English -Malayalam Vision aid with Multi Modal Machine Learning Technologies
H O Lekshmy, Bharat Jayaraman · 2022 6th International Conference on Intelligent Computing and Control Systems (ICICCS) · 2022
Most of the blind assistive devices accessible in the market operate in the English language. This narrow downs the product’s scope in multilingual countries. To deal with this, some assistive devices make use of translation APIs and modules to make an interpretation of English into the regional language, but these translations are not up to the mark when comparing with the original text. Most of the translation APIs make use of traditional machine translation models that intakes single context as input and produce translations out of it. Linguistic ambiguity is very common among such models. This is mainly because of the morphological and syntactical divergence between the languages. This makes English-Indic language translation a challenging task. This research work proposes an efficient translation algorithm exclusively for blind assistive devices. This approach employs an image-guided multi-modal machine translation methodology, in which an image is used as an additional input to generate more accurate translations/scene distributions on Malayalam.