CN108717423A - A kind of code segment recommendation method excavated based on deep semantic - Google Patents
A kind of code segment recommendation method excavated based on deep semantic Download PDFInfo
- Publication number
- CN108717423A CN108717423A CN201810371788.5A CN201810371788A CN108717423A CN 108717423 A CN108717423 A CN 108717423A CN 201810371788 A CN201810371788 A CN 201810371788A CN 108717423 A CN108717423 A CN 108717423A
- Authority
- CN
- China
- Prior art keywords
- code segment
- natural language
- vector
- collection
- training
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F8/00—Arrangements for software engineering
- G06F8/30—Creation or generation of source code
- G06F8/31—Programming languages or programming paradigms
- G06F8/315—Object-oriented languages
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention discloses a kind of code segments excavated based on deep semantic to recommend method, effect of the depth learning technology in natural language processing and its advantage in the excavation of natural language semanteme is utilized, and combined the characteristics of inquiry code segment is recommended.According to the Natural Language Search of input and code segment itself and its annotation of institute's band, depth excavates natural language semanteme and code segment concrete function, generate sentence vector sum paragraph vector, so that therefore the consistent code segment and natural language querying of semantic attribute is mapped to similar vector space, recommend most matched, similarity sorted N number of code segment from high to low for given inquiry.This method not only increases the accuracy of recommendation, moreover it is possible to improve the recall ratio of recommendation, and have preferable fault-tolerant ability to the natural language querying of input inquiry.
Description
Technical field
The technical field that the invention belongs to have the code of inquiry to recommend, was referred specifically to for a kind of generation excavated based on deep semantic
Code section recommends method.
Background technology
During actual written in code, developer is frequently encountered unfamiliar programmed tasks or needs to realize
Certain specific functions, in this case, if developer can find existing similar code segment to learn its use
Method either is directly replicated to paste and then modify and be improved to carry out code reuse, can so be saved to developer
Save plenty of time, energy and meaningless repeated work；However, how to recommend the generation of high quality based on developer's actual demand
Code Duan Ze is the major issue of software repeated usage.
During actual development, developer would generally select to inquire the code segment of needs using search engine.But
Since software code has globality, the keyword in code segment can accurately not describe the function of one section of code, so looking into
Asking result generally can not be fully up to expectations.In addition, existing recommendation method is usually concerned only with code segment itself and ignores its description letter
Breath, and the function with the natural language description code segment of the description information of code segment most simple, intuitive.In recent years, due to depth
The extensive use of study is spent, Language Processing field also achieves breakthrough so that carry out to natural language and programming language
Deep semantic and information excavating can also obtain good effect.So it is one that language processing techniques are combined with code recommendation
The novel effective recommendation method of kind.
Invention content
Above-mentioned the deficiencies in the prior art are directed to, the purpose of the present invention is to provide a kind of generations excavated based on deep semantic
Code section recommends method, supports the code segment towards natural language querying to recommend using depth learning technology；The present invention being capable of basis
The annotation and code segment ontology of Natural Language Search and code segment itself the institute band of user, depth excavate natural language semanteme and generation
Code section concrete function so that the consistent band annotated code section of semantic attribute and natural language querying be therefore mapped to it is similar to
Quantity space recommends most matched code segment for given inquiry.
In order to achieve the above objectives, the technical solution adopted by the present invention is as follows：
A kind of code segment excavated based on deep semantic of the present invention recommends method, includes the following steps：
Step 1)：Construct the code segment collection S with method description information on a large scale；
Step 2)：Building method description information collection D1With method main body collection D2, construction collection of comments D1', and utilize and construct
Data set train Encoder-Decoder natural language sentence vector generator model Ms1, training Encoder-Decoder programmings
Language passages vector generator model M2；
Step 3)：The method name Name of each code segment in code segment collection S is extracted, and vector after being mapped with the code segment
Indicate that α ' constitutes key-value pair form<Name,α'>, as index file used when recommending；
Step 4)：The natural language querying given to one obtains its corresponding natural language sentence vector, then in band side
Most matched N number of sorted code segment is recommended to each inquiry in the code segment collection S of method description information.
Preferably, the step 1) specifically includes：Detailed programs are obtained from open source software platform, to source generation in detailed programs
Code file is that unit is cut by method, obtains the code segment collection S with method description information, the title form of each code segment
For packet name & class name & method names；.
Preferably, the step 1) further includes specifically：The detailed programs be Java projects, Android projects and its
Its project.
Preferably, the step 2) specifically includes：
21) with method description information collection D1It is training set to Encoder-Decoder natural language sentence vector generator moulds
Type M1It is trained, converges to designated state, complete the training of natural language sentence vector generator；By method description information collection D1
In each code segment annotation first sentence extraction out as input, then generate natural language sentences vector α1And by its
As the corresponding part indicated with annotated code vector paragraph；
22) with method main body collection D2It is training set to Encoder-Decoder programming language paragraph vector generator model Ms2
Be trained, in training to completing model training when specifying convergence state, and at the same time generate the section of each code segment main body to
Measure α2；
23) by natural language sentences vector α1With the vector paragraph α of code segment main body2Weighting summation, obtain vectorial α and by its
As the vector that can finally characterize entirely with annotated code section, then by the set and collection of comments D of institute directed quantity α1' from
Right language sentence vector is denoted as training set to train neural network mapping model M3, and pass through nerve net after the completion of training
Network mapping model M3Mapping vectorial α is mapped to obtain band annotated code section mapping after vector table show α '.
Preferably, the step 4) specifically includes：
41) for trained Encoder-Decoder natural languages sentence vector generator model M1, give a nature
The query statement sentence vector β of a specified dimension is calculated in language in-put；
42) similarity between the two vectors is indicated with two vectorial angle cos θ, vector table shows after calculating mapping
Similarity value between α ' and query statement sentence vector β recommends and its most similar N number of code given natural language querying
Section, and sort from high to low according to similarity.
Beneficial effects of the present invention：
The present invention using effect of the depth learning technology in natural language processing and its in language semantic excavation
Advantage is used to solve the problems, such as how according to given natural language querying to recommend the band annotated code section of high quality；With following
Advantage：
(1) carrying out natural language processing using deep learning really can deeply excavate natural language semanteme, rather than only
Matched merely with text key word so that sentence vector corresponding to semantic identical sentence semantic space distance closer to,
Inquiry meaning to be expressed thus can be really excavated, and then keeps matching when recommendation more accurate, improves recommendation
Accuracy.
(2) code segment structure can be excavated by carrying out the processing method of paragraph vectorization to code segment main body using deep learning
Information and semantic information in programming language level are extracted rather than just simple Feature Words, are thus fully excavated
The information that code segment main body itself is included, and then the effect of code segment recommendation can be improved.
(3) match to obtain the most similar N number of code segment of annotating semantic as recommendation results, and according to language using deep semantic
Adopted similarity sorts from high to low, also can be accordingly even when the query express of input is not clear enough or have slight deviations
Suitable recommendation results are found in relatively low position, not only increase recall ratio, also certain fault-tolerant ability.
Description of the drawings
Fig. 1 is used frame model schematic diagram to generate sentence vector sum paragraph vector in the present invention.
Fig. 2 is the Encoder-Decoder model schematics used in the present invention.
Fig. 3 is the basic unit schematic diagram in the Encoder-Decoder models used in the present invention.
Fig. 4 is the principle of the present invention figure.
Specific implementation mode
For the ease of the understanding of those skilled in the art, the present invention is made further with reference to embodiment and attached drawing
Bright, the content that embodiment refers to not is limitation of the invention.
1- Fig. 4 is described in detail the technical solution of invention by taking the recommendation of Java code section as an example below in conjunction with the accompanying drawings：
Step 1：Construct the code segment collection S with method description information on a large scale；Wherein,
11) Java projects are obtained on the software platform (such as GitHub) increased income, to Java files in project according to side
Method is that unit is cut, and obtains the method with method description information, the text of the entitled filename of write packet name class name method
In part.
12) the code segment collection S with method description information tentatively obtained is screened, by inferior (such as without method
Description information) or the deletion of useless (such as test method) code segment, the S set of the high quality after being simplified.
