A device for automatically generating entity training results by artificial sentence formation

By automating the data collection and labeling process, the problem of cumbersome and time-consuming traditional data labeling methods is solved, thereby improving the accuracy and efficiency of model training.

CN118734811BActive Publication Date: 2024-11-15JIANGSU ZHONGWEI TECH SOFTWARE SYST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411215783.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2024-11-15
Estimated Expiration
2044-09-02

AI Technical Summary

Technical Problem

Traditional data annotation methods are cumbersome and time-consuming, making it difficult to efficiently annotate and retrieve text content, which affects model training efficiency.

Method used

The system employs a data collection module, a data analysis module, an entity training module, and an entity verification module. Data is automatically labeled and verified using Excel spreadsheets and databases, and the training model is optimized by combining online and offline update modules.

Benefits of technology

It automates and improves the efficiency of data annotation, enhances the accuracy and efficiency of training models, and simplifies the data collection and annotation process.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a device for automatically generating entity training results by artificial sentence making, comprising a data collection module, a data analysis module, an entity training module, an entity verification module and a data update module. The data collection module uses Excel to collect data and establishes data list rules; the data analysis module analyzes the collected data and automatically marks the data; the entity training module performs rule field conversion on the data, obtains training model data, inputs the training model data into the training model, and performs training; the entity verification module performs double-layer verification on the marked entity data; the data update module performs secondary training on the executed data, and specifically covers and solves the original erroneous data online. The invention innovates a method that can directly use a data table or a database, and establishes a logical relationship between a training object and a training element while establishing a training model for artificial sentence making, thereby automatically achieving a training goal, which is beneficial to improving the training accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a device for automatically generating entity training results by artificial sentence formation. Background Art

[0002] In the field of machine learning, model training usually requires a large amount of labeled data as input. Traditional data labeling methods often involve multiple links, including data collection, organization, labeling and review. The collection and organization are too complicated, which is not conducive to subsequent labeling work. In addition, during the labeling process, the text content is usually manually labeled online, which is complicated, time-consuming and difficult to retrieve. Summary of the invention

[0003] The purpose of the present invention is to provide a device for automatically generating entity training results by artificial sentence formation, so as to solve the problems raised in the above-mentioned background technology.

[0004] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a device for automatically generating entity training results by artificial sentence formation, comprising a data collection module, a data analysis module, an entity training module, and an entity verification module.

[0005] The data collection module uses Excel to collect data and set up row and column rules in the data list;

[0006] The data analysis module analyzes and obtains the data collected and imported into the system, and automatically annotates the data;

[0007] The entity training module is used to convert the above data into rule fields, obtain training model data, input the training model data into the training model, and perform model training;

[0008] The entity verification module is used to perform double-layer verification on the marked entity data, and if the data is incorrect, it is marked again;

[0009] The data updating module is used to perform secondary training on the executed data in an online or offline state, and to cover the original data in a targeted manner.

[0010] Preferably, the row and column rules in the list in the data collection module are that the first row represents the content that needs to be marked, wherein in the first row of fields, the first field represents the sentence content, and subsequent fields represent the content information that needs to be collected and marked for the sentence; the column field rules can be set to represent time, ID card, and telephone number.

[0011] Preferably, the specific analysis process of the data analysis module is: analyzing the Excel data imported into the system, automatically parsing the table header in Excel, obtaining the entity attributes corresponding to the table header, and automatically marking the entities on the statements in the system that imports the data.

[0012] Preferably, the entity training module performs rule field conversion on the original text corresponding to the annotation and the entity association information, wherein the field is converted into json format, and then the training model data is obtained and directly connected to the model. At the same time, the training model data can be updated, deleted and adjusted in real time.

[0013] Preferably, the original text corresponding to the annotation refers to the content used for sentence construction; the entity association information refers to the content of each information to be extracted.

