Abstract: This project presents an innovative offline speech translator that utilizes Natural Language Processing (NLP) and Natural Language Toolkit (NLTK) to enable real-time language translation. Our system integrates automatic speech recognition (ASR) and machine translation (MT) to facilitate accurate and efficient language translation. We employ a cascaded architecture, incorporating NLTK's tokenization, stemming, and lemmatization techniques to enhance text preprocessing. Experimental results demonstrate the effectiveness of our approach, achieving competitive translation accuracy on benchmark datasets. Our offline speech translator has far-reaching implications for global communication, enabling individuals to transcend language barriers and connect with others in real-time, regardless of internet connectivity.

Keywords: Offline speech translation, NLP, NLTK, ASR, MT, real-time translation, gTTs, pyttsx3, Argos Translate, Vosk

| DOI: 10.17148/IARJSET.2025.12438