Tales

product result
model generative ai

Inspiration

Due to the fact that many Indonesians still lack the ability to understand the context of what they read, several problems have arisen in Indonesian education, particularly in the area of research.

We are inspired by ChatGPT from OpenAI, which features a chatbot for uploading PDFs. However, in ChatGPT, users are limited in their requests. This presents an opportunity for us to create our own chatbot for people eager to learn, cultivate efficient research skills, and help people understand the context of their reading material.

What it does

Tales' main focus is to create a chatbot for specific dialogues between learners and the bot, especially for learning and understanding articles, journals, and scientific works.

How we built it

We use machine learning models to generate text. For this project, we rely on LangChain and LLaMA as the main tools for text generation. We also added a feature that can capture user speech and turn it into text using a simple transformer model we trained. Additionally, we use several open-source AI models from Hugging Face, which we retrain to understand the context of documents like PDFs and DOCs.

Challenges we ran into

Ananda Creatix has to train the LLaMA model to produce results that provide answers according to the context given by the user.

Accomplishments that we're proud of

Despite the limited time and knowledge, we still finished the product. These are the things that make us proud of it.

What we learned

We learned how to fine-tune a generative text model and deploy it on a local host using Streamlit.

What's next for Tales

Our next main focus is to develop Tales' as advanced technology, we want to develop kinda, mobile apps form, update new features, add generative image for the user, and various generative text like AI to learn math, history, language, etc.