Simple English

• 

  Uranus, after a few expansions.

• 

  A simplified page on the University of California, Berkeley.

• 

  An article for Andrej Karpathy, after clicking into his home, and "deep learning."

• 

Inspiration

• Wikipedia has versions of some pages that are "simplified", e.g., https://simple.wikipedia.org/wiki/Association_football
• However, not every page has this; and, it's fun to be able to delve deeper after getting the initial simplified version
• Inspired by explorer.globe.engineer and telescopictext.org, we tried to enable a spelunking-like feature for the summarized article

What it does

• On a Wikipedia page, gives the user the ability to (1) 'translate' the content into Simple English, even if that translation didn't exist and (2) click into key words or phrases identified by the LLM, for exploring rabbitholes

How we built it

• The client is an extension written in TypeScript (Plasmo)
• The server is written in Python (Flask)

Challenges we ran into

• Working correctly with assistants; prompt engineering

Accomplishments that we're proud of

• Managing to get typed output with keyed words
• Being able to expand on those words in-text

What we learned

• Plasmo is great for building extensions
• gpt4-o is powerful! Assistants are cool!

What's next for Simple English

• Use the existing simple translation, if it exists
• Render more Wikipedia-like summaries (e.g., preserve links, tables, images)

Built With

• chrome
• javascript
• openai
• plasmo
• python
• typescript