Towards English - Mizo Legal Machine Translation: Corpus Creation, Analysis, and Benchmarks

Key points

• Indian court proceedings are held in English, so speakers of languages with limited translation technology have limited access to them.1
• Professional native Mizo translators translated complete documents; sentence pairs were manually aligned, and two independent reviewers checked the results.1
• Our study finds that English words are borrowed in a very selective way: official and legal terms are almost always kept in English, but words for everyday things like actions, situations, and results are usually expressed in native Mizo.1
• Mizo diacritics are also used inconsistently across the corpus.1

Sources (1)

• [1]Towards English - Mizo Legal Machine Translation: Corpus Creation, Analysis, and Benchmarks
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 04:11 PM

  We present the first English Mizo parallel dataset of Indian court judgments.

  Indian court proceedings are held in English, so speakers of languages with limited translation technology have limited access to them.

Extractive summary: sentences quoted from the sources.