A statistical semantic language model for source code

Nguyen, Tung; Nguyen, Anh Tuan; Nguyen, Hoan Anh; Nguyen, Tien N.

doi:10.1145/2491411.2491458

Cited by 180 publications

(117 citation statements)

References 15 publications

Supporting

Mentioning

117

Contrasting

Order By: Relevance

“…They evaluate the practical potential of the high regularity that language models found in source code by building a code completion tool and showing that it can significantly improve the default suggestions given by the Eclipse IDE. 1 Subsequent work investigates the potential of these models in a number of settings, such as their applicability to mining repositories at massive scale [2], and the influence factors such as semantic information [24] and local context [35] on code completion performance. In particular, Allamanis et al investigate whether coding conventions can be extracted automatically from OSS projects using language models [1].…”

Section: B Language Modelsmentioning

confidence: 99%

Will They Like This? Evaluating Code Contributions with Language Models

Hellendoorn¹,

Dévanbu²,

Bacchelli³

2015

2015 IEEE/ACM 12th Working Conference on Mining Software Repositories

View full text Add to dashboard Cite

Popular open-source software projects receive and review contributions from a diverse array of developers, many of whom have little to no prior involvement with the project. A recent survey reported that reviewers consider conformance to the project's code style to be one of the top priorities when evaluating code contributions on Github. We propose to quantitatively evaluate the existence and effects of this phenomenon. To this aim we use language models, which were shown to accurately capture stylistic aspects of code. We find that rejected changesets do contain code significantly less similar to the project than accepted ones; furthermore, the less similar changesets are more likely to be subject to thorough review. Armed with these results we further investigate whether new contributors learn to conform to the project style and find that experience is positively correlated with conformance to the project's code style.

show abstract

Section: B Language Modelsmentioning

confidence: 99%

Will They Like This? Evaluating Code Contributions with Language Models

Hellendoorn¹,

Dévanbu²,

Bacchelli³

2015

2015 IEEE/ACM 12th Working Conference on Mining Software Repositories

View full text Add to dashboard Cite

show abstract

“…NLP techniques have been shown to be useful in requirements engineering [19,43,44], usability of API documents [69,70], generation of program comments [23,34,47], code completion [22,37,41], and other tasks [50]. For example, Zhong et al [70] employed NLP and machine learning (ML) techniques to infer resource specifications from API documents.…”

Section: Text Analysis For Software Engineeringmentioning

confidence: 99%

Proactive detection of inadequate diagnostic messages for software configuration errors

Zhang

Ernst

2015

Proceedings of the 2015 International Symposium on Software Testing and Analysis

View full text Add to dashboard Cite

This paper presents a technique to detect inadequate (i.e., missing or ambiguous) diagnostic messages for configuration errors issued by a configurable software system.The technique injects configuration errors into the software under test, monitors the software outcomes under the injected configuration errors, and uses natural language processing to analyze the output diagnostic message caused by each configuration error. The technique reports diagnostic messages that may be unhelpful in diagnosing a configuration error.We implemented the technique for Java in a tool, ConfDiagDetector. In an evaluation on 4 real-world, mature configurable systems, ConfDiagDetector reported 43 distinct inadequate diagnostic messages (25 missing and 18 ambiguous). 30 of the detected messages have been confirmed by their developers, and 12 more have been identified as inadequate by users in a user study. On average, ConfDiagDetector required 5 minutes of programmer time and 3 minutes of compute time to detect each inadequate diagnostic message.

show abstract

“…SLAMC [12] is a statistical semantic language model for source code that introduces semantic information into the language models, which represents additional information in the language model that involves, for instance, the global context of source files to help predict the next token. Tu et al [28] however, argue that code tokenization is enough for the n-gram language models.…”

Section: Methods Call Recommendersmentioning

confidence: 99%

“…On the one hand, API usability was studied through empirical experiments (e.g., [2,3,4,5,6]), which revealed the types of barriers that API users face (e.g., relationships between types and instantiation of abstract types). On the other hand, recommendation systems were proposed as an aid to assist API users through IDE (Integrated Development Environment) code completion mechanisms, which rely either on structural analysis (e.g., [7,8]) or on mining patterns from source code (e.g., [9,10,11,12,13]). Although code completion is an IDE feature that boosts programmer productivity with respect to code writing using a familiar API, empirical studies demonstrated that code completion is often used to explore, and hence, learn an unfamiliar API (e.g., [3,6]).…”

Section: Introductionmentioning

confidence: 99%

Stepwise API usage assistance using n -gram language models

Santos

Prendi

Sousa

et al. 2017

Journal of Systems and Software

View full text Add to dashboard Cite

Reusing software involves learning third-party APIs, a process that is often time-consuming and error-prone. Recommendation systems for API usage assistance based on statistical models built from source code corpora are capable of assisting API users through code completion mechanisms in IDEs. A valid sequence of API calls involving different types may be regarded as a well-formed sentence of tokens from the API vocabulary. In this article we describe an approach for recommending subsequent tokens to complete API sentences using n-gram language models built from source code corpora. The provided system was integrated in the code completion facilities of the Eclipse IDE, providing contextualized completion proposals for Java taking into account the nearest lines of code. The approach was evaluated against existing client code of four widely used APIs, revealing that in more than 90% of the cases the expected subsequent token is within the 10-top-most proposals of our models. The high score provides evidence that the recommendations could help on API learning and exploration, namely through the assistance on writing valid API sentences.

show abstract

A statistical semantic language model for source code

Cited by 180 publications

References 15 publications

Will They Like This? Evaluating Code Contributions with Language Models

Will They Like This? Evaluating Code Contributions with Language Models

Proactive detection of inadequate diagnostic messages for software configuration errors

Stepwise API usage assistance using n -gram language models

Contact Info

Product

Resources

About