CN110334080B - Knowledge base construction method for realizing autonomous learning - Google Patents
Knowledge base construction method for realizing autonomous learning Download PDFInfo
- Publication number
- CN110334080B CN110334080B CN201910562032.3A CN201910562032A CN110334080B CN 110334080 B CN110334080 B CN 110334080B CN 201910562032 A CN201910562032 A CN 201910562032A CN 110334080 B CN110334080 B CN 110334080B
- Authority
- CN
- China
- Prior art keywords
- knowledge
- knowledge point
- model
- new
- points
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/211—Schema design and management
-
- 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
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Human Computer Interaction (AREA)
- Artificial Intelligence (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The invention discloses a knowledge base construction method for realizing autonomous learning, which comprises the steps of obtaining a plurality of knowledge points, inputting the knowledge points into a knowledge point identification model to generate a candidate knowledge point set, inputting a candidate new knowledge point set into a new knowledge point judgment model to generate a new knowledge point set, then filtering the new knowledge point set to generate a first set of all knowledge points which are new knowledge points, and obtaining a first accuracy rate of learning the new knowledge points according to the first set and the new knowledge point set, when the first accuracy reaches a preset threshold value, the knowledge point identification model is combined with the new knowledge point judgment model to obtain an autonomous learning knowledge base, and by adopting the embodiment provided by the invention, the knowledge base can automatically distinguish similar knowledge points, and new knowledge is found from unknown data, manual guidance is continuously reduced, and the working efficiency is greatly improved.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a knowledge base construction method for realizing autonomous learning.
Background
The dialog knowledge base is an important knowledge base in the dialog system, and contains a large number of artificial ideas and actually accumulated knowledge points (questions and corresponding answers which are concerned by a user). Data can be classified from a large amount of dialogue data through various clustering or classification algorithms and manual consulting methods, and each class corresponds to a problem concerned by a user.
However, in the existing algorithm, similar but different knowledge points cannot be distinguished, low-frequency but important new knowledge points cannot be found, a large amount of manual guidance is needed, the labor cost is too high, and self-learning cannot be achieved.
Disclosure of Invention
The embodiment of the invention aims to provide a knowledge base construction method for realizing autonomous learning, which can distinguish similar knowledge points, automatically discover new knowledge from unknown data, continuously reduce manual guidance and greatly improve the working efficiency.
In order to achieve the above object, an embodiment of the present invention provides a knowledge base construction method for implementing autonomous learning, including the following steps:
acquiring a plurality of knowledge points, inputting the knowledge points to a pre-established knowledge point identification model, and generating a candidate knowledge point set; the knowledge points comprise questions corresponding to the knowledge and answers corresponding to the knowledge point questions;
inputting the candidate new knowledge point set into a pre-established new knowledge point judgment model to generate a new knowledge point set;
filtering the new knowledge point set to generate a first set of which all knowledge points are new knowledge points;
obtaining a first accuracy rate of learning new knowledge points according to the first set and the new knowledge point set, and judging whether the first accuracy rate reaches a preset threshold value;
if so, combining the knowledge point identification model with the new knowledge point judgment model to obtain an autonomous learning knowledge base;
if not, optimizing the knowledge point identification model and the new knowledge point judgment model so as to enable the accuracy rate of learning new knowledge points to reach the preset threshold value.
Further, the pre-established knowledge point identification model is constructed by the following method:
taking a BERT text classification model as a first basic model, and acquiring a plurality of common knowledge point sets;
carrying out positive and negative classification on each knowledge point problem in the common knowledge point sets to obtain a positive knowledge point set belonging to required knowledge points and a negative knowledge point set belonging to unneeded knowledge points;
and taking the common knowledge point sets as the input of the first basic model, taking the positive knowledge point set as the output of the first basic model, and taking the first basic model as a knowledge point identification model when the identification accuracy of the first basic model reaches a first threshold value.
Further, the pre-established new knowledge point judgment model is constructed by the following method:
taking the BERT text similarity model as a second basic model, and acquiring a plurality of new knowledge points and a plurality of common knowledge point sets;
respectively calculating the problems of the new knowledge points according to a similarity algorithm, and obtaining a similarity set with the similarity exceeding a second threshold value and a dissimilarity set with the similarity not higher than the second threshold value by the similarity of the problems of the new knowledge points and the problems of each knowledge point in the common knowledge point sets;
and constructing a new knowledge point judgment model by taking the new knowledge points and the common knowledge point sets as the input of the second basic model and the dissimilar sets as the output of the second basic model.
