Debanjan Ghosh

Title: Innovations by Debanjan Ghosh

Introduction

Debanjan Ghosh is an accomplished inventor based in Cambridge, MA. He has made significant contributions to the field of machine learning and natural language processing. With a total of three patents to his name, Ghosh is recognized for his innovative approaches to evaluating argumentative writing and text classification.

Latest Patents

One of Ghosh's latest patents is focused on machine learning-based argument component segmentation. This invention involves receiving data that includes a passage of text generated in response to a prompt, which comprises multiple sentences. The passage is tokenized into a plurality of tokens, each corresponding to a different word in the text. A first classification head of an adaptive fine-tuned transformer classifies each token into one of several classes. Additionally, a second classification head classifies each sentence as either including or not including an argument. This patent describes related apparatus, systems, techniques, and articles that enhance the understanding of argumentative structures in text.

Another notable patent by Ghosh is aimed at the automatic evaluation of argumentative writing by young students. This invention provides systems and methods for evaluating argument critique essays written by students in response to prompts. A transformer pre-trained for natural language processing is utilized as a machine learning model, which is fine-tuned with a first training dataset of unannotated essays written by college students. It is then further fine-tuned with a second dataset of annotated essays written by middle school students, where each sentence is marked for the presence of valid critiques. The fine-tuned model classifies each sentence in an essay as either containing a valid critique or not.

Career Highlights

Debanjan Ghosh currently works at Educational Testing Service, where he applies his expertise in machine learning to improve educational assessments. His work focuses on enhancing the evaluation processes for student writing, making it more efficient and effective.

Collaborations

Ghosh collaborates with notable colleagues such as Vikram Ramanarayanan and Matthew Mulholland, contributing to advancements in educational technology and assessment methodologies.

Conclusion

Debanjan Ghosh is a prominent inventor whose work in machine learning and natural language processing is shaping the future of educational assessments. His innovative patents reflect his commitment to improving the evaluation of argumentative writing among students.