Team02-Enhanced Text Communication System using Sentiment based Prompting for End-User

Indian Institute of Science Summer 2025 DA225o Submission6 Authors

Keywords: sentimentclassification, chatapplication, textprompting, textsuggestions, transformers, bert, llm, nlp, mlp

TL;DR: A sentiment-driven prompting framework that analyzes emotional context in real-time to generate more empathetic and contextually appropriate responses in conversational applications

Abstract:

> Today's chat applications are built around the same building blocks which were used when they were emerging - one-on-one chats, groups, multiple modalities (text, audio, video, emojis, etc.), and any new chat application has pretty much the same functionality.
>
> On the other hand, more and more people are communicating over text than any other form of communication. But text has one problem - the tone (or body) of the conversation is not always apparent, and may require further clarification. It may eventually even lead to misunderstandings while communicating. This happens so often that multiple modalities are then equipped in order to clarify, elaborate or communicate the same *sentiment*.
> This paper presents an application to enhance our texting experience, and add another dimension to communication through this medium. The chat application presented in this paper does the following:

> 1. Understand the ongoing conversation in a text (chat) session in real-time
> 2. Provide live analysis about the *sentiment* of parts of the conversation or last (immediate) text
> 3. Based on the sentiment understanding and context of the conversation, provide suggestions/prompts to the user to respond appropriately, or tailor it before sending

> This application aims to solidify communication through text seamlessly by improving user understanding and judgment, while retaining the existing methods.

Submission Number: 6