Further, the optimizing the knowledge point identification model and the new knowledge point determination model specifically includes:
when the first accuracy rate does not reach a preset threshold value, improving the first threshold value in the knowledge point identification model so that the knowledge point identification model identifies a more needed knowledge point;
reducing a second threshold in the new knowledge point determination model such that the new knowledge point determination model determines more dissimilar knowledge points.
Further, the required knowledge point is a question of interest to the user and an answer to the question of interest to the user.
Further, the similarity algorithm is a binary algorithm.
Further, the preset threshold is 0.95.
Further, the first threshold is 0.8.
Further, the second threshold is 0.2.
Compared with the prior art, the method has the following beneficial effects:
the method for constructing the knowledge base for realizing the autonomous learning comprises the steps of obtaining a plurality of knowledge points, inputting the knowledge points into a knowledge point identification model to generate a candidate knowledge point set, inputting a candidate new knowledge point set into a new knowledge point judgment model to generate a new knowledge point set, filtering the new knowledge point set to generate a first set of which all knowledge points are new knowledge points, obtaining a first accuracy rate of learning the new knowledge points according to the first set and the new knowledge point set, and judging whether the first accuracy rate reaches a preset threshold value or not; if so, combining the knowledge point identification model with the new knowledge point judgment model to obtain an autonomous learning knowledge base; if not, the knowledge point identification model and the new knowledge point judgment model are optimized, so that the accuracy rate of learning new knowledge points reaches a preset threshold value.
Drawings
Fig. 1 is a schematic flow chart of a knowledge base construction method for implementing autonomous learning according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, fig. 1 is a schematic flowchart of an embodiment of a knowledge base construction method for implementing autonomous learning according to the present invention; the embodiment of the invention provides a knowledge base construction method for realizing autonomous learning, which comprises the steps of S1-S6;
and S1, acquiring a plurality of knowledge points, inputting the knowledge points into a pre-established knowledge point identification model, and generating a candidate knowledge point set.
The knowledge point comprises a question corresponding to the knowledge and an answer corresponding to the knowledge point question.
In this embodiment, the pre-established knowledge point identification model is constructed by the following method: taking a BERT text classification model as a first basic model, and acquiring a plurality of common knowledge point sets; carrying out positive and negative classification on each knowledge point problem in the common knowledge point sets to obtain a positive knowledge point set belonging to required knowledge points and a negative knowledge point set belonging to unneeded knowledge points; and taking the common knowledge point sets as the input of the first basic model, taking the positive knowledge point set as the output of the first basic model, and taking the first basic model as a knowledge point identification model when the identification accuracy of the first basic model reaches a first threshold value.
It should be noted that the required knowledge point is a question concerned by the user and an answer to the question concerned by the user, and the first threshold is optimally 0.8. In an experiment, the higher the first threshold value is, the more the obtained knowledge points are needed by the user, but the too high threshold value can cause the working efficiency of the model to be too low and the obtained knowledge points to be less; however, when the first threshold is 0.8, the knowledge points required by the user can be obtained, and the working efficiency of the model is not affected.
And S2, inputting the candidate new knowledge point set into a pre-established new knowledge point judgment model to generate a new knowledge point set.
In this embodiment, the pre-established new knowledge point determination model is constructed by the following method: taking the BERT text similarity model as a second basic model, and acquiring a plurality of new knowledge points and a plurality of common knowledge point sets; respectively calculating the problems of the new knowledge points according to a similarity algorithm, and obtaining a similarity set with the similarity exceeding a second threshold value and a dissimilarity set with the similarity not higher than the second threshold value by the similarity of the problems of the new knowledge points and the problems of each knowledge point in the common knowledge point sets; and constructing a new knowledge point judgment model by taking the new knowledge points and the common knowledge point sets as the input of the second basic model and the dissimilar sets as the output of the second basic model.
Wherein the similarity algorithm is a binary algorithm, and the second threshold is optimally 0.2. In an experiment, the lower the second threshold value is, the more dissimilar the obtained dissimilar knowledge points are, but the too low threshold value is the too low similarity, so that the working efficiency of the model is too low, and the obtained knowledge points are too few; however, when the second threshold is 0.2, the dissimilar knowledge points required by the user can be obtained, and the working efficiency of the model is not influenced.
