XTest: A Parallel Multilingual Corpus with Test Cases for Code Translation and Its Evaluation*

The advancement of artificial intelligence has significantly improved the automated generation, summarization and translation of source code. Researchers are working together to improve the automated coding tasks to bring the next generation of programming systems. Currently there are few corpora available for automated code generation, summarization and translation tasks but these corpora and models are limited within common languages like C, C++, Python, Java, etc. Also none of the existing corpus contains appropriate test cases to evaluate a program properly. This paper introduces XTest: A Parallel Multilingual Corpus with Test Cases for Code Translation. Our dataset contains parallel programs in 9 languages, Problem statement and test cases. Also in this work, we have built 30 systems to translate code between some high-resourced(C++, Python, etc) and low-resourced (Go, Ruby, etc) programming languages.