Sign-language is the primary mode of communication for the hearing- or speech-impaired people. With the technological advancements in machine-learning and computational technologies the models are developed for the translating sign-language to other natural language. In this paper, a two-way translation model based on CNN-LSTM is presented for the translation of Indian sign language to one of the India's regional language, Tamil. The proposed model is trained and tested for a subset of commonly used words and resulted in an accuracy of 92.9% and a precision of 93.7%. This two-way ISL communication system aims to revolutionize the way ISL is taught and learned.