As a preferred embodiment of the invention, the invention can also train another new knowledge point judgment model by using a BERT text similarity model for distinguishing the new knowledge point from other knowledge points as non-homogeneous knowledge points, the training method is to use the problems of the homogeneous knowledge points as similar texts and the problems of different knowledge points as dissimilar texts, splicing the problems together two by two, then judging whether the two problems are similar by using a binary model after BERT coding, and using the probability value of the positive label finally output by the model as the similarity. Each candidate knowledge point is distinguished from the common knowledge point set from the candidate knowledge point set through the new knowledge point judging model, the knowledge point with the highest score is selected to be judged as a new knowledge point, and the knowledge point with the lowest score is judged as a non-new knowledge point, wherein the score calculating method comprises the following steps: and (3) taking the maximum value of the similarity of the candidate knowledge problem and each problem in the common knowledge point set as the similarity of the new knowledge and the common knowledge set, and then using the (1-similarity) to obtain a score. Since the higher the similarity, the more relevant the new knowledge is to the common knowledge points, the score should be lower.
And S3, filtering the new knowledge point set to generate a first set of which all knowledge points are new knowledge points.
S4, obtaining a first accuracy rate of learning new knowledge points according to the first set and the new knowledge point set, and judging whether the first accuracy rate reaches a preset threshold value.
Note that, the preset threshold is 0.95. In the experiment, when the selected threshold value is 0.95, namely 95% of the extracted new knowledge is correct, the effect of finding and classifying the new knowledge points is the best.
And S5, if yes, combining the knowledge point identification model with the new knowledge point judgment model to obtain an autonomous learning knowledge base.
And S6, if not, optimizing the knowledge point identification model and the new knowledge point judgment model so that the accuracy of learning new knowledge points reaches the preset threshold value.
As a preferred embodiment of the present invention, when the first accuracy does not reach a preset threshold, the first threshold in the knowledge point identification model is increased, so that the knowledge point identification model identifies a more required knowledge point; and reducing a second threshold value in the new knowledge point judgment model so that the new knowledge point judgment model judges more dissimilar knowledge points, and the accuracy rate of learning new knowledge points can reach a preset threshold value.
Preferably, in the process of training the model, the classification performance of the candidate knowledge points can be improved by optimizing and improving the first threshold in the knowledge point identification model, so that the candidate knowledge points with higher quality are extracted; by optimizing and reducing the second threshold of the new knowledge point judgment model, the judgment performance of the new knowledge points can be improved, and the quality of the knowledge point set is improved.
In order to better explain the working principle of the invention, the following steps of the working principle of one embodiment provided by the invention comprise an initialization stage, a knowledge point discovery stage, a knowledge point judgment stage, a manual review stage, a system optimization stage and an evaluation and iteration stage;
an initialization stage: the method comprises the steps of manually marking a plurality of common knowledge point sets and a new knowledge point, training a knowledge point identification model based on a BERT text binary classification model, and identifying whether input data are knowledge points required by a user, wherein the knowledge points required by the user can be set in a self-defining mode through a background.
And then training a new knowledge point judgment model based on the BERT text similarity model, wherein the new knowledge point judgment model is used for judging whether the knowledge points in the candidate knowledge point set generated by the knowledge point identification model are new knowledge points.
A knowledge point discovery stage: and identifying positive knowledge points from the acquired data through a knowledge point identification model, and taking the positive knowledge points as a candidate knowledge point set.
A knowledge point judging stage: and inputting the candidate knowledge point set and the common knowledge point set into the new knowledge point judgment model to obtain a knowledge point set which is dissimilar to the common knowledge point set and is used as a new knowledge point set.
And (3) a manual auditing stage: and manually checking new knowledge points and non-new knowledge points in the knowledge point judgment.
And (3) an optimization stage: the working performance of the model is improved while the manual participation is continuously reduced. In the above stage, the collected new knowledge points are added to the knowledge point recognition model again for training as new data.
