• 

  Result from Research Radar given a prompt of "online learning"

ResearchRadar

Contributors: Emily Duire-Johnson, Anubha Thapliyal
Hackathon: Developed for the 2024 hackOMSCS Competition

Inspiration

• We met and were inspired by the 2024 OMSCS Conference
• How do we continue to learn (via top researchers or labs)
• Google Scholar is great, but can be overwhelming and enforces its own page ranking algorithm

What it does

Our app simplifies the task of searching for top researchers in a particular area and summarizes their latest work.

How we built it

Our demo + supplemental slide deck provides an overview of the planning process and system design. At a high level, it is a web app that uses Streamlit for the front-end, and Python + Google Scholar + open-source LLMs for the back-end and processing of information.

Challenges we ran into

• We were unable to use OpenAI's API due to limited free tier options
• Our back-up options for open-source LLM summarization did not yeild results of as high of quality as the ChatGPT results
• Google Scholar also imposes a monthly limit of 100 search queries for the free tier, which limited our testing options

Accomplishments and what we learned

We learned or got better at working with:

• Google Scholar API
• Canva
• Open-source models via Hugging Face
• Streamlit
• OpenAI's API
• Prompt engineering

What's next for ResearchRadar

• Scaling it up: increasing the number of authors returned in the result, and the number of papers summarized
• Integrating ChatGPT into the results
• integrating Google Patent API

Built With

• python
• serpapi
• streamlit