Enhancing ASR by MT using Semantic Information from HindiWordNet

Aniruddha Tammewar, Karan Singla, Srinivas Bangalore, Michael Carl

    Research output: Contribution to conferencePosterResearchpeer-review


    In a conventional CAT (Computer Assisted Translation) system a human translator post-edits an automatically generated target language text using the keyboard. In this paper we extend a CAT system with speech input by which the translator speaks the translation, a process refered to as sight translation. We report several experiments to improve the performance of an automatic speech recognition system, taking advantage of machine translation output and information fromWordNet. Overall we outperform a baseline system which has no semantic information by an increased 1.6% word accuracy for the English to Hindi translation.
    Original languageEnglish
    Publication date2013
    Publication statusPublished - 2013
    Event10th International Conference on Natural Language Processing - Centre for Development of Advanced Computing, Noida, India
    Duration: 18 Dec 201320 Dec 2013
    Conference number: 10


    Conference10th International Conference on Natural Language Processing
    LocationCentre for Development of Advanced Computing
    Internet address

    Cite this