CN109918127A - A kind of defect error correction method based on code revision pattern differentials - Google Patents
A kind of defect error correction method based on code revision pattern differentials Download PDFInfo
- Publication number
- CN109918127A CN109918127A CN201910170495.5A CN201910170495A CN109918127A CN 109918127 A CN109918127 A CN 109918127A CN 201910170495 A CN201910170495 A CN 201910170495A CN 109918127 A CN109918127 A CN 109918127A
- Authority
- CN
- China
- Prior art keywords
- defect
- code
- report
- cluster
- value
- 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.)
- Granted
Links
Landscapes
- Stored Programmes (AREA)
Abstract
The present invention provides a kind of defect error correction methods based on code revision pattern differentials in software maintenance technology field, the following steps are included: first passing through crawler to obtain state in defect searching platform Bugzilla is settled defect report, the defect report connected with not replicated relation is formed into a cluster according to the sequence of defect report ID；It recycles TF-IDF algorithm to carry out keyword extraction to the title of defect in each cluster and description, and keyword extraction is carried out to remaining defect report, and be added in the cluster of corresponding keyword；Secondly, the source code that comparison every two report is submitted, obtains Code Clones pair, finds out the diff information of source code cloned segment in each defect report, make an amendment pattern differentials figure to every part of diff information respectively；Finally, calculating difference ratio, judge the modification of defect with the presence or absence of problem；The present invention help developer more rapidly easily find out modified defect there are the problem of.
Description
Technical field
The invention belongs to software maintenance technology field, in particular to a kind of defect error correction based on code revision pattern differentials
Method.
Background technique
Now, with the continuous development of science and technology, the rise in intelligent epoch, demand of the people to software are more and more.When opening
When hair personnel encounter some indeterminable defect problems during software development and maintenance, they generally require to pass through retrieval
The software defect platform of some open sources obtains more defect relevant informations, if GitHub(mono- is towards open source and privately owned soft
The hosted platform of part project), Stack Overflow(mono- IT Questions & Answers website relevant to program), Bugzilla(mono-
The defect tracking system of a open source) etc. platforms.But developer has found that state is resolved, verified or closed sometimes
Defect report reappeared problem, so state is changed to reopened, such as Fig. 1.Currently, it is directed to software defect platform,
Having some research contents keeps the defect repair of developer more convenient, such as version control system and defect tracking system will
Existing defect report is integrated in software history library, and is supplied to software developer and is searched and refer to, and is come with this
It helps them and carries out defect repair according to obtained defect relevant information, but be to have repaired defect for detection there is no research
With the presence or absence of the research of problem.
In defect platform, only when defect report presenter has found defect repair, there are just can be by the shape of defect when problem
State is changed to reopened, and carries out subsequent research and reparation, and the existing research for defect platform is only exploit person
The defect that newly proposes of member provides some information reparations for reference, but have ignored repaired in defect with the presence or absence of problem this
Point.
Summary of the invention
For the defects in the prior art, it is an object of the invention to overcome shortcoming in the prior art, one is provided
Defect error correction method of the kind based on code revision pattern differentials, solution can not detect to have repaired whether defect deposits in the prior art
In the technical problem of problem, the method can detect settled defect there may be the problem of, save discovery defect repair problem
Time, enhance the defect repair experience of developer.
The object of the present invention is achieved like this: the defect error correction method based on code revision pattern differentials, including following
Step:
(1) defect report that state in defect searching platform Bugzilla is resolved, including its title are obtained by crawler
(title), description(description), relationship and code section between defect, and data cleansing is carried out, according still further to defect report
ID(serial number) sequence by the defect report connected with duplicate relationship formed a cluster；
(2) using TF-IDF(term frequency-inverse document frequency) algorithm to defect in each cluster
Title and description carry out keyword extraction, the descriptor as the cluster；
(3) after having confirmed the descriptor of cluster, TF-IDF algorithm is recycled to carry out keyword extraction to remaining defect report, and add
In the cluster for entering corresponding keyword；
(4) it for the report in each cluster, is mentioned with a kind of Code Clones detection algorithm comparison every two report for being known as CDLH
The source code of friendship, and Code Clones pair are obtained, the diff information of source code cloned segment in each defect report is found out, and
Pattern differentials figure is made an amendment to every part of diff information respectively；
(5) calculate difference ratio, if difference ratio be not less than 65~75%, then it is assumed that the two has made modification of the same race mostly, then remaining
There may be problems for the modification of defect, and the defect for having reparation problem is sent to and reports that presenter is examined again, repaired.
