Lakshmi Mohandas

Enthusiastic and accomplished researcher with a passion for AI and data science, seeking to leverage my expertise in statistical analysis, machine learning,deep learning and NLP techniques to contribute in the field of AI. Proven ability to translate complex concepts into actionable insights, combined with a strong background in academia and a commitment to excellence.

Recognition of Indian Sign Language (ISL) is critical in overcoming communication gaps among the hearing impaired and the hearing world. To comprehend the nuanced motions of ISL, multiple approaches such as Long Short-Term Memory (LSTM) and Recurrent Neural Networks (RNN) are used in this difficult endeavor. Hand recognition algorithms recognise sign gestures and use convex hull techniques for precise tracking by utilising technologies such as MediaPipe. Furthermore, preprocessing techniques such as unsharp mask filtering improve image quality, which aids in correct recognition. The system uses Pyttsx3 for text- to-speech functionality to enable low latency communication. The incorporation of these cuttingedge technologies enables the system to transform ISL gestures into textual form and vice versa, encouraging inclusion and enabling the hearing-impaired community to fully participate in discussions and information exchange.

As part of my research in Natural Language Processing, I led a project focused on Automatic Text Summarization. The primary objective was to develop algorithms and models capable of summarising large bodies of text into concise and coherent summaries, replicating the essential information present in the original text using state-of-the-art NLP models, such as transformers, and training them on diverse datasets to enhance the model's ability to understand and summarize different types of content. The research findings were documented and published in a peer-reviewed paper.

Through my research, I built a system that can generate questions and answer pairs on images. It consists of two separate modules—Visual Question Generation (VQG) which generates questions based on the image, and a Visual Question Answering (VQA) module that produces a befitting answer that the VQG module generates. Through this approach, not only questions were generated, but questions were also evaluated by using a question answering system. It eliminates the need for human intervention in dataset annotation and also finds applications in the field of the educational sector, where the requirement of human input for textual questions has been essential till now.