Overview

This repository contains the data and implementation for the paper Hierarchical Learning of Cross-Language Mappings through Distributed Vector Representations for Code , in Proceedings of the IEEE/ACM 40th International Conference on Software Engineering: New Ideas and Emerging Technologies Results Track (ICSE-NIER), Gothenburg, Sweden, 2018 .

Since we found that the implementation of BiSkip2Vec from Multivec has better performance than our implementation, we recommend the researchers that want to reproduce our result use Multivec for training the token embeddings.

Data

• DATA/DATA_RAW.tar.gz: contains the raw data of these open source projects: antlr, cordova, datastax, factual, fpml, log4j, lucene, spring, uap, zeromq
• DATA/TRAINING_SENTENCES: contains the processed data of aligned sentences between C# and Java, which will use as the input for Multivec.
• DATA/SIGNATURE: all of the method signatures for each of the project

Result

• The mapping that we found can be access at mapping results

• We also use the mappings mined from our work in the configuration file of Java2Csharp: https://github.com/bdqnghi/sharpen-java2csharp.