In the present invention by by state in defect platform be resolved(solved) defect report according to duplicate
Relationship formed cluster, using TF-IDF algorithm to each cluster extract descriptor, then by remaining defect according to itself keyword carry out
Match and sort out, defect report can be carried out the classification based on keyword by this method, be more advantageous to the later period and searched in each cluster
Code Clones pair out；The Code Clones detection algorithm based on grammer is recycled to detect identical code, renaming/parameter
The code of change, almost the same code and semantic similar code, can more fully detect code in this way
Clone couple；Code Clones are being detected to rear, comparison code is modified to diff() information makes an amendment pattern differentials figure, there are differences
Different defect report is then likely to become the problematic defect of modification mode, carries out error correction to existing defect report, can help
Developer more rapidly easily find out modified defect there are the problem of, enhance the defect repair experience of developer, mention
The validity and high efficiency of high defect retrieval；It can be applied in the work to settled defect problem error correction.
In order to further realize the extraction of keyword, in the step (2), the specific steps of keyword are extracted are as follows: TF-
IDF algorithm is made of TF and IDF two parts, and TF is that the appearance to word each in text carries out frequency statistics, and IDF is for assigning
The specific weight of each word is given, after calculating separately two values, the two is multiplied to obtain TF-IDF value, and to word each in document
Value be ranked up, it is highest several as descriptor to choose its intermediate value.
In order to further realize whether be Code Clones pair judgement, in the step (4), judge in two defect reports
Two sections of codes whether be that the steps of Code Clones pair is specifically,
(401) source code segment is inputted, and is translated into AST；
(402) using the real value representation for obtaining each code snippet based on the LSTM of AST；
(403) by hash function that real value obtained in step (402) is encoded translated for binary system Hash codes；
(404) a pair of of hash code is given, with function
To determine code aiWith code ajIt whether is Code Clones pair, then return function g(a if the conditions are meti,aj) value be 1, that is, sentence
Determine code aiWith code ajFor Code Clones pair；Otherwise return function g(ai,aj) value be -1, code aiWith code ajIt is not generation
Code clone couple；
Wherein, aiAnd ajFor the code segment that two Hash indicate, k is the dimension in Hamming space, and m is that the dimension in Hamming space is total
Number, aI, kIndicate code segment aiCryptographic Hash in dimension k, aj,kFor code segment ajCryptographic Hash in dimension k,Refer to the offer of tender
Number, thr is threshold value and value is 2.
In order to calculate the difference ratio of two defect reports, in the step (5), the calculation formula of difference ratio is specific
Are as follows:(1)；
Wherein,The number of certain modification mode in respectively two defect reports,Respectively two defects
The weight of mode is modified in report,The gross sample number of middle modification mode in respectively two defects,For the difference of the two
Ratio value.
Detailed description of the invention
Fig. 1 is the condition conversion figure of Bugzilla platform defect report in the present invention.
Fig. 2 is flow diagram of the invention.
Fig. 3 is the bug report example screenshot enumerated in the present invention.
Fig. 4 is that Bugzilla platform is referred to the defects of cluster figure in example of the invention.
Fig. 5 is the diff information screenshot of a bug in example of the invention.
Fig. 6 is the diff information screenshot of another bug in example of the invention.
Fig. 7 is that diff information modifies pattern differentials figure in example of the invention.
