Facebook Research - Unsupervised Translation of Programming Languages
Published:Jun 24, 2020 16:50
•1 min read
•ML Street Talk Pod
Analysis
The article highlights a new approach to programming language translation by Facebook Research, focusing on unsupervised learning. The core innovation is the use of word-piece embeddings to leverage token overlap between languages, eliminating the need for parallel data. The article also mentions the researchers involved, the source of the information (ML Street Talk Pod), and provides links to the paper and a related video.
Key Takeaways
- •Facebook Research developed an unsupervised method for translating programming languages.
- •The method uses word-piece embeddings to exploit token overlap.
- •The approach eliminates the need for parallel language data.
- •The research was presented on the ML Street Talk Pod and has a related video.
Reference
“The article doesn't contain a direct quote, but it references the paper's abstract, which describes the problem of transcompilation and the limitations of existing methods.”