Evaluation and iteration stage: and evaluating the accuracy of the judgment results of the new knowledge points and the non-new knowledge points in the optimization stage. Different accuracy rates correspond to different states, when the accuracy rate is high, the knowledge base construction technology can be completed with high quality by completely depending on the two models, when the accuracy rate is low, the knowledge base construction can not be directly carried out by using the current two models, and manual guidance is required to be added. In the iteration process, the above stages need to be circulated continuously until the two models can construct a knowledge base with high accuracy, the accuracy for measuring the critical value is an adjustable threshold, and when the threshold is too high and the manual quitting time is delayed, the self-learning performance of the system is low, but the extracted knowledge quality is high; when the threshold value is too low, the manual quitting is advanced, the self-learning performance of the system is high, and the extracted knowledge quality is low.
The method for constructing the knowledge base for realizing the autonomous learning comprises the steps of obtaining a plurality of knowledge points, inputting the knowledge points into a knowledge point identification model to generate a candidate knowledge point set, inputting a candidate new knowledge point set into a new knowledge point judgment model to generate a new knowledge point set, filtering the new knowledge point set to generate a first set of which all knowledge points are new knowledge points, obtaining a first accuracy rate of learning the new knowledge points according to the first set and the new knowledge point set, and judging whether the first accuracy rate reaches a preset threshold value or not; if so, combining the knowledge point identification model with the new knowledge point judgment model to obtain an autonomous learning knowledge base; if not, the knowledge point identification model and the new knowledge point judgment model are optimized, so that the accuracy rate of learning new knowledge points reaches a preset threshold value.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention.
Claims (7)
1. A knowledge base construction method for realizing autonomous learning is characterized by comprising the following steps:
acquiring a plurality of knowledge points, inputting the knowledge points to a pre-established knowledge point identification model, and generating a candidate knowledge point set; the knowledge points comprise questions corresponding to the knowledge and answers corresponding to the knowledge point questions; the pre-established knowledge point identification model is constructed by the following method:
taking a BERT text classification model as a first basic model, and acquiring a plurality of common knowledge point sets;
carrying out positive and negative classification on each knowledge point problem in the common knowledge point sets to obtain a positive knowledge point set belonging to required knowledge points and a negative knowledge point set belonging to unneeded knowledge points;
taking the common knowledge point sets as the input of the first basic model, taking the positive knowledge point set as the output of the first basic model, and taking the first basic model as a knowledge point identification model when the identification accuracy of the first basic model reaches a first threshold value;
inputting the candidate knowledge point set into a pre-established new knowledge point judgment model to generate a new knowledge point set; the pre-established new knowledge point judgment model is constructed by the following method:
taking the BERT text similarity model as a second basic model, and acquiring a plurality of new knowledge points and a plurality of common knowledge point sets;
respectively calculating the problems of the new knowledge points according to a similarity algorithm, and obtaining a similar set with the similarity exceeding a second threshold value and a dissimilar set with the similarity not higher than the second threshold value by the similarity of the problems of the new knowledge points and the problems of each knowledge point in the common knowledge point sets;
constructing a new knowledge point judgment model by taking the plurality of new knowledge points and the plurality of common knowledge point sets as the input of the second basic model and the dissimilar sets as the output of the second basic model;
filtering the new knowledge point set through manual examination to generate a first set of which all knowledge points are new knowledge points;
obtaining a first accuracy rate of learning new knowledge points according to the first set and the new knowledge point set, and judging whether the first accuracy rate reaches a preset threshold value;
if so, combining the knowledge point identification model with the new knowledge point judgment model to obtain an autonomous learning knowledge base;
if not, optimizing the knowledge point identification model and the new knowledge point judgment model so as to enable the accuracy rate of learning new knowledge points to reach the preset threshold value.
2. The method for constructing a knowledge base for implementing autonomous learning according to claim 1, wherein the optimization processing is performed on the knowledge point recognition model and the new knowledge point determination model, specifically:
when the first accuracy rate does not reach a preset threshold value, improving the first threshold value in the knowledge point identification model so that the knowledge point identification model identifies a more needed knowledge point;
reducing a second threshold in the new knowledge point determination model such that the new knowledge point determination model determines more dissimilar knowledge points.
3. The method for building a knowledge base realizing autonomous learning according to claim 1, wherein the required knowledge points are questions of interest to a user and answers to the questions of interest to the user.
4. The method for building a knowledge base realizing autonomous learning according to claim 1, wherein the similarity algorithm is a binary algorithm.
5. The knowledge base construction method for realizing autonomous learning according to any one of claims 1 to 4, wherein the preset threshold is 0.95.
