We Translate Spoken Language Translator

Abstract

Considering the low literacy rate of Pakistan (58%) and an even poor literacy criteria, 60% of Pakistanis are unable to read and write a sentence in English. Therefore, when Pakistanis go abroad, they face such language barriers where they are unable to communicate with the locals there. Similarly for the tourists who visit Pakistan and specifically visit the northern areas, they face problems while communicating. Hence to overcome these problems, we have made a system called, “WeTranslate”, a spoken language translator that translates speech input from one language to another which in our case is restricted to English-Urdu and vice-versa. The input is given in the form of speech, which gets recognized and then is converted from the source language (the language that it is spoken in) to the target language (the language that we want it to get translated to). Existing solutions such as Google Translate, does provide English to Urdu and vice-versa translation but it doesn’t work offline which means that one needs a continuous internet connection for the application to work which is a constraint but with WeTranslate, the system can work offline. Also, speech recognition of Urdu is required considering the literacy rate and criteria which makes a lot of Pakistanis incompetent in writing a full sentence in either language hence to give their input in speech remains more suitable but Google Translate doesn’t provide speech recognition for Urdu therefore we provide recognition for those who are unable to write sentences in either language. The accuracy of the combined system is around 75% which includes recognition for recorded audios and those recognized texts getting translated. Work on live audio is still in progress so that the accuracy gets better for a more accurate integrated system. Aso, this system will be used to translate English videos to urdu to facilitate those who are unable to understand English.