Welcome to our submission for the Samsung AI Hackathon! Below you'll find details about our project, the inspiration behind it, and the components we've developed to address the challenges faced by users in the Indian market.
The inspiration behind our project stems from a deep understanding and appreciation of the linguistic and cultural diversity that characterizes India. With approximately 22 Indic languages spoken across the nation, we recognized the need for an AI-based Conversational Assistant tailored specifically to cater to this diverse linguistic landscape.
Our journey began with the realization that a large portion of India's population communicates using languages that are often overlooked by mainstream technology solutions. This realization ignited a passion within us to bridge this gap and create a transformative solution that not only understands but celebrates India's linguistic diversity.
By developing an AI Assistant capable of seamless communication in multiple languages, maintaining context, and authenticity, we aim to set new standards in inclusivity. Leveraging state-of-the-art language models such as Mistral Instruct 7b v0.2 and llava-1.5-7b-hf, our solution not only meets current needs but also positions us favorably for future implications of multimodal interactions.
This notebook contains the implementation of our AI Conversational Assistant for India. Here's a brief overview:
- Mission: Building an AI Conversational Assistant tailored for the Indian market.
- Languages: Supporting 22 Indic languages for seamless communication.
- Technology: Utilizing Mistral Instruct 7b v0.2 and llava-1.5-7b-hf for advanced language processing.
- Functionality: Seamlessly handles textual queries without any visual tasks and incorporates cutting-edge multimodal capabilities.
In this notebook, we've developed a Custom Samsung Chatbot Recommender System. Key highlights include:
- Intent Classifier: Utilizing a custom-created dataset from Samsung prompts, open-sourced on Kaggle.
- Dataset: Comprising 10 different classes with a total of 967 diverse datapoints, created using GPT 3.5.
- Goal: Enhancing ease-of-use among users when interacting with Samsung IoT Devices.
To explore our project and its functionalities, follow these steps:
- Clone the repository to your local machine.
- Open the respective notebooks in a Jupyter environment.
- Follow the instructions within each notebook to execute and explore the code.
Arav Jain: samsung-prism-round-2.ipynb
Vatsal Jha: samsungrecommender.ipynb
Ayushman Kar: Led backend implementation, database integration, and deployment.
Dilshad Sukheswala: Managed backend development and user interface design.
This project is a submission for the Samsung AI Hackathon and is developed by StraightOuttaVellore