Step 2：Building method description information collection D1With method main body collection D used in training programming language paragraph vector2；
The methodical description information of institute is extracted, obtains training method description information collection used in natural language sentence vector
D1, first of abstracting method description information obtain collection of comments D1', the methodical code segment ontology of institute is extracted, obtains training generation
Method main body collection D used in the vector paragraph of code section2。
Step 3：Natural language sentences vector generator and programming language paragraph vector generator are constructed and trained, and is obtained
Vector with annotated code section indicates α, then maps vectorial α to obtain vector by trained neural network mapping model
α'；Wherein,
31) the sentence vector generator of natural language is with method description information collection D1For training set, then to Encoder-
Decoder natural language sentence vector generator model Ms1Be trained until converge to designated state, complete natural language sentence to
Measure the training of generator；By method description information collection D1In first sentence extraction of each code segment annotation be out used as M1's
Input generates natural language sentences vector α1And as the corresponding part indicated with annotated code vector paragraph；
32) the paragraph vector generator of programming language is with method main body collection D2Encoder-Decoder is programmed for input
Language passages vector generator model M2It is trained, is completed in the when training of training to specified convergence state, and training
The vector paragraph α of each code segment main body is generated in journey2；
33) by natural language sentences vector α1With the vector paragraph α of code segment main body2Weighting summation, obtains vectorial α, and by its
As the vector that can finally characterize entirely with annotated code section, then by each α with corresponding to annotated code section and the band
The corresponding vector annotated of annotated code section is denoted as training set, training neural network mapping model M3, and completed in training
Vector table shows α ' after the mapping of mapping model is mapped α afterwards, which can characterize with annotated code section
Semantic vector α is indicated in the vector of natural language semantic space.
Step 4：The method name Name of each code segment is extracted from the code segment collection S with method description information, also
It is the form of packet name class name method name, and shows that α ' constitutes the form of key-value pair with vector table after the mapping of the code segment<Name,
α'>, as index file used when recommending.
Step 5：The natural language querying given to one obtains its corresponding natural language sentence vector, then in band side
Most matched N number of sorted code segment is recommended to each inquiry in the code segment collection S of method description information；Wherein,
51) natural language querying given to one uses trained Encoder-Decoder natural languages sentence vector
Maker model M1 is calculated the query statement sentence vector β corresponding to natural language querying；
52) similarity between two vectors is indicated with two vectorial angle cos θ, to given natural language querying
It is calculated in query statement sentence vector β and index file that vector table shows the similarity of α ', then root after the mapping of each code segment
According to index file recommendation and its most similar N number of code segment, and sort from high to low according to similarity.
Embodiment：
The Java projects obtained on the software platform GitHub that increases income are cut first, obtain band annotation one by one
Code segment, and be written into file.By taking project assertj-core-master as an example, obtained after cutting
“main.java.org.assertj.core.internal.Strings.java&Strings&
doCommonCheckForCharSequence.java”、
" main.java.org.assertj.core.util.diff.Delta.java&Delta&De lta.java " ..., single generation
Code section form is as follows：
35 are obtained in project assertj-core-master has standalone feature, high quality code segment.
After data set processing is completed, the annotation collection D of code segment is further obtained1', D1' in comment statement be " Remove
the first instance of a value if found in the list and replaces it with the
last item","get file content","......".Method description information collection D1, D1In descriptive statement be：
“Remove the first instance of a value if found in the list and
replaces it with the last item in the list.This saves a copy down of all
Items at the expense of not preserving list order. " " ... " etc and code segment side
Method main body collection D2.
After all model trainings are completed, its corresponding sentence vector and band can be obtained to arbitrary natural language querying
The natural language sentences vector α of annotated code section1With the vector paragraph α of each code segment main body2, and addition calculation obtains final α
It is indicated as the vector with annotated code section, in code segment example as above：
α=[0.0501139,0.0799258,0.0690878 ...]
It is after mapping model conversion：
α '=[0.1001695,0.060278,0.0700396 ...]
Input inquiry sentence y=(y1,y2,...,yt), specially " remove the first instance of a
List " then obtains sentence vector (" the remove the first of the corresponding specified dimensions of y after by model treatment
Instance of a list "=[0.0703125,0.0869141,0.0878906 ...]).
