Knowledge Grounded Dialogue Generation
By Deeksha Varshney

Deeksha Varshney will give a talk on Knowledge Grounded Dialogue Generation

Abstract

Dialogue systems have emerged as a vital tool that helps people with their everyday tasks. Advancements in AI have led to the integration of conversational systems, such as personal assistants, into smartphones. This integration provides users with constant companionship and assistance in achieving their goals. Language modelling and natural language generation (NLG) techniques have advanced rapidly, allowing for completely data-driven conversation models that produce natural language responses directly. Building end-to-end dialogue models has been a long interest of natural language research. However, it is a common problem that these dialogue generation models may devolve into boring and repetitive material, resulting in off-topic and pointless responses in conversation assistants. In order to understand and generate coherent dialog, it is necessary to base it on extra information such as aspect, emotion, sentiment, images, structured and unstructured knowledge.

Biography

Deeksha Varshney has completed her Doctor of Philosophy (Ph.D.) in Computer Science and Engineering from the Indian Institute of Technology Patna (IIT Patna), India. Her doctoral research (2018–2023) focused on Enhancing Dialogue Generation Models Via Multi-Source Heterogeneous Information. Previously, she has done her graduation and post-graduation in Computer Science from Banaras Hindu University, Varanasi, India. She is currently a Research Fellow at the Department of Mechanical Engineering at the National University of Singapore (NUS), working under the supervision of Dr. Gianmarco Mengaldo. Her ongoing work involves tackling Entity level hallucinations in Dialogue Systems, Interpretability and Climate NLP.