Graph-Based Statistical Language Model for Code

By Anh Tuan Nguyen, Tien Nhut Nguyen. 2015

219 citations recorded by the source corpus.

Subfields: Information Systems, Software, Natural Language Processing, Signal Processing

How researchers see this paper

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

In Information Systems

In Software

In Natural Language Processing

In Signal Processing

Citations in the corpus

A Survey of Machine Learning for Big Code and Naturalness (2017, background): “n et al. [140] PD(c|s) Token Phrase Migration Nguyen et al. [144] PD(c) Token + parse info n-gram Code Completion Nguyen et al. [141] PD(c|s) Token + parse info Phrase SMT Migration Nguyen and Nguyen [139] PD(c) Partial PDG n-gram Code Completion Oda et al. [146] PD(c|s) Syntax + Token Tree-to-Strin”
Learning to Generate Pseudo-Code from Source Code Using Statistical Machine Translation (T) (2015, neutral): “…this study: MeCab to tokenize Japanese sentences [29], pialign to train word alignments [22], KenLM to train Kneser-Ney smoothed language model [30], Moses to train and generate target sentences using the PBMT model [31], and Travatar to train and generate target sentences using T2SMT models [32].”
DeepBugs: a learning approach to name-based bug detection (2018, neutral): “As observed by others [Bielik et al. 2016; Nguyen and Nguyen 2015; Raychev et al. 2015], the huge amount of existing code provides ample of examples of likely correct code.”
Automatically Learning Semantic Features for Defect Prediction (2016, background): “It has been shown that pro-grams’ semantic information is useful for tasks such as code completion and bug detection [15,28,44,46,60].”
Deep Learning to Find Bugs (2017, background): “As observed by others [9, 28, 35], the huge amounts of existing code provide ample of examples of likely correct code.”
Bayesian specification learning for finding API usage errors (2017, methodology): “Some recent efforts [40, 41, 52] have also used n-gram models to learn specifications on source code structure.”
When Code Completion Fails: A Case Study on Real-World Completions (2019, methodology): “…without exception, code completion tools are evaluated against synthetic benchmarks , typically produced by adding “holes” to existing code (e.g., removing an identiﬁer) and tasking the model with predicting the correct expression for that location using the remaining context [4], [6], [9]–[12].”
Exploring API Embedding for API Usages and Applications (2017, background): “Typical natural language processing (NLP) models for API call suggestion include n -gram model [3], [4], deep neural network model [5], and graph-based generative model [6].”
Are deep neural networks the best choice for modeling source code? (2017, methodology): “Dynamism Evolution is normal for well-used software systems; bug fixes and new features keep rolling in. NLP corpora evolvemuch more slowly.”
StaQC: A Systematically Mined Question-Code Dataset from Stack Overflow (2018, methodology): “Moreover, we can consider encoders similar to [31, 33] for capturing the intrinsic structure of programming language.”
Semantic code search via equational reasoning (2020, methodology): “These tools are based on synthesis, statistical modeling, or both [4, 19, 52, 53, 62].”
Evaluating the evaluations of code recommender systems: A reality check (2016, background): “Examples include recommending the next syntactic token [19], links to related code snippets [15], and which windows to close [14].”
Bayesian Sketch Learning for Program Synthesis (2017, background): “Another closely related literature focuses on learning hidden specifications [37] and syntactic patterns [4, 10, 31, 32, 37] from code in large software repositories.”
Learning to Fuzz: Application-Independent Fuzz Testing with Probabilistic, Generative Models of Input Data (2016, methodology): “, based on n-grams [7, 9, 38], graphs [37, 42] and recurrent neural networks [43].”