[0014] Preferably, when performing entity analysis, the entity verification module first automatically marks the input statement as an entity, and then performs a double-layer verification on the marked entity data, wherein if the data is incorrect, the entity or specific data is marked in the marking system; the double-layer verification refers to restrictions in the Excel spreadsheet and the database respectively.

[0015] Preferably, the data update module includes an online data update module and an offline data update module; the online data update module refers to the data executed in the system being corrected in an online state, and then directly imported into the training model for secondary training, so as to cover and solve the original erroneous data in a targeted manner; the offline data update module refers to the data after training, using the data content inferred by the training model, reversely exporting it as Excel, and after debugging and marking the data, importing it into the training model for a secondary time, and then training again, and covering and solving the original erroneous data.

[0016] Preferably, the training result information can be automatically summarized and counted, and based on the training results, the templates of the data training statements that are missing in the autonomous analysis model can be used to supplement the new data templates for displaying the data information.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] The present invention innovates a method that can directly use a data table or a database to establish a training model by manually constructing sentences, and at the same time, establish a logical relationship between the training object and the training elements, thereby automatically achieving the training goal, which is beneficial to improving the accuracy of training and improving efficiency. DETAILED DESCRIPTION

[0019] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Example

[0020] The present invention provides a device for automatically generating entity training results by artificial sentence formation, comprising a data collection module, a data analysis module, an entity training module, and an entity verification module.

[0021] The data collection module uses an Excel table to collect data and establish row and column rules in the data list. When setting the list rules, the row and column rules in the list are that the first row represents the content that needs to be marked, where in the first row field, the first field represents the sentence content, and the subsequent fields represent the content information that needs to be collected and marked for the sentence; the column field rule can be set to represent time, ID card, telephone number, etc. By setting the row and column rules, the accuracy of the marked content or format can be guaranteed. For example, when using Excel to collect data, the first field in the first row of the table can be set to represent "add schedule sentence", then "add schedule sentence" is the sentence content, which is the content that needs to be trained, and the subsequent fields of the first row can be set to represent "importance", "schedule time", "business content", so as to indicate that the content related to "importance", "schedule time", and "business content" needs to be extracted and marked from the "add schedule sentence", that is, the content to be trained;

[0022] After data collection, the data analysis module is used to analyze and obtain the collected data imported into the system, and the data is automatically labeled. The specific analysis process is: analyze the Excel data imported into the system, automatically parse the table header in Excel, obtain the entity attributes corresponding to the table header, and automatically label the entities on the statement in the system that imports the data. For example, according to the above rule "the first field in the first row of the table represents "add schedule statement", and the subsequent fields of the first row can be set to represent "importance", "schedule time", "transaction content", if the content to be trained entered in the field representing "add schedule statement" is "add an important schedule, hold a promotion meeting this afternoon", then "importance", "this afternoon", "hold a promotion meeting" are labeled from the input statement according to "importance", "transaction content", and "transaction content" respectively;

[0023] The entity training module converts the data analyzed in the data analysis module into regular fields, thereby converting the original text corresponding to the annotation and the entity association information into regular fields, obtaining the training model data, and directly inputting the training model data into the training model to train the training model. At the same time, the training model data can be updated, deleted, and adjusted in real time;

[0024] When converting in the rule field, the conversion format is json format, for example:

[0025] [{

[0026] "text": "I plan to take a day off next Monday to deal with some family matters.",

[0027] "result": [{

[0028] "annotation": "time",

[0029] "content": "Next Monday",

[0030] "position": "4,7"

[0031] }]

[0032] }, {

[0033] "text": "I need to take a sick day next Tuesday because I have an appointment to see a doctor.",

[0034] "result": [{

[0035] "annotation": "time",

[0036] "content": "Next Tuesday",

[0037] "position": "0,3"

[0038] }]

[0039] }];

[0040] After the data is marked, the entity verification module performs a double-level verification on the marked entity data. The double-level verification refers to the restrictions in the Excel table and the database respectively. Specifically, when analyzing the entity, the entity verification module first automatically marks the input statement as an entity, and then performs a double-level verification on the marked entity data. If the data is incorrect, there will be problems with the marked entity or the marked specific data in the marking system.