Specific embodiment
Defect error correction method based on code revision pattern differentials as shown in Figure 2, comprising the following steps:
(1) defect report that state in defect searching platform Bugzilla is resolved, including its title are obtained by crawler
(title), description(description), relationship and code section between defect, and data cleansing is carried out, according still further to defect report
ID(serial number) sequence will be repeated with duplicate() defect report that connects of relationship forms a cluster；
(2) using TF-IDF(term frequency-inverse document frequency) algorithm to defect in each cluster
Title and description carry out keyword extraction, the descriptor as the cluster；
(3) after having confirmed the descriptor of cluster, TF-IDF algorithm is recycled to carry out keyword extraction to remaining defect report, and add
In the cluster for entering corresponding keyword；
(4) it for the report in each cluster, is mentioned with a kind of Code Clones detection algorithm comparison every two report for being known as CDLH
The source code of friendship, and Code Clones pair are obtained, the diff information of source code cloned segment in each defect report is found out, and
Pattern differentials figure is made an amendment to every part of diff information respectively；
(5) according to difference ratio calculation formula, if difference ratio is not less than 65~75%, then it is assumed that the two has done of the same race repair mostly
Change, then there may be problems for the modification of remaining defect, and the defect for having reparation problem is sent to report presenter and is carried out again
It examines, repair；
The calculation formula of difference ratio is specific are as follows:(1)；
Wherein,The number of certain modification mode in respectively two defect reports,Respectively two defects
The weight of mode is modified in report,The gross sample number of middle modification mode in respectively two defects,For the difference of the two
Ratio value.
In step (2), extract the specific steps of keyword are as follows: TF-IDF algorithm is by TF(word frequency) and IDF(against text word
Frequently two parts form), and TF is that the appearance to word each in text carries out frequency statistics, and IDF is specific for assigning each word
Weight, after calculating separately two values, the two is multiplied to obtain TF-IDF value, and arrange the value of word each in document
It is highest several as descriptor to choose wherein TF-IDF value for sequence.
In step (4), judge whether two sections of codes in two defect reports are that the steps of Code Clones pair is specifically,
(401) source code segment is inputted, and is translated into AST(abstract syntax tree)；
(402) the LSTM(shot and long term memory network based on AST is used, is a kind of time recurrent neural network, is capable of Chief Learning Officer, CLO's
Dependence) obtain the real value representation of each code snippet；
(403) by hash function by real value obtained in step (402) it is encoded translated be binary system Hash codes, make clone pair
Hamming distance (hamming distance) of the code snippet in Hamming space (hamming space) can connect each other
Closely, facilitated with this and judge whether two sections of codes are cloned codes pair；
(404) a pair of of hash code is given, with function
To determine code aiWith code ajIt whether is Code Clones pair, then return function g(a if the conditions are meti,aj) value be 1, that is, sentence
Determine code aiWith code ajFor Code Clones pair；Otherwise return function g(ai,aj) value be -1, code aiWith code ajIt is not generation
Code clone couple；
Wherein, aiAnd ajThe code segment that respectively two Hash indicate, k are the dimension in Hamming space, and m is the dimension in Hamming space
Sum, ai,kIndicate code segment aiCryptographic Hash in dimension k, aj,kFor code segment ajCryptographic Hash in dimension k,Refer to
Scalar functions, thr is threshold value and preferred value is 2.
In the present invention by by state in defect platform be resolved(solved) defect report according to duplicate
Relationship formed cluster, using TF-IDF algorithm to each cluster extract descriptor, then by remaining defect according to itself keyword carry out
Match and sort out, defect report can be carried out the classification based on keyword by this method, be more advantageous to the later period and searched in each cluster
Code Clones pair out；The Code Clones detection algorithm based on grammer is recycled to detect identical code, renaming/parameter
The code of change, almost the same code and semantic similar code, can more fully detect code in this way
Clone couple；Code Clones are being detected to rear, comparison code is modified to diff() information makes an amendment pattern differentials figure, it has differences
Defect report be then likely to become the problematic defect of modification mode, to existing defect report carry out error correction, can help open
Hair personnel more rapidly easily find out modified defect there are the problem of, enhance the defect repair experience of developer, improve
The validity and high efficiency of defect retrieval；It can be applied in the work to settled defect problem error correction.
Specific embodiment is given below, and the invention will be further described.
