Skip to main content Skip to navigation

Comuniqa : Exploring Large Language Models for improving speaking skills

Project Overview

The document explores the use of generative AI in education, focusing on the Comuniqa application, which leverages a Large Language Model (LLM) to enhance English speaking skills among non-native speakers in India. It critiques traditional language learning methods and underscores the benefits of AI in delivering scalable, accessible, and personalized educational experiences. Through a study comparing the effectiveness of the Comuniqa app with feedback from human experts, the findings indicate that while AI can provide precise assessments, it falls short in delivering the empathy and cognitive engagement characteristic of human instructors. Consequently, the study advocates for a blended learning approach that combines the strengths of both AI and human feedback to achieve the best educational outcomes. This highlights a significant shift in educational practices, illustrating how generative AI can complement traditional teaching methods to foster improved language acquisition and learning efficiency.

Key Applications

Comuniqa

Context: Improving English speaking skills for non-native speakers in India

Implementation: The application was implemented as a mobile app allowing users to practice speaking and receive instant feedback using LLMs.

Outcomes: Provided scalable and accessible feedback on speaking skills, enhancing learning experiences for those without access to human experts.

Challenges: Lacks human-level cognitive capabilities and empathy, which may impact user trust and engagement.

Implementation Barriers

Technological Limitations

AI systems, including LLMs, lack the emotional intelligence and empathy that human instructors provide.

Proposed Solutions: Integrating human feedback into the AI learning process to enhance user experience and outcomes.

Accessibility

Not all learners may have the same access to technology or the internet, limiting the reach of AI tools.

Proposed Solutions: Developing offline versions of the app or providing access through community resources.

User Trust

Users may doubt the accuracy of AI-generated assessments, affecting their willingness to rely on the technology.

Proposed Solutions: Enhancing transparency in how scores are calculated and providing users with more context on AI's capabilities.

Project Team

Manas Mhasakar

Researcher

Shikhar Sharma

Researcher

Apurv Mehra

Researcher

Utkarsh Venaik

Researcher

Ujjwal Singhal

Researcher

Dhruv Kumar

Researcher

Kashish Mittal

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Manas Mhasakar, Shikhar Sharma, Apurv Mehra, Utkarsh Venaik, Ujjwal Singhal, Dhruv Kumar, Kashish Mittal

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies