Rujul Gandhi, an Indian-origin scholar at the MIT-IBM Watson AI Lab, is at the forefront of groundbreaking investigations that examine the intersection of computer science and linguistics.
According to a release by MIT, Gandhi, who began exploring the fusion of these disciplines during her undergraduate years at MIT, is now a key figure in projects that have the potential to revolutionize communication between humans and artificial intelligence.
The main focus of her efforts is the creation of a parser that facilitates the translation of natural language instructions into formats that are compatible with machines. When robots or interactive AI systems are required to comprehend and carry out commands, the task at hand is to convert human-friendly language instructions into a machine-readable format.
In order to address this, Gandhi and her advisors, Yang Zhang of IBM and MIT assistant professor Chuchu Fan, are leveraging the pre-trained encoder-decoder model T5.
Furthermore, Gandhi is delving into the realm of low-resource languages in the context of speech recognition. In an era where data-driven language models dominate, low-resource languages face challenges due to limited transcribed speech or the absence of a written form.
Her interest in language processing for such languages led her to devise a unique approach. By recognizing patterns in sound sequences, her research group creates a pseudo-vocabulary for these languages, providing labeled data for diverse applications.
Gandhi envisions a future where people can interact with software and devices in their native language or dialect, where voice assistants are enhanced, and where language technology contributes to translation and interpretation. Her work reflects a commitment to inclusivity, ensuring that the benefits of advanced AI technologies extend to diverse linguistic and cultural contexts, the release by the university mentioned.
Comments
Start the conversation
Become a member of New India Abroad to start commenting.
Sign Up Now
Already have an account? Login