Syntax errors just aren't natural: improving error reporting with language models

By Hazel Victoria Campbell, Abram Hindle, José Nelson Amaral. 2014

92 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Information Systems, Software

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Information Systems

In Software

Citations in the corpus

The adverse effects of code duplication in machine learning models of code (2018, neutral): “ould be appreciated that not all metrics and tasks are equally affected by code duplication.For example, if an application requires predicting code’s non-identifier tokens (e.g. as in Campbell et al. [10]), duplication would have a much smaller effect compared to an autocompletion application for pr”
Big Code != Big Vocabulary: Open-Vocabulary Models for Source Code (2020, methodology): “Campbell et al. [18] used n -gram language models to detect syntax error locations in Java code, and later used an NLM for the same purpose [78].”
Learning to Fix Build Errors with Graph2Diff Neural Networks (2019, background): “Previous work on syntax errors has been very successful generating repairs resolving missing delimiters and parentheses [10, 15].”
Bugram: Bug detection with n-gram language models (2016, methodology): “[5] built n-gram models with historical correct source code to locate the cause of syntax errors.”
Toward Deep Learning Software Repositories (2015, background): “human oracle cost [7], predicting programmer comments to improve search over code bases and code categorization [8], improving error reporting [10], generating feasible test cases to improve coverage [11], improving stylistic consistency to aid readability and maintainability [13], and code migratio”
On the localness of software (2014, unknown): “As mentioned earlier, good language models will be useful for other applications: code porting [35], coding standards checking [3], correcting syntax errors [12] We emulate code suggestion in two scenarios: coding in new files where only prolog is available (Section 5.”
Neuro-Symbolic Program Corrector for Introductory Programming Assignments (2018, methodology): “Campbell et al. [6] presented a technique to use unnaturalness of code with syntax errors to locate the actual source of errors.”
Mining idioms from source code (2014, methodology): “N -gram language models have been used to improve code autocompletion performance [2, 18, 39], learn coding conventions [3] and find syntax errors [8].”
Deep learning code fragments for code clone detection (2016, methodology): “We enlist a particular type of language model to map the terms in source code to continuous-valued vectors called embeddings (Sec.”
On the "Naturalness" of Buggy Code (2015, background): “In fact, prior work [7] showed that syntactically incorrect code is flagged as improbable by language models.”
Maybe Deep Neural Networks are the Best Choice for Modeling Source Code (2019, methodology): “Also, improved language models for code have the potential to enable new tools for aiding code readability [2], program repair [11, 15, 35, 62], program synthesis [34] and translation be-tween programming languages [45, 56].”
On the naturalness of software (2016, background): “For example, Campbell et al.(10) diagnose syntax error locations by exploiting the fact that language models trained on working code will be surprised by syntax errors.”
Are deep neural networks the best choice for modeling source code? (2017, background): “As we shall see, performance in this final setting is best-in-class, and has ramifications for code synthesis [14, 30], fault localization [8, 31], code naturalization [2] and many real-world code suggestions settings [15, 29].”
Products, developers, and milestones: how should I build my N-Gram language model (2015, methodology): “For example, Movshovitz-Attias and Co-hen demonstrated how to use LMs to generate code comments [4] and Campbell et al. showed they could improve error reporting [6].”