The index file of the code segment of N number of band annotation is in data set S<N4S1,αi’1>,<N4S2,αi’2>...,<
N4SN,αi’N>, specifically such as<"Remove the first instance of a value if found in the list
and replaces it with the last item in the list”,M1>......<“Input and output
of a file”,Mk>..., there are cos θ (y, αi') value be respectively 0.0054,0.062, number as 0.785......
Value, and to be minimum N number of in S, and cos θ (y, α '1)<Cos θ (y, α '2)<......<Cos θ (y, α 'N), then the knot recommended
Fruit is：
1:<main.java.org.assertj.core&Delta&fastUnorderedRemoveInt,
[0.1001695,......]>
2:<……,……>
...
N:<……,……>
Here N number of sorted code segment is index, that is, the corresponding link of code segment, performance in actual recommendation
For the form of the packet name & class name & method names in S, when user wants to check some specific code segment, it is only necessary to click
Check real source code code segment, design is based on the considerations of users'comfort and aesthetics in this way.
There are many concrete application approach of the present invention, the above is only a preferred embodiment of the present invention, it is noted that for
For those skilled in the art, without departing from the principle of the present invention, it can also make several improvements, this
A little improve also should be regarded as protection scope of the present invention.
Claims (5)
1. a kind of code segment excavated based on deep semantic recommends method, which is characterized in that include the following steps：
Step 1)：Construct the code segment collection S with method description information on a large scale；
Step 2)：Building method description information collection D1With method main body collection D2, construction collection of comments D1', and utilize the number constructed
According to collection training Encoder-Decoder natural language sentence vector generator model Ms1, training Encoder-Decoder programming languages
Paragraph vector generator model M2；
Step 3)：The method name Name of each code segment in code segment collection S is extracted, and is shown with vector table after code segment mapping
α ' constitutes key-value pair form<Name,α'>, as index file used when recommending；
Step 4)：The natural language querying given to one obtains its corresponding natural language sentence vector, is then retouched in band method
It states in the code segment collection S of information and most matched N number of sorted code segment is recommended to each inquiry.
2. the code segment according to claim 1 excavated based on deep semantic recommends method, which is characterized in that the step
1) it specifically includes：Detailed programs are obtained from open source software platform, source code file in detailed programs is carried out by method for unit
Cutting obtains the code segment collection S with method description information, and the title form of each code segment is packet name & class name & method names.
3. the code segment according to claim 2 excavated based on deep semantic recommends method, which is characterized in that the step
1) further include specifically：The detailed programs are Java projects or Android projects.
4. the code segment according to claim 1 excavated based on deep semantic recommends method, which is characterized in that the step
2) it specifically includes：
21) with method description information collection D1It is training set to Encoder-Decoder natural language sentence vector generator model Ms1Into
Row training, converges to designated state, completes the training of natural language sentence vector generator；By method description information collection D1In it is each
Then first sentence extraction of code segment annotation generates natural language sentences vector α out as input1And as right
The part that should be indicated with annotated code vector paragraph；
22) with method main body collection D2It is training set to Encoder-Decoder programming language paragraph vector generator model Ms2It carries out
Training completes model training, and at the same time generating the vector paragraph α of each code segment main body when training is to specified convergence state2；
23) by natural language sentences vector α1With the vector paragraph α of code segment main body2Weighting summation, obtain vectorial α and as
The vector entirely with annotated code section can be finally characterized, then by the set and collection of comments D of institute directed quantity α1' natural language
Speech sentence vector is denoted as training set to train neural network mapping model M3, and reflected by neural network after the completion of training
Penetrate model M3Mapping vectorial α is mapped to obtain band annotated code section mapping after vector table show α '.