A defect report being illustrated in figure 3 in Bugzilla platform, in the defect report of bug#1215247,
The defect of duplicate relationship has bug#1215252 and bug#1280804, and the two defect reports do not have the pass duplicate
The defect of system, so these three defects temporarily become a cluster；The cluster that defect in Fig. 3 is formed is carried out using TF-IDF algorithm
The extraction of keyword, available keyword ECMAScript Internationalization API, and as the cluster
Descriptor；It is ECMAScript that some defects, which are referred to descriptor by keyword, in platform
The cluster of Internationalization API, if Fig. 4 is to be referred to the defects of cluster example, Fig. 5 and Fig. 6 are respectively bug#
864843 diff information similar with source code in bug#866301 process comparison；Understand how to judge in order to facilitate the public
Cloned codes pair provide following example, 94 row (a in Fig. 5i) and Fig. 6 in 90 row (aj) are as follows: case " OS_TARGET "
In, the then a indicated with HashiFor [1,1,1], a indicated with HashjFor [1,1,1], then, then above-mentioned two generations
Code is cloned codes pair；Can be obtained according to Fig. 7, there are three types of the modes of modification altogether by bug#864843 --- deleting redundancy, (weight is taken as
0.5) parameter (weight is taken as 1) and addition if conditional statement (weight is taken as 1), are modified, and each pattern occurs once, bug#
866301 altogether there are two types of modify mode --- modification parameter (weight is taken as 1) and add if conditional statement (weight is taken as 1), (its
In, the weight of each modification mode is given by technical staff according to obtained diff information experience), and each pattern only occurs one
It is secondary, according to difference ratio calculation formula, the difference ratio value between above two bug can be obtained are as follows:, in the present embodiment, difference ratio is not less than 70%, then it is assumed that
The two has made modification of the same race mostly, so the difference ratio value of the two is 80%70%, then it can be determined that its modification is roughly the same,
Without examining again.
The present invention is not limited to the above embodiments, on the basis of technical solution disclosed by the invention, the skill of this field
For art personnel according to disclosed technology contents, one can be made to some of which technical characteristic by not needing creative labor
A little replacements and deformation, these replacements and deformation are within the scope of the invention.
Claims (4)
1. a kind of defect error correction method based on code revision pattern differentials, which comprises the following steps:
(1) defect report that state in defect searching platform Bugzilla is resolved, including its are obtained by crawler
Relationship and code section between title, description, defect, and data cleansing is carried out, according still further to the suitable of defect report ID
The defect report connected with duplicate relationship is formed a cluster by sequence；
(2) keyword extraction is carried out using title and description of the TF-IDF algorithm to defect in each cluster, as the cluster
Descriptor；
(3) after having confirmed the descriptor of cluster, TF-IDF algorithm is recycled to carry out keyword extraction to remaining defect report, and add
In the cluster for entering corresponding keyword；
(4) it for the report in each cluster, is mentioned with a kind of Code Clones detection algorithm comparison every two report for being known as CDLH
The source code of friendship, and Code Clones pair are obtained, the diff information of source code cloned segment in each defect report is found out, and
Pattern differentials figure is made an amendment to every part of diff information respectively；
(5) calculate difference ratio, if difference ratio be not less than 65~75%, then it is assumed that the two has made modification of the same race mostly, then remaining
There may be problems for the modification of defect, and the defect for having reparation problem is sent to and reports that presenter is examined again, repaired.
2. a kind of defect error correction method based on code revision pattern differentials according to claim 1, which is characterized in that institute
It states in step (2), extracts the specific steps of keyword are as follows: TF-IDF algorithm is made of TF and IDF two parts, and TF is in text
The appearance of each word carries out frequency statistics, and IDF be for assigning each word specific weight, after calculating separately two values,
The two is multiplied to obtain TF-IDF value, and the value of word each in document is ranked up, chooses the highest several conducts of its intermediate value
Descriptor.
