Facebook AI Open-Sourced ‘TransCoder’: A Deep Learning Based Self-Supervised Neural Transcompiler System

0
2778
https://ai.facebook.com/blog/deep-learning-to-translate-between-programming-languages/

There are many programming languages utilized today, from the earliest COBOL, Fortran to a current significant-level programming language like C++, Java, and Python. The absolute most famous advances today incorporate JavaScript, Python, Java, C#, and C++.

Business, account, and regulatory frameworks for organizations and governments mainly utilize COBOL. Because of its declining prominence and the retirement of experienced COBOL developers, programs are being relocated to new stages or revamped into modern-day languages. But this is causing fortune to companies as the translation is an asset concentrated undertaking that requires mastery in both the source and target languages. 

Advertisement

The group at Facebook AI Research (FAIR) has built up an AI model, TransCoder. FAIR plans to take care of this issue via automation, with the assistance of cutting edge Deep Learning technology to translate between 3 accessible programming languages utilized today.

The main differentiating factor of TransCoder is that it is self-supervised training.TransCoder depends solely on source code written in only one programming language, as opposed to requiring instances of similar codes in both the source and target language. It requires no expertise in programming languages. 

TransCoder utilizes a sequence-to-sequence (seq2seq) model with an encoder and a decoder with a transformer design. The model uses the three principles of unsupervised Machine Translation, as shown below. 

https://ai.facebook.com/blog/deep-learning-to-translate-between-programming-languages/

Real-world applications of TransCoder in future:

  1. Programmers working in a company or on an open-source project can integrate programs of other languages into their plan for making it more efficient.
  2. Will help people who don’t have time or resource to learn programming in multiple languages
  3. Helps in reducing the effort and expense of updating an old codebase written in archaic language
  4. Companies can update their codebase to any modern languages available.

However, with upgrades in source-to-source translation, intelligent machines will enhance automation and may pose a threat to a couple of employments, particularly in testing. The eventual fate of programming is still unknown!

https://ai.facebook.com/blog/deep-learning-to-translate-between-programming-languages/

Paper: https://arxiv.org/pdf/2006.03511.pdf

GitHub: https://github.com/facebookresearch 

Source: https://ai.facebook.com/blog/deep-learning-to-translate-between-programming-languages/

LEAVE A REPLY

Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.