CN109299007A - A kind of defect repair person's auto recommending method - Google Patents
A kind of defect repair person's auto recommending method Download PDFInfo
- Publication number
- CN109299007A CN109299007A CN201811088636.0A CN201811088636A CN109299007A CN 109299007 A CN109299007 A CN 109299007A CN 201811088636 A CN201811088636 A CN 201811088636A CN 109299007 A CN109299007 A CN 109299007A
- Authority
- CN
- China
- Prior art keywords
- report
- defect
- defect report
- similarity
- person
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3692—Test management for test results analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/194—Calculation of difference between files
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention proposes a kind of defect repair person's auto recommending methods, and first after a latest report is submitted to system, this method will search for similar report;Then correlated characteristic, such as candidate defect repair person are extracted from similar report.Finally factor in summary is recommended.Experiments have shown that the method for the invention not only increases a possibility that reparation, avoids excessive repetition and recommend, and reduce repair time.
Description
Technical field
The invention belongs to technical field of software engineering, more particularly to a kind of defect repair person's auto recommending method.
Background technique
With the increase of software system complexity and scale, the generation of software defect is also more and more frequent, or even causes huge
Huge economic loss, repairing defect becomes a very urgent job.As the defect tracking system of Bugzilla one kind uses
A large amount of developer manually carries out unified investigation to defect, and the exploitation for needing defect information is fed back in the form of defect report
Person.So they are faced with some problems.It repeats to recommend to be one of them, the defect and formation generated daily due to software systems
Defect report substantial amounts, content it is similar report, similar defect easily recommend different developers, cause repeat push away
It recommends, so as to cause the difficulty and waste of time in reparation.
Summary of the invention
The invention aims to solve problems of the prior art, proposes a kind of defect repair person and recommend automatically
Method.The method of the invention not only increases a possibility that reparation, avoids excessive repetition and recommends, and reduces reparation
Time.
The purpose of the present invention is achieved through the following technical solutions: a kind of defect repair person's auto recommending method, including following
Step:
Step 1: newly-generated defect report is added to database;
Step 2: being pre-processed using natural language processing technique to defect report, and being calculated using the cosine law should
The similarity of defect report and other defect report;
Step 3: newly-generated defect report is referred to and the highest one kind of his similarity using support vector machines
In;
Step 4: extracting correlated characteristic;
Step 5: executing reparation person's proposed algorithm, final reparation person's recommendation list is obtained.
Further, the similarity that the defect report and other defect report are calculated using the cosine law, specifically:
The similarity between two defect reports is calculated using cosine similarity, formula is as follows:
Wherein, ViAnd VjIndicate the vector of different defect reports, WkiRefer to the weight of k-th of word in defect report di, WkjRefer to
The weight of k-th of word in defect report dj, n represent the size of word set;
The value of two weights is calculated by TF-IDF, and formula is as follows:
Wherein, tfkiIt is the frequency of k-th of word in defect report di, N is total defect report number, nkRepresent k-th of word extremely
Occurs primary defect report number less;
Title and the description of defect report are only extracted, is independently calculated between two defect report titles between description
Text similar value, obtain the similarity (SBBR) between two defect reports according to the following formula:
SBBR(bi, bj)=α × tsij+(1-α)×dsij
Wherein, tsijRefer to the text similar value between defect report bi and bj title, dsijRefer to defect report bi and bj
Text similar value between description, α index topic relative weighting shared in defect report.
Further, the step 3 specifically:
Using K-means clustering algorithm, existing defect report is clustered to obtain similar report collection, the K-means is poly-
Class algorithm description are as follows: 1) K defect report is in space arbitrarily placed as reference data points, and guarantees that they disperse enough,
The reference data points are by the centroid as initial cluster;2) each defect report is assigned to that closest to him of centroid
In cluster;3) position of form center of each cluster is recalculated after all defect reports are all assigned;4) step 2) and 3) straight is repeated
Until centroid is no longer mobile;
The similarity degree between two defect reports is being defined as described two defect reports and centroid in clustering algorithm
The distance between;The clustering algorithm will be iterated always process until error metric E reaches minimum, claim centroid not at this time
It moves again;The formula of E is as follows:
Wherein, K is the number of cluster, and N is the sum for participating in the defect report of cluster, and Xn is the vector of n-th of defect report,
μjRefer to the centroid of j-th of cluster;
All defect report in training set is all assigned in each similar report collection { B1, B2 ..., Bn }, report collection
Each of report it is similar;When a new defect report arrives, the new defect report and similar report are calculated
The similarity for concentrating each to report, be just regarded as when finding maximum similarity MAXSBBR new defect report to it is similar
Similarity between report collection, then distinguishes whether new defect report belongs to some collection using SVM;SVMPredict is exactly
SVM distinguishes the expression of model, and SVMPredict will find out a possibility that new defect report belongs to each collection, when obtaining
After the value of some SVMPredict, its maximum value is taken to be compared with threshold θ, if more than θ, then belongs to some collection, if being less than
Then new defect report is independent at a collection, dissimilar with other reports.
Further, the correlated characteristic is reparation person and repair time.
Further, the step 5 specifically:
After the correlated characteristic for extracting defect report, the factor for influencing presentee is quantified, the factor is to repair
Efficiency and reparation experience;
Remediation efficiency:
Wherein, deviCandidate restoration person is represented, bj refers to the person of being repaired deviThe defect report repaired, F (bj,devi) refer to
The repair time of defect report b, n refer to reparation person deviThe report total of reparation;
Reparation experience:
Wherein, " # " indicates " ... number ";
It executes reparation person's proposed algorithm (DRA):
Wherein, β and γ is the weight of two different factors, meets β+γ=1, M, which indicates to work as, has new defect report to reach
When, all recommendable numbers.
The advantages of the method for the invention are as follows: when carrying out similarity calculation using the cosine law, report is described
It is assigned to different weights respectively with report heading, indicates the importance difference of the corresponding word in these two types of texts to improve information
The accuracy of retrieval.Consider the association attributes of recommended reparation person, such as the efficiency of reparation person, the experience etc. of reparation person,
As the whether suitable foundation of Standard Judgement reparation person.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of defect repair person's auto recommending method of the present invention;
Fig. 2 is that F-measure value-reparation person of each method recommends number curve figure.
Specific embodiment
Technical solution in the embodiment of the present invention that following will be combined with the drawings in the embodiments of the present invention carries out clear, complete
Ground description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based on this
Embodiment in invention, every other reality obtained by those of ordinary skill in the art without making creative efforts
Example is applied, shall fall within the protection scope of the present invention.
In conjunction with Fig. 1, the present invention proposes a kind of defect repair person's auto recommending method, comprising the following steps:
Step 1: newly-generated defect report is added to database;
Step 2: pre-processing using natural language processing technique (NLP) to defect report, and utilize cosine law meter
Calculate the similarity of the defect report and other defect report;The pretreatment specifically:
NLTK and TEXTBLOB are performed the steps of in the library system loads python
(1) segment: defect report or user comment information are cut into several vocabulary, these vocabulary are used to calculate text
This similarity.
(2) stop-word removes: some stop-words (such as " the " " a " " are " etc.) frequently occur in English text but
There is no any specific meaning to defect location.Stop word list according to WordNet, system will remove these vocabulary.
(3) root: all words will be converted into their root morphology, that is to say, that third-person singular, mistake
When going and the tenses such as future tense can be converted into the original form of vocabulary.
(4) it noun and verb screening: is identified in defect report and user comment information by load POS labeling module
Verb and noun.Only these vocabulary are used to calculate text similarity, because they are most representative meanings in text
The vocabulary of justice.
Step 3: newly-generated defect report is referred to and the highest one kind of his similarity using support vector machines
In;
Step 4: extracting correlated characteristic;The correlated characteristic is reparation person and repair time.
Step 5: executing reparation person's proposed algorithm, final reparation person's recommendation list is obtained.
The similarity that the defect report and other defect report are calculated using the cosine law, specifically:
The similarity between two defect reports is calculated using cosine similarity, formula is as follows:
Wherein, ViAnd VjIndicate the vector of different defect reports, WkiRefer to the weight of k-th of word in defect report di, WkjRefer to
The weight of k-th of word in defect report dj, n represent the size of word set;
The value of two weights is calculated by TF-IDF, and formula is as follows:
Wherein, tfkiIt is the frequency of k-th of word in defect report di, N is total defect report number, nkRepresent k-th of word extremely
Occurs primary defect report number less;
Title and the description of defect report are only extracted, is independently calculated between two defect report titles between description
Text similar value, obtain the similarity (SBBR) between two defect reports according to the following formula:
SBBR(bi, bj)=α × tsij+(1-α)×dsij
Wherein, tsijRefer to the text similar value between defect report bi and bj title, dsijRefer to defect report bi and bj
Text similar value between description, α index topic relative weighting shared in defect report.
The step 3 specifically:
When a new defect report arrives, the similar defect report retrieval based on SVM will start to execute, for new report
Announcement finds one and the highest class of his similarity.It needs to create example before this to enable SVM to be used, this is just needed using K-
Means clustering algorithm clusters existing defect report to obtain similar report collection, the K-means clustering algorithm description are as follows: 1)
K defect report is arbitrarily placed in space as reference data points, and guarantees that they disperse enough, and the reference data points will
Centroid as initial cluster;2) each defect report is assigned in that closest to him cluster of centroid;3) when all
Defect report recalculates the position of form center of each cluster after being all assigned;4) step 2) is repeated and 3) until centroid no longer movement is
Only;
The similarity degree between two defect reports is being defined as described two defect reports and centroid in clustering algorithm
The distance between (or data point);The clustering algorithm will be iterated always process until error metric E reaches minimum,
Centroid is claimed no longer to move at this time;The formula of E is as follows:
Wherein, K is the number of cluster, and N is the sum for participating in the defect report of cluster, and Xn is the vector of n-th of defect report,
μjRefer to the centroid of j-th of cluster;
All defect report in training set is all assigned in each similar report collection { B1, B2 ..., Bn }, report collection
Each of report it is similar;When a new defect report arrives, the new defect report and similar report are calculated
The similarity for concentrating each to report, be just regarded as when finding maximum similarity MAXSBBR new defect report to it is similar
Similarity between report collection, then distinguishes whether new defect report belongs to some collection using SVM;SVMPredict is exactly
SVM distinguishes the expression of model, and SVMPredict will find out a possibility that new defect report belongs to each collection, when obtaining
After the value of some SVMPredict, its maximum value is taken to be compared with threshold θ, if more than θ, then belongs to some collection, if being less than
Then new defect report is independent at a collection, dissimilar with other reports.
The step 5 specifically:
After the correlated characteristic for extracting defect report, the factor for influencing presentee is quantified, the factor is to repair
Efficiency and reparation experience;
Remediation efficiency:
Wherein, deviCandidate restoration person is represented, bj refers to the person of being repaired deviThe defect report repaired, F (bj, devi) refer to
The repair time of defect report b, n refer to reparation person deviThe report total of reparation;
Reparation experience:
Wherein, " # " indicates " ... number ";
It executes reparation person's proposed algorithm (DRA):
Wherein, β and γ is the weight of two different factors, meets β+γ=1, M, which indicates to work as, has new defect report to reach
When, all recommendable numbers.
The present invention has chosen 1888 parts labeled as settled defect report from the open defect warehouse open source projects JBOSS
Accuse, the method for the invention be trained, is assessed, Fig. 2 shows that F value is better than other methods, it is maximum up to 82.5% and
The DREX with Frequency to behave oneself best in other methods, only 80%.The method of the invention is then embodied in table 1
The significant effect of defect repair time is reduced, average cost is 01 days two months, DREX with more second-best than effect
The 01 days three months reduction one months that Degree is spent.(note: F-measure score value is a kind of measurement model accuracy
Index, be considered as the weighted harmonic mean value for accuracy rate and recall rate).
Table 1 predicts repair time comparison result
Note: SVM-based, which refers to, calls the report being solved in SVM study defect library to make when there is latest report to enter
It is assigned in existing classification with classifier, and the suitable developer of sub-fraction is recommended to repair defect.DREX with
Frequency, Degree are a kind of measurements for reparation person's grading, and Frequency refers to comment of the reparation person in report
Record, Degree refer to the connection number of a certain node.
Above to a kind of defect repair person's auto recommending method provided by the present invention, it is described in detail, herein
Apply that a specific example illustrates the principle and implementation of the invention, the explanation of above example is only intended to help
Understand method and its core concept of the invention;At the same time, for those skilled in the art, according to the thought of the present invention,
There will be changes in the specific implementation manner and application range, in conclusion the content of the present specification should not be construed as to this
The limitation of invention.
Claims (5)
1. a kind of defect repair person's auto recommending method, which comprises the following steps:
Step 1: newly-generated defect report is added to database;
Step 2: pre-processing using natural language processing technique to defect report, and the defect is calculated using the cosine law
The similarity of report and other defect report;
Step 3: using support vector machines by newly-generated defect report be referred to his similarity it is highest one kind in;
Step 4: extracting correlated characteristic;
Step 5: executing reparation person's proposed algorithm, final reparation person's recommendation list is obtained.
2. the method according to claim 1, wherein described calculate the defect report and other using the cosine law
The similarity of defect report, specifically:
The similarity between two defect reports is calculated using cosine similarity, formula is as follows:
Wherein, ViAnd VjIndicate the vector of different defect reports, WkiRefer to the weight of k-th of word in defect report di, WkjRefer to defect
Report the weight of k-th of word in dj, n represents the size of word set;
The value of two weights is calculated by TF-IDF, and formula is as follows:
Wherein, tfkiIt is the frequency of k-th of word in defect report di, N is total defect report number, nkK-th of word is represented at least to go out
Now primary defect report number;
Title and the description of defect report are only extracted, the text between two defect report titles between description is independently calculated
This similar value obtains the similarity (SBBR) between two defect reports according to the following formula:
SBBR(bi, bj)=α × tsij+(1-α)xdsij
Wherein, tsijRefer to the text similar value between defect report bi and bj title, dsijRefer to that defect report bi and bj are described
Between text similar value, α index topic relative weighting shared in defect report.
3. according to the method described in claim 2, it is characterized in that, the step 3 specifically:
Using K-means clustering algorithm, existing defect report is clustered to obtain similar report collection, the K-means cluster is calculated
Method description are as follows: 1) K defect report is in space arbitrarily placed as reference data points, and guarantees that they disperse enough, it is described
Reference data points are by the centroid as initial cluster;2) each defect report is assigned to that closest to him cluster of centroid
In;3) position of form center of each cluster is recalculated after all defect reports are all assigned;4) repeat step 2) and 3) until
Until centroid no longer moves;
The similarity degree between two defect reports is being defined as between described two defect reports and centroid in clustering algorithm
Distance;The clustering algorithm will be iterated always process until error metric E reaches minimum, and title centroid is no longer moved at this time
It is dynamic;The formula of E is as follows:
Wherein, K is the number of cluster, and N is the sum for participating in the defect report of cluster, and Xn is the vector of n-th of defect report, μjRefer to
The centroid of j-th of cluster;
All defect report in training set is all assigned in each similar report collection { B1, B2 ..., Bn }, reports concentration
Each report is similar;When a new defect report arrives, calculates the new defect report and similar report is concentrated
The similarity of each report is just regarded as new defect report and similar report when finding maximum similarity MAXSBBR
Then similarity between collection distinguishes whether new defect report belongs to some collection using SVM;SVMPredict is exactly SVM
Distinguish the expression of model, SVMPredict will find out a possibility that new defect report belongs to each collection, when being owned
SVMPredict value after, take its maximum value to be compared with threshold θ, if more than θ, then belong to some collection, if be less than if
New defect report is independent at a collection, dissimilar with other reports.
4. according to the method described in claim 3, it is characterized in that, the correlated characteristic is reparation person and repair time.
5. according to the method described in claim 4, it is characterized in that, the step 5 specifically:
After the correlated characteristic for extracting defect report, the factor for influencing presentee is quantified, the factor is remediation efficiency
With the experience of reparation;
Remediation efficiency:
Wherein, deviCandidate restoration person is represented, bj refers to the person of being repaired deviThe defect report repaired, F (bj,devi) refer to defect report
The repair time of b is accused, n refers to reparation person deviThe report total of reparation;
Reparation experience:
Wherein, " # " indicates " ... number ";
It executes reparation person's proposed algorithm (DRA):
Wherein, β and γ is the weight of two different factors, meets β+γ=1, and M is indicated when there is new defect report to reach, institute
There is recommendable number.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811088636.0A CN109299007A (en) | 2018-09-18 | 2018-09-18 | A kind of defect repair person's auto recommending method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811088636.0A CN109299007A (en) | 2018-09-18 | 2018-09-18 | A kind of defect repair person's auto recommending method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109299007A true CN109299007A (en) | 2019-02-01 |
Family
ID=65163590
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811088636.0A Pending CN109299007A (en) | 2018-09-18 | 2018-09-18 | A kind of defect repair person's auto recommending method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109299007A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110442514A (en) * | 2019-07-11 | 2019-11-12 | 扬州大学 | The method that defect repair is recommended is realized based on learning algorithm |
CN112667492A (en) * | 2020-11-06 | 2021-04-16 | 北京工业大学 | Recommendation method for software defect report repairer |
WO2021143056A1 (en) * | 2020-01-16 | 2021-07-22 | 平安科技(深圳)有限公司 | Text conclusion intelligent recommendation method and apparatus, computer device and computer-readable storage medium |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107957929A (en) * | 2017-11-20 | 2018-04-24 | 南京大学 | A kind of software deficiency report based on topic model repairs personnel assignment method |
-
2018
- 2018-09-18 CN CN201811088636.0A patent/CN109299007A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107957929A (en) * | 2017-11-20 | 2018-04-24 | 南京大学 | A kind of software deficiency report based on topic model repairs personnel assignment method |
Non-Patent Citations (3)
Title |
---|
GEUNSEOK YANG 等: "An Emotion Similarity Based Severity Prediction of Software Bugs:A Case Study of Open Source Projects", 《IEICE TRANS. INF. & SYST》 * |
TAO ZHANG 等: "How to Recommend Appropriate Developers for Bug Fixing?", 《2012 IEEE 36TH ANNUAL COMPUTER SOFTWARE AND APPLICATIONS CONFERENCE》 * |
史高翔 等: "基于缺陷相似度与再分配图的软件缺陷分配方法", 《计算机科学》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110442514A (en) * | 2019-07-11 | 2019-11-12 | 扬州大学 | The method that defect repair is recommended is realized based on learning algorithm |
CN110442514B (en) * | 2019-07-11 | 2024-01-12 | 扬州大学 | Method for realizing defect repair recommendation based on learning algorithm |
WO2021143056A1 (en) * | 2020-01-16 | 2021-07-22 | 平安科技(深圳)有限公司 | Text conclusion intelligent recommendation method and apparatus, computer device and computer-readable storage medium |
CN112667492A (en) * | 2020-11-06 | 2021-04-16 | 北京工业大学 | Recommendation method for software defect report repairer |
CN112667492B (en) * | 2020-11-06 | 2024-03-08 | 北京工业大学 | Software defect report repairman recommendation method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110799981B (en) | Systems and methods for domain-independent aspect level emotion detection | |
Swathi et al. | An optimal deep learning-based LSTM for stock price prediction using twitter sentiment analysis | |
Deepak et al. | A novel firefly driven scheme for resume parsing and matching based on entity linking paradigm | |
Jain et al. | Sentiment classification of twitter data belonging to renewable energy using machine learning | |
CN111767725B (en) | Data processing method and device based on emotion polarity analysis model | |
Sanborn et al. | Deep learning for semantic similarity | |
Chen et al. | Improving distributed representation of word sense via wordnet gloss composition and context clustering | |
US11030533B2 (en) | Method and system for generating a transitory sentiment community | |
CN109299007A (en) | A kind of defect repair person's auto recommending method | |
Chirawichitchai | Sentiment classification by a hybrid method of greedy search and multinomial naïve bayes algorithm | |
Hasan et al. | Sentiment analysis using out of core learning | |
CN108694176B (en) | Document emotion analysis method and device, electronic equipment and readable storage medium | |
US20220148049A1 (en) | Method and system for initiating an interface concurrent with generation of a transitory sentiment community | |
Hashemzadeh et al. | Improving keyword extraction in multilingual texts. | |
CN114722176A (en) | Intelligent question answering method, device, medium and electronic equipment | |
US11605004B2 (en) | Method and system for generating a transitory sentiment community | |
Jawad et al. | Combination of convolution neural networks and deep neural networks for fake news detection | |
CN112905796A (en) | Text emotion classification method and system based on re-attention mechanism | |
Syn et al. | Using latent semantic analysis to identify quality in use (qu) indicators from user reviews | |
CN111563361A (en) | Text label extraction method and device and storage medium | |
Kostkina et al. | Document categorization based on usage of features reduction with synonyms clustering in weak semantic map | |
Zhang et al. | Sentiment identification by incorporating syntax, semantics and context information | |
CN114328903A (en) | Text clustering-based customer service log backflow method and device | |
CN110609892A (en) | OA intelligent recommendation system based on AI | |
Khodorchenko | Distant supervision and knowledge transfer for domain-oriented text classification in online social networks |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190201 |