Mistakes in binary conditions are a source of error in many software systems. They happen when developers use < or > instead of <= or >=. These boundary mistakes are hard to find for developers and pose a manual labor-intensive work. While researches have been proposing solutions to identify errors in boundary conditions, the problem remains a challenge. In this thesis, we propose deep learning models to learn mistakes in boundary conditions and train our model on approximately 1.6M examples with faults in different boundary conditions. We achieve an accuracy of 85.06%, a precision of 85.23% and a recall of 84.82% on a controlled dataset. Additionally, we perform tests on 41 real-world boundary condition bugs found from GitHub and try to find bugs from the Java project of Adyen. However, the false-positive rate of the model remains an issue. We hope that this work paves the way for future developments in using deep learning models for defect prediction.

Introduction: Software development is difficult and requires knowledge on many different levels such as understanding programming algorithms, languages, and frameworks. In addition, before code is being worked on, the system requirements and functionality are first discussed in natural language, after which it is sometimes visualized for the developers in a more formal language such as Unified Modeling Language. Recently, researchers have tried to close the gap between natural language description of the system and the actual implementation in code using natural language processing techniques. The techniques from NLP have also proven to be useful at generating code snippets while developers work on source code. This literature survey aims to present an overview of the field of code generation using Natural Language Processing techniques. Method: Google Scholar search engine was used to search for papers regarding code generation using NLP. Results: A total of 428 abstracts were screened to reveal 36 papers suitable for the survey. The found papers were categorized into 6 groups by application type. Conclusion: Source code has similarities to natural language, hence NLP techniques have been successfully used to generate code. Additionally, the area has also benefited from recent deep learning based advances in NLP.