TransCoder to translate programming languages: Facebook AI


Facebook AI’s new technology is an open-sourced ‘TransCoder’ which is self-supervised. It is a neural trans compiler system based on deep learning.

C++, Java, and Python are the current significant level of programming languages. Earlier, the most commonly used programming languages were COBOL and Fortran. JavaScript, Python, Java, C#, and C++ are the absolute most famous advances which are incorporated in today’s world.

Organizations and governments use COBOL for business, account, and regulatory frameworks. COBOL is facing a declining prominence and the experienced COBOL developers are retiring because of which the program is relocated to new stages and revamped into modern-day languages. Relocation to new stages is causing fortune to companies and translation is an asset concentrated undertaking. Translation requires mastery in both the source and target languages.

TransCoder, an AI model is built by the group at Facebook AI Research (FAIR). The issue of translation is planned to be taken care of by FAIR via automation. The assistance of cutting edge Deep Learning technology can be used to translate between 3 accessible programming languages that are used in today’s world.

TransCoder provides self-supervised training which is its main differentiating factor. TransCoder depends only on source code which is written in only one programming language. Instances of requiring similar codes in both the source and target language will be opposed by the TransCoder. Expertise in programming languages is not required for TransCoder.

A sequence-to-sequence (seq2seq) model is utilized by TransCoder which contains an encoder and a decoder with a transformer design. The three principles of unsupervised Machine Translation which include Cross-lingual masked model pretraining, Denoising auto-encoding, Back-translation are being used by TransCoder.

The TransCoder can be used in the future with real-world applications and they are as follows:

  • Programs of other languages can be integrated by programmers working in a company or an open-source project to make their plan more efficient.
  • The transCoder will help people to learn programming in multiple languages for those with no access to time and resources.
  • Updating an old codebase written in an archaic language requires effort and expense which will be reduced by TransCoder.
  • The codebase can be updated to any modern languages available is made possible for the companies.

 Source-to-source translation, intelligent machines when upgraded can enhance automation and when it comes to testing it may pose a threat to a couple of employments. However, the future of programming is still unknown. 


Please enter your comment!
Please enter your name here