3. a kind of defect error correction method based on code revision pattern differentials, which is characterized in that in the step (4), judge two
Whether two sections of codes in a defect report are that the steps of Code Clones pair is specifically,
(401) source code segment is inputted, and is translated into AST；
(402) using the real value representation for obtaining each code snippet based on the LSTM of AST；
(403) by hash function that real value obtained in step (402) is encoded translated for binary system Hash codes；
(404) a pair of of hash code is given, with function
To determine code aiWith code ajIt whether is Code Clones pair, then return function g(a if the conditions are meti,aj) value be 1, that is, sentence
Determine code aiWith code ajFor Code Clones pair；Otherwise return function g(ai,aj) value be -1, code aiWith code ajIt is not generation
Code clone couple；
Wherein, aiAnd ajFor the code segment that two Hash indicate, k is the dimension in Hamming space, and m is the dimension sum in Hamming space,
ai,kIndicate code segment aiCryptographic Hash in dimension k, aj,kFor code segment ajCryptographic Hash in dimension k,Refer to the offer of tender
Number, thr is threshold value and value is 2.
4. a kind of defect error correction method based on code revision pattern differentials according to claim 1, which is characterized in that
In the step (5), the calculation formula of difference ratio is specific are as follows:(1)；
Wherein,The number of certain modification mode in respectively two defect reports,Respectively two defects
The weight of mode is modified in report,The gross sample number of middle modification mode in respectively two defects,For the difference of the two
Ratio value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910170495.5A CN109918127B (en) | 2019-03-07 | 2019-03-07 | Defect error correction method based on code modification mode difference |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910170495.5A CN109918127B (en) | 2019-03-07 | 2019-03-07 | Defect error correction method based on code modification mode difference |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109918127A true CN109918127A (en) | 2019-06-21 |
| CN109918127B CN109918127B (en) | 2022-02-11 |
Family
ID=66963608
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910170495.5A Active CN109918127B (en) | 2019-03-07 | 2019-03-07 | Defect error correction method based on code modification mode difference |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109918127B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110851176A (en) * | 2019-10-22 | 2020-02-28 | 天津大学 | Clone code detection method capable of automatically constructing and utilizing pseudo clone corpus |
| CN112286807A (en) * | 2020-10-28 | 2021-01-29 | 北京航空航天大学 | Software defect positioning system based on source code file dependency relationship |
| CN112433756A (en) * | 2020-11-24 | 2021-03-02 | 北京京航计算通讯研究所 | Rapid code clone detection method and device based on weighted recursive self-encoder |
| CN112463424A (en) * | 2020-11-13 | 2021-03-09 | 扬州大学 | End-to-end program repair method based on graph |
| CN112612516A (en) * | 2020-12-30 | 2021-04-06 | 瑞庭网络技术（上海）有限公司 | Code processing method and device |
| CN114461196A (en) * | 2022-02-21 | 2022-05-10 | 广州图创计算机软件开发有限公司 | Intelligent auxiliary method and system for software development |
| CN117331849A (en) * | 2023-12-01 | 2024-01-02 | 杭银消费金融股份有限公司 | Testing method and system based on accurate analysis of codes |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130346424A1 (en) * | 2012-06-21 | 2013-12-26 | Microsoft Corporation | Computing tf-idf values for terms in documents in a large document corpus |
| CN106156633A (en) * | 2016-06-23 | 2016-11-23 | 扬州大学 | The risk analysis method of software-oriented amendment |
| CN107808278A (en) * | 2017-10-11 | 2018-03-16 | 河海大学 | A kind of Github open source projects based on sparse self-encoding encoder recommend method |
| US20180248895A1 (en) * | 2017-02-27 | 2018-08-30 | Amazon Technologies, Inc. | Intelligent security management |
-
2019
- 2019-03-07 CN CN201910170495.5A patent/CN109918127B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130346424A1 (en) * | 2012-06-21 | 2013-12-26 | Microsoft Corporation | Computing tf-idf values for terms in documents in a large document corpus |
| CN106156633A (en) * | 2016-06-23 | 2016-11-23 | 扬州大学 | The risk analysis method of software-oriented amendment |
| US20180248895A1 (en) * | 2017-02-27 | 2018-08-30 | Amazon Technologies, Inc. | Intelligent security management |