5. the code segment according to claim 1 excavated based on deep semantic recommends method, which is characterized in that the step
4) it specifically includes：
41) for trained Encoder-Decoder natural languages sentence vector generator model M1, give a natural language
The query statement sentence vector β of a specified dimension is calculated in input；
42) indicate the similarity between the two vectors with two vectorial angle cos θ, calculate vector table after mapping show α ' and
Similarity value between query statement sentence vector β, to given natural language querying recommend with its most similar N number of code segment,
And it sorts from high to low according to similarity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810371788.5A CN108717423B (en) | 2018-04-24 | 2018-04-24 | Code segment recommendation method based on deep semantic mining |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810371788.5A CN108717423B (en) | 2018-04-24 | 2018-04-24 | Code segment recommendation method based on deep semantic mining |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108717423A true CN108717423A (en) | 2018-10-30 |
| CN108717423B CN108717423B (en) | 2020-07-07 |
Family
ID=63899075
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810371788.5A Active CN108717423B (en) | 2018-04-24 | 2018-04-24 | Code segment recommendation method based on deep semantic mining |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108717423B (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670022A (en) * | 2018-12-13 | 2019-04-23 | 南京航空航天大学 | A kind of java application interface use pattern recommended method based on semantic similarity |
| CN110716749A (en) * | 2019-09-03 | 2020-01-21 | 东南大学 | Code searching method based on function similarity matching |
| CN110806861A (en) * | 2019-10-10 | 2020-02-18 | 南京航空航天大学 | API recommendation method and terminal combining user feedback information |
| CN111061935A (en) * | 2019-12-16 | 2020-04-24 | 北京理工大学 | Science and technology writing recommendation method based on self-attention mechanism |
| CN111142850A (en) * | 2019-12-23 | 2020-05-12 | 南京航空航天大学 | Code segment recommendation method and device based on deep neural network |
| CN111191002A (en) * | 2019-12-26 | 2020-05-22 | 武汉大学 | Neural code searching method and device based on hierarchical embedding |
| CN111459491A (en) * | 2020-03-17 | 2020-07-28 | 南京航空航天大学 | Code recommendation method based on tree neural network |
| CN111522839A (en) * | 2020-04-25 | 2020-08-11 | 华中科技大学 | Natural language query method based on deep learning |
| CN111857660A (en) * | 2020-07-06 | 2020-10-30 | 南京航空航天大学 | Context-aware API recommendation method and terminal based on query statement |
| US11645054B2 (en) | 2021-06-03 | 2023-05-09 | International Business Machines Corporation | Mapping natural language and code segments |
| US11720346B2 (en) | 2020-10-02 | 2023-08-08 | International Business Machines Corporation | Semantic code retrieval using graph matching |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105190597A (en) * | 2012-12-13 | 2015-12-23 | 微软技术许可有限责任公司 | Social-based information recommendation system |
| US9557972B2 (en) * | 2014-03-25 | 2017-01-31 | Electronics And Telecommunications Research Institute | System and method for code recommendation and share |
| CN106462399A (en) * | 2014-06-30 | 2017-02-22 | 微软技术许可有限责任公司 | Code recommendation |
| CN107506414A (en) * | 2017-08-11 | 2017-12-22 | 武汉大学 | A kind of code based on shot and long term memory network recommends method |
-
2018
- 2018-04-24 CN CN201810371788.5A patent/CN108717423B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105190597A (en) * | 2012-12-13 | 2015-12-23 | 微软技术许可有限责任公司 | Social-based information recommendation system |
| US9557972B2 (en) * | 2014-03-25 | 2017-01-31 | Electronics And Telecommunications Research Institute | System and method for code recommendation and share |
| CN106462399A (en) * | 2014-06-30 | 2017-02-22 | 微软技术许可有限责任公司 | Code recommendation |
| CN107506414A (en) * | 2017-08-11 | 2017-12-22 | 武汉大学 | A kind of code based on shot and long term memory network recommends method |
Non-Patent Citations (1)
| Title |
|---|
| 吕飞: "基于搜索的代码推荐技术研究", 《万方学位论文数据库 硕士论文》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670022A (en) * | 2018-12-13 | 2019-04-23 | 南京航空航天大学 | A kind of java application interface use pattern recommended method based on semantic similarity |
| CN109670022B (en) * | 2018-12-13 | 2023-09-29 | 南京航空航天大学 | Java application program interface use mode recommendation method based on semantic similarity |
| CN110716749A (en) * | 2019-09-03 | 2020-01-21 | 东南大学 | Code searching method based on function similarity matching |
| CN110716749B (en) * | 2019-09-03 | 2023-08-04 | 东南大学 | Code searching method based on functional similarity matching |
| CN110806861A (en) * | 2019-10-10 | 2020-02-18 | 南京航空航天大学 | API recommendation method and terminal combining user feedback information |
| CN111061935B (en) * | 2019-12-16 | 2022-04-12 | 北京理工大学 | Science and technology writing recommendation method based on self-attention mechanism |
| CN111061935A (en) * | 2019-12-16 | 2020-04-24 | 北京理工大学 | Science and technology writing recommendation method based on self-attention mechanism |
| CN111142850A (en) * | 2019-12-23 | 2020-05-12 | 南京航空航天大学 | Code segment recommendation method and device based on deep neural network |
| CN111191002A (en) * | 2019-12-26 | 2020-05-22 | 武汉大学 | Neural code searching method and device based on hierarchical embedding |
| CN111459491B (en) * | 2020-03-17 | 2021-11-05 | 南京航空航天大学 | Code recommendation method based on tree neural network |
| CN111459491A (en) * | 2020-03-17 | 2020-07-28 | 南京航空航天大学 | Code recommendation method based on tree neural network |
| CN111522839A (en) * | 2020-04-25 | 2020-08-11 | 华中科技大学 | Natural language query method based on deep learning |
| CN111522839B (en) * | 2020-04-25 | 2023-09-01 | 华中科技大学 | Deep learning-based natural language query method |
| CN111857660A (en) * | 2020-07-06 | 2020-10-30 | 南京航空航天大学 | Context-aware API recommendation method and terminal based on query statement |
| US11720346B2 (en) | 2020-10-02 | 2023-08-08 | International Business Machines Corporation | Semantic code retrieval using graph matching |
| US11645054B2 (en) | 2021-06-03 | 2023-05-09 | International Business Machines Corporation | Mapping natural language and code segments |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108717423B (en) | 2020-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108717423A (en) | A kind of code segment recommendation method excavated based on deep semantic | |
| CN108446540B (en) | Program code plagiarism type detection method and system based on source code multi-label graph neural network | |
| Shi et al. | Cast: Enhancing code summarization with hierarchical splitting and reconstruction of abstract syntax trees | |
| WO2020010834A1 (en) | Faq question and answer library generalization method, apparatus, and device | |
| Ockeloen et al. | BiographyNet: Managing Provenance at Multiple Levels and from Different Perspectives. | |
| CN110377751A (en) | Courseware intelligent generation method, device, computer equipment and storage medium | |
| CN109885665A (en) | A kind of data query method, apparatus and system | |
| CN110532358A (en) | A kind of template automatic generation method towards knowledge base question and answer | |
| CN107656921A (en) | A kind of short text dependency analysis method based on deep learning | |
| Ell et al. | SPARQL query verbalization for explaining semantic search engine queries | |
| CN110162297A (en) | A kind of source code fragment natural language description automatic generation method and system | |
| Agrawal et al. | Semantic analysis of natural language queries using domain ontology for information access from database | |
| Simperl et al. | Combining human and computation intelligence: the case of data interlinking tools | |
| RU2008104155A (en) | METHOD AND SYSTEM OF ORGANIZATION AND FUNCTIONING OF A DATABASE OF REGULATORY DOCUMENTATION | |
| Yang et al. | User story clustering in agile development: a framework and an empirical study | |
| Giordani et al. | Automatic generation and reranking of sql-derived answers to nl questions | |
| CN113779062A (en) | SQL statement generation method and device, storage medium and electronic equipment | |
| CN103019924B (en) | The intelligent evaluating system of input method and method | |
| CN103177089A (en) | Sentence meaning composition relationship lamination identification method based on central blocks | |
| Mvumbi | Natural language interface to relational database: a simplified customization approach | |
| CN109725894A (en) | A kind of elasticsearch code generating method and tool based on java language | |
| Zhang et al. | Beqain: An effective and efficient identifier normalization approach with bert and the question answering system | |
| CN107818078B (en) | Semantic association and matching method for Chinese natural language dialogue | |
| WO2022142015A1 (en) | Process migration method and apparatus, device, and medium | |
| Nguyen et al. | Systematic knowledge acquisition for question analysis |
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 |