The objective of SMarT is to develop models for machine translation and speech recognition systems. This modelisation is based on mathematical methods to identify, extract and find relationships between language patterns which are necessary for translation and speech recognition.
Languages are studied through monolingual, parallel or comparable corpora for under-resourced languages (Arabic dialects) and Arabic, French and English. At present we restrict ourselves to these languages, but nothing prevents us from considering others since the proposed modeling is not based on a linguistic but machine learning study.
SMarT is interested also by the issue of the transformation of the voices altered by certain diseases to make them more intelligible.
Keywords: Statistical modeling of language, Machine translation, Study of under-resourced anguages, Quality of estimation, Evolutionary algorithms, Speech translation, Mining comparable and social network corpora.