| CN107808278A (en) * | 2017-10-11 | 2018-03-16 | 河海大学 | A kind of Github open source projects based on sparse self-encoding encoder recommend method |
Non-Patent Citations (1)
| Title |
|---|
| 孙小兵 等: "软件修改影响分析研究与进展", 《电子学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110851176A (en) * | 2019-10-22 | 2020-02-28 | 天津大学 | Clone code detection method capable of automatically constructing and utilizing pseudo clone corpus |
| CN110851176B (en) * | 2019-10-22 | 2023-07-04 | 天津大学 | Clone code detection method capable of automatically constructing and utilizing pseudo-clone corpus |
| CN112286807A (en) * | 2020-10-28 | 2021-01-29 | 北京航空航天大学 | Software defect positioning system based on source code file dependency relationship |
| CN112463424A (en) * | 2020-11-13 | 2021-03-09 | 扬州大学 | End-to-end program repair method based on graph |
| CN112463424B (en) * | 2020-11-13 | 2023-06-02 | 扬州大学 | Graph-based end-to-end program repairing method |
| CN112433756A (en) * | 2020-11-24 | 2021-03-02 | 北京京航计算通讯研究所 | Rapid code clone detection method and device based on weighted recursive self-encoder |
| CN112612516A (en) * | 2020-12-30 | 2021-04-06 | 瑞庭网络技术（上海）有限公司 | Code processing method and device |
| CN112612516B (en) * | 2020-12-30 | 2024-05-03 | 瑞庭网络技术（上海）有限公司 | Code processing method and device |
| CN114461196A (en) * | 2022-02-21 | 2022-05-10 | 广州图创计算机软件开发有限公司 | Intelligent auxiliary method and system for software development |
| CN114461196B (en) * | 2022-02-21 | 2022-09-27 | 广州图创计算机软件开发有限公司 | Intelligent auxiliary method and system for software development |
| CN117331849A (en) * | 2023-12-01 | 2024-01-02 | 杭银消费金融股份有限公司 | Testing method and system based on accurate analysis of codes |
| CN117331849B (en) * | 2023-12-01 | 2024-02-27 | 杭银消费金融股份有限公司 | Testing method and system based on accurate analysis of codes |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109918127B (en) | 2022-02-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109918127A (en) | A kind of defect error correction method based on code revision pattern differentials | |
| Wang et al. | Detecting code clones with graph neural network and flow-augmented abstract syntax tree | |
| Wang et al. | Exploring graph neural networks for semantic enrichment: Room type classification | |
| Xia et al. | Collective personalized change classification with multiobjective search | |
| CN109799990A (en) | Source code annotates automatic generation method and system | |
| Wang et al. | Detecting code clones with graph neural networkand flow-augmented abstract syntax tree | |
| CN113961241A (en) | Code clone detection method based on GAT (generic antigen-based) graph neural network model | |
| Fei et al. | Making decision like human: Joint aspect category sentiment analysis and rating prediction with fine-to-coarse reasoning | |
| Ferrag et al. | Securefalcon: The next cyber reasoning system for cyber security | |
| CN114332519A (en) | Image description generation method based on external triple and abstract relation | |
| Wu et al. | Turn tree into graph: Automatic code review via simplified ast driven graph convolutional network | |
| CN115562721A (en) | Clone code detection method and system for mining features from assembly language | |
| CN117787253B (en) | Triple extraction method based on double gating mechanism and depending on directed attention network | |
| CN117540035B (en) | RPA knowledge graph construction method based on entity type information fusion | |
| Ni et al. | Defect identification, categorization, and repair: Better together | |
| Yang et al. | Smart contracts vulnerability auditing with multi-semantics | |
| Yang et al. | Improving domain-specific neural code generation with few-shot meta-learning | |
| CN116663010A (en) | Intelligent contract vulnerability detection system construction method based on GPT model | |
| CN105119961A (en) | Semantic Web service automatic combination method based on body | |
| Torres et al. | Comparison of Clang abstract syntax trees using string kernels | |
| Wu et al. | Contrastive learning for multi-modal automatic code review | |
| Geng | Media bias detecting based on word embedding | |
| Xu et al. | Data Quality Matters: A Case Study of Obsolete Comment Detection | |
| Dong et al. | Multi-passage extraction-based machine reading comprehension based on verification sorting | |
| Zubkov et al. | Evaluation of Contrastive Learning with Various Code Representations for Code Clone Detection |
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 |