[0041] When data is marked as problematic, the data can be updated using the data update module, which includes an online data update module and an offline data update module. Users can choose a more appropriate data update module when using it. If the online data update module is online, the executed data is trained twice, and the original erroneous data is covered online in a targeted manner.

[0042] The data offline update module is for the trained data in an offline state. It uses the data content inferred by the training model to reversely export it as it is in Excel. After debugging and marking the data through various means such as format, drop-down format, length, etc., it is imported into the model for the second time and then trained again, so as to cover and solve the original erroneous data.

[0043] The result information after manual sentence training can be automatically summarized and counted. According to the results of model training, the template of the data training sentence that is lacking in the model can be analyzed autonomously, and the new data template can be supplemented to display the data information. Therefore, the present invention can directly use the data table or database to establish the logical relationship between the training object and the training elements while establishing the training model for manual sentence creation, so as to automatically achieve the training goal, which is conducive to improving the accuracy of training and improving efficiency.

[0044] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments are to be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the present invention.

Claims

1. A device for automatically generating entity training results by artificial sentence formation, Features: Including data collection module, data analysis module, entity training module, entity verification module, data update module, The data collection module uses Excel to collect data and set up row and column rules in the data list. The row and column rules in the list in the data collection module are that the first row represents the content that needs to be marked, wherein in the first row field, the first field represents the sentence content, and the subsequent fields represent the content information that needs to be collected and marked for the sentence; the column field rule can be set to represent time, ID card, and telephone number; The data analysis module analyzes and obtains the collected data imported into the system, and automatically annotates the data. The specific analysis process of the data analysis module is: analyzing the Excel data imported into the system, automatically parsing the table header in Excel, obtaining the entity attributes corresponding to the table header, and automatically annotating the entities on the statements in the system that imported the data; The entity training module is used to convert the above data into rule fields, obtain training model data, input the training model data into the training model, and perform model training; The entity verification module is used to perform double-layer verification on the marked entity data, and if the data is incorrect, it is marked again; The data update module is used to perform secondary training on the executed data in an online or offline state, and to cover the original data in a targeted manner; the data update module includes an online data update module and an offline data update module; The data online update module is in an online state. After the executed data is corrected in the system, it is directly imported into the training model for secondary training, so as to cover and solve the original erroneous data in a targeted manner. The data offline update module refers to the data content inferred by the training model for the trained data, which is reverse exported as Excel, and after error correction and data labeling, it is imported into the training model for the second time, and then trained again to overwrite and resolve the original erroneous data.

2. The device for automatically generating entity training results by artificial sentence formation according to claim 1, Features: In the entity training module, the original text corresponding to the annotation and the entity association information are converted into regular fields, wherein the fields are converted into json format, and then the training model data is obtained and directly connected to the model. At the same time, the training model data can be updated, deleted and adjusted in real time.

3. The device for automatically generating entity training results by artificial sentence formation according to claim 1, Features: The original text corresponding to the annotation refers to the content used for sentence formation; the entity-related information refers to the content of each piece of information to be extracted.

4. The device for automatically generating entity training results by artificial sentence formation according to claim 1, Features: When analyzing entities, the entity verification module first automatically marks the input statements as entities, and then performs a double-level verification on the marked entity data. If the data is incorrect, the entity or specific data is marked in the marking system; the double-level verification refers to restrictions in the Excel table and the database respectively.

5. The device for automatically generating entity training results by artificial sentence formation according to claim 1, Features: The training result information can be automatically summarized and counted. According to the training results, the template of the data training statement that is missing in the autonomous analysis model can be used to supplement the new data template to display the data information.