated and improved, meanwhile, the reinforcement learning model also needs to weight the model evaluation value of the human feedback, and finally, the correction and improvement of the model effect are completed.
And a data feedback system. The system is responsible for feeding back the generated financial document result to the demand applicant, and the system has functions of auditing, downloading, browsing and the like.
As shown in fig. 3, the terminal device 6 may include: processor 601, storage medium 602, and bus 603, storage medium 602 storing machine-readable instructions executable by processor 601, when the terminal device is running, the processor 601 communicates with storage medium 602 via bus 603, and processor 601 executes the machine-readable instructions to perform the steps of the deep learning model training method as described in the previous embodiments. The specific implementation manner and the technical effect are similar, and are not repeated here.
For ease of illustration, only one processor is described in the above terminal device. It should be noted, however, that in some embodiments, the terminal device of the present invention may also include multiple processors, and thus, the steps performed by one processor described in the present invention may also be performed jointly by multiple processors or separately.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily appreciate variations or alternatives within the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (9)
1. The financial document automatic generation method based on the GPT model architecture is characterized by comprising the following specific steps:
firstly, building a pre-training model, and obtaining a document pre-training model by pre-training an existing document;
specifically, the related documents of the business are arranged, the text, the table text, the text in the image and the like in the documents are classified by using an NLP text classification model according to the business types, the classification result is adjusted by using manpower, the classified text is input into a pre-training model, and the characteristic processing is carried out through an embedded network and a self-attention network in the pre-training model, so that the text representation of the text of the documents is obtained, and further, a financial text pre-training model is obtained;
step two, building an image theme extraction model, and extracting key useful information from the picture through an image recognition technology:
specifically, key useful information in the picture is extracted through an image recognition technology so as to increase a corpus of the pre-training model;
step three, a special system for text labeling is built, text labeling efficiency is improved, and labeling of contents such as text categories, association relations, standard answers, processing results and the like is achieved:
specifically, the system automatically provides one to five-level document classification categories, all documents are manually input into the system and the category of the document is selected to be saved, after the text is input into the system, the system provides an operation interface which enables a annotator to mark related texts in a dragging mode on a system interface and marks sub-categories and affiliations among the documents, and for isolated problems without relevance, affiliation, category information and reply content, all the annotation information can be manually annotated and document categories can be newly established;
step four, using the respective type distinguishing data to make fine adjustment of a specific scene text generation model;
specifically, the image recognition model is utilized to recognize image characters, the classified financial documents are added, then manual text labeling is carried out on all the classified documents, then fine tuning training is carried out on the pre-training model based on the labeled text, and the text generation accuracy of the model on a specific scene is improved by utilizing a supervised learning method;
fifthly, using a human feedback reinforcement learning model to realize that the model automatically improves the model effect in a subsequent task;
specifically, after the user uses the financial text generated by the invention, the user needs to manually modify and adjust the financial text according to the actual business scene, and finally the document which is finally modified and put into business use is automatically uploaded to a model background by the system and put into a reinforcement learning model after being classified by the system, so that the generation effect of the model text is improved;
step six, an AI image self-training model module is realized by using the CLIP model, and further, a flow chart, a system architecture diagram, a physical architecture diagram and a PPT diagram are automatically generated according to the description of the requirement;
and step seven, constructing a result auditing and feedback system.
2. The method of claim 1, wherein the document related to the total financial business stored in the commercial bank is classified by using a self-training NLP text classification model with the business type such as financial service contract, bidding, internal official document, external notice, regulation, etc., and manually rechecked and fine-tuned after the classification is completed. Meanwhile, training a pre-training model based on the GPT framework by utilizing the full quantity of financial business related documents.
3. The method of claim 1, wherein the text in the picture is extracted from the relevant document using image recognition techniques and the document classification is added to the training model corpus.
4. The method of claim 1, wherein the classification result of the full-scale financial business related document is manually marked, including text content marking and question-answer result marking, and the marking text is used for performing fine-tuning training on the pre-training model based on the GPT architecture to promote the actual effect of the model in generating the financial document for the business scene.
5. The method according to claim 1, wherein the selectable action with the largest corresponding loss value is selected from the selectable action set as the current round of business action; applying the business action of the present round to the business environment to obtain the feedback of the present round and the next state of the business environment, which are made by the business environment; calculating a plurality of profit label values corresponding to a plurality of alternative service actions in the current state, wherein the profit label values corresponding to the service actions of the current round are calculated based on the feedback of the current round, the next state and the neural network; for any other action in the plurality of alternative service actions, if the action belongs to the alternative action set, determining a benefit label value as a larger value in a loss value and a first threshold value; if the value belongs to the forbidden action set, determining the profit label value as a smaller value of the loss value and the second threshold value; the first threshold value and the second threshold value are smaller than the income label value corresponding to the business action of the round.
6. A financial document automatic generation system of a GPT model architecture, comprising: the system comprises a large language model data storage unit, a number demand management unit, a Chinese vector generation unit, a large language model training unit, a reinforcement learning unit and a data feedback system;
the large language model data storage unit is used for storing business requirement documents, financial documents, text vectors, model scripts, model parameters, model intermediate variables and model effect evaluation values;
the digital demand management unit is used for providing an application platform for compiling, auditing and submitting the type and content requirements of the generated financial document;
the Chinese vector generation unit is used for reading the Chinese demand text from the large language model data storage unit, converting the Chinese demand text into digital vectors and then storing the digital vectors into the large language model data storage unit;
the large language model training unit is deployed on the model service cluster server, a pre-training model of a common document is trained by using a large language pre-training model, after the model training is completed, fine adjustment of a targeted model is carried out according to different types of documents, and the model parameters are stored in the large language model data storage unit and related parameters are called when the model is executed;
the reinforcement learning unit is deployed on the model service cluster server, evaluates and promotes the document effect automatically generated by the model, and meanwhile, the reinforcement learning model also needs to weight the model evaluation value fed back manually, and finally, the correction and promotion of the model effect are completed;
and the data feedback system is responsible for feeding back the generated financial document result to the demand applicant, and the system has functions of auditing, downloading, browsing and the like.
7. A terminal device, comprising: a processor, a storage medium and a bus, the storage medium storing machine-readable instructions executable by the processor, the processor and the storage medium communicating over the bus when the terminal device is operating, the processor executing the machine-readable instructions to perform the steps of the method of any of claims 1 to 4 when executed.
8. A storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method according to any of claims 1 to 4.
9. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311388756.3A CN117350253A (en) | 2023-10-25 | 2023-10-25 | Financial document automatic generation method and system of GPT model architecture |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311388756.3A CN117350253A (en) | 2023-10-25 | 2023-10-25 | Financial document automatic generation method and system of GPT model architecture |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117350253A true CN117350253A (en) | 2024-01-05 |
Family
ID=89361033
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311388756.3A Pending CN117350253A (en) | 2023-10-25 | 2023-10-25 | Financial document automatic generation method and system of GPT model architecture |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117350253A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118898663A (en) * | 2024-09-30 | 2024-11-05 | 深圳市智慧城市科技发展集团有限公司 | Conference material generation method, device and storage medium |
-
2023
- 2023-10-25 CN CN202311388756.3A patent/CN117350253A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118898663A (en) * | 2024-09-30 | 2024-11-05 | 深圳市智慧城市科技发展集团有限公司 | Conference material generation method, device and storage medium |
| CN118898663B (en) * | 2024-09-30 | 2025-04-04 | 深圳市智慧城市科技发展集团有限公司 | Conference material generation method, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US12159112B2 (en) | Hierarchical machine learning architecture including master engine supported by distributed light-weight real-time edge engines | |
| US11281860B2 (en) | Method, apparatus and device for recognizing text type | |
| CN110765765B (en) | Contract key term extraction method, device and storage medium based on artificial intelligence | |
| CN113392209A (en) | Text clustering method based on artificial intelligence, related equipment and storage medium | |
| CN104978587B (en) | A kind of Entity recognition cooperative learning algorithm based on Doctype | |
| CN111767403A (en) | Text classification method and device | |
| CN111339303B (en) | Text intention induction method and device based on clustering and automatic abstracting | |
| CN113886553B (en) | A text generation method, device, equipment and storage medium | |
| CN113806537B (en) | Commodity category classification method and its device, equipment, medium, and product | |
| Riba et al. | Table detection in business document images by message passing networks | |
| CN113761910A (en) | Comment text fine-grained emotion analysis method integrating emotional characteristics | |
| CN118733748B (en) | Training method and device for generating large language model and value system identification application | |
| CN113449808A (en) | Multi-source image-text information classification method and corresponding device, equipment and medium | |
| CN113837307A (en) | Data similarity calculation method and device, readable medium and electronic equipment | |
| CN113449103A (en) | Bank transaction flow classification method and system integrating label and text interaction mechanism | |
| CN112287215A (en) | Intelligent employment recommendation method and device | |
| CN111930944B (en) | File label classification method and device | |
| CN109635289B (en) | Entries Classification Method and Audit Information Extraction Method | |
| CN117350253A (en) | Financial document automatic generation method and system of GPT model architecture | |
| CN113590755A (en) | Word weight generation method and device, electronic equipment and storage medium | |
| CN115577051A (en) | Data item classification method and device, electronic equipment and storage medium | |
| CN120196736B (en) | Supply and demand matching method for digital science and technology personalized service | |
| CN119808866B (en) | Training method of search word generation model, search word generation method and device | |
| CN120876124A (en) | Method, device and storage medium for generating electronic bill | |
| CN119312792A (en) | Text annotation method, system and computer device |
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 |