6. The method for building a knowledge base realizing autonomous learning according to claim 5, wherein the first threshold is 0.8.
7. The method for building a knowledge base realizing autonomous learning according to claim 6, wherein the second threshold is 0.2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910562032.3A CN110334080B (en) | 2019-06-26 | 2019-06-26 | Knowledge base construction method for realizing autonomous learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910562032.3A CN110334080B (en) | 2019-06-26 | 2019-06-26 | Knowledge base construction method for realizing autonomous learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110334080A CN110334080A (en) | 2019-10-15 |
| CN110334080B true CN110334080B (en) | 2020-05-12 |
Family
ID=68143005
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910562032.3A Active CN110334080B (en) | 2019-06-26 | 2019-06-26 | Knowledge base construction method for realizing autonomous learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334080B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110781672B (en) * | 2019-10-30 | 2024-01-30 | 北京爱学习博乐教育科技有限公司 | Question bank production method and system based on machine intelligence |
| CN112860889A (en) * | 2021-01-29 | 2021-05-28 | 太原理工大学 | BERT-based multi-label classification method |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105824933A (en) * | 2016-03-18 | 2016-08-03 | 苏州大学 | Automatic question-answering system based on theme-rheme positions and realization method of automatic question answering system |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7096179B2 (en) * | 2001-08-15 | 2006-08-22 | Siemens Corporate Research, Inc. | Text-based automatic content classification and grouping |
| CN103810170B (en) * | 2012-11-06 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Intercommunion platform file classification method and device |
| CN103838886A (en) * | 2014-03-31 | 2014-06-04 | 辽宁四维科技发展有限公司 | Text content classification method based on representative word knowledge base |
| CN105095969B (en) * | 2015-09-25 | 2018-08-17 | 沈阳农业大学 | The device of autonomous learning model towards shared knowledge |
| US10304444B2 (en) * | 2016-03-23 | 2019-05-28 | Amazon Technologies, Inc. | Fine-grained natural language understanding |
-
2019
- 2019-06-26 CN CN201910562032.3A patent/CN110334080B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105824933A (en) * | 2016-03-18 | 2016-08-03 | 苏州大学 | Automatic question-answering system based on theme-rheme positions and realization method of automatic question answering system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110334080A (en) | 2019-10-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108960409B (en) | Method and device for generating annotation data and computer-readable storage medium | |
| CN112069310B (en) | Text classification method and system based on active learning strategy | |
| CN109960808B (en) | Text recognition method, device and equipment and computer readable storage medium | |
| CN101710490A (en) | Method and device for compensating noise for voice assessment | |
| CN110610193A (en) | Method and device for processing labeled data | |
| CN111159404B (en) | Text classification method and device | |
| CN109325106A (en) | A kind of U.S. chat robots intension recognizing method of doctor and device | |
| CN110334080B (en) | Knowledge base construction method for realizing autonomous learning | |
| CN112926045B (en) | Group control equipment identification method based on logistic regression model | |
| CN111368096A (en) | Knowledge graph-based information analysis method, device, equipment and storage medium | |
| CN110544481A (en) | S-T classification method and device based on voiceprint recognition and equipment terminal | |
| CN112906790B (en) | Solitary old man identification method and system based on electricity consumption data | |
| CN111046930A (en) | Power supply service satisfaction influence factor identification method based on decision tree algorithm | |
| CN110458600A (en) | Portrait model training method, device, computer equipment and storage medium | |
| CN111597821B (en) | Method and device for determining response probability | |
| CN107766560A (en) | The evaluation method and system of customer service flow | |
| CN114781688A (en) | Method, device, equipment and storage medium for identifying abnormal data of business expansion project | |
| CN112052686B (en) | Voice learning resource pushing method for user interactive education | |
| CN113095511A (en) | Method and device for judging in-place operation of automatic master station | |
| CN107909090A (en) | Learn semi-supervised music-book on pianoforte difficulty recognition methods based on estimating | |
| CN112738724B (en) | Method, device, equipment and medium for accurately identifying regional target crowd | |
| CN109308565B (en) | Crowd performance grade identification method and device, storage medium and computer equipment | |
| CN112819527A (en) | User grouping processing method and device | |
| CN111353553A (en) | Method and device for cleaning error labeling data, computer equipment and storage medium | |
| CN110659356B (en) | Information matching method, device and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |