Automated Construction Method of Domain Knowledge Graph for Software Defects

Abstract

The invention relates to a method for automatically constructing domain knowledge graphs oriented to software defects. The present invention extracts bug defect relationship triples, and then classifies the bug defect field, that is, collects bug related information in the bug defect database and processes bug description information, and obtains verb phrases and noun phrases in description sentences through natural language processing. Then, according to the dependency relationship between the words related to the bug information, the bug relationship triples are obtained, and the characteristics of the bug defect field are extracted, and a semi-supervised classifier is used for training and learning, so that the classifier automatically classifies the triples that have not been marked before. group, store all classified relational triples in a graph database, and construct a domain knowledge graph for software defects. The invention overcomes the defect that there is no effective processing method in the bug defect field. The present invention analyzes bug defect descriptions from the perspective of natural language processing, helps software developers and maintainers to understand bug defect information and domain knowledge more intuitively, quickly and accurately, improves efficiency, realizes software defect domain knowledge map, and establishes a relationship between bugs connection relation.

Description

technical field

[0001] The invention belongs to the field of software maintenance, in particular to a method for automatically constructing a domain knowledge graph for software defects. Background technique

[0002] In order to solve some bug problems encountered by software developers and maintainers, they usually need to search for bug related information by means of some websites containing bug defect libraries, such as Bugzilla. However, most of these websites use search methods based on traditional relational databases. Although most of the search results contain a large amount of information and attributes of bugs, the relationship between bugs and defects is not particularly obvious. In addition, when the searcher searches through keywords, the website adopts the method of keyword matching one by one, which will make the search results inaccurate, and when the searcher enters fewer keywords or more keywords, the results will produce two extremes. (1) The search resu...

Examples

Embodiment Construction

[0027] The technical idea of ​​the present invention is:

[0028] This method mainly consists of two parts. First, the first part is the extraction of bug defect relation triples; the second part is the classification of bug defect fields. Specifically, first collect the bug-related information in the bug defect library and focus on processing the description information of the bug, obtain the verb phrases and noun phrases in the description sentence through natural language processing technology, and then use the word and word related to the bug information according to the information. The dependencies between the bugs are obtained to obtain the relationship triplet of bugs. Extract the bug defect domain features, use a semi-supervised classifier for training and learning, and finally make the classifier automatically classify the previously unlabeled triples. Then, all the classified relation triples are stored in the graph database to construct a domain knowledge graph f...

technical field

[0001] The invention belongs to the field of software maintenance, in particular to a method for automatically constructing a domain knowledge graph for software defects. Background technique

[0002] In order to solve some bug problems encountered by software developers and maintainers, they usually need to search for bug related information by means of some websites containing bug defect libraries, such as Bugzilla. However, most of these websites use search methods based on traditional relational databases. Although most of the search results contain a large amount of information and attributes of bugs, the relationship between bugs and defects is not particularly obvious. In addition, when the searcher searches through keywords, the website adopts the method of keyword matching one by one, which will make the search results inaccurate, and when the searcher enters fewer keywords or more keywords, the results will produce two extremes. (1) The search resu...

Examples

Embodiment Construction

[0027] The technical idea of ​​the present invention is:

[0028] This method mainly consists of two parts. First, the first part is the extraction of bug defect relation triples; the second part is the classification of bug defect fields. Specifically, first collect the bug-related information in the bug defect library and focus on processing the description information of the bug, obtain the verb phrases and noun phrases in the description sentence through natural language processing technology, and then use the word and word related to the bug information according to the information. The dependencies between the bugs are obtained to obtain the relationship triplet of bugs. Extract the bug defect domain features, use a semi-supervised classifier for training and learning, and finally make the classifier automatically classify the previously unlabeled triples. Then, all the classified relation triples are stored in the graph database to construct a domain knowledge graph f...

Images