Reviewddit

Inspiration

We found that frequently when browsing Amazon reviews, a lot of them tend to be low-quality reviews, joke reviews, or seemingly botted or AI-generated. We all have turned to searching Reddit for reviews in the past, as it generally has more accurate information, but it's a pain compared to the reviews just being right there.

What it does

Reviewddit, given a product name, will automatically seek out Reddit reviews for a given product, and using those reviews will use AI technology to generate stats about the item, including a generated star count, the number of total reviews, and how many were positive or negative, and the like--all in the convenience of your context menu.

How we built it

We decided to make it a browser extension for ease of access, so we started with that as the front end.

The backend is made up of a few major components. The first is the API itself, which is done with FastAPI. Searching and parsing Reddit posts is done with a Google custom search engine and the Python Reddit API wrapper. Natural language processing is done with spaCy, specifically the eng_spacysentiment model for sentiment analysis. Auth0 is used for authentication, combined with MongoDB, allowing for search history to be shared across machines, and object caching to reduce requests to Google and Reddit's servers.

Challenges we ran into

Integration was the hardest part, with CORS integration always being kind of finicky, and some extension-specific security measures that I had to rewrite bits of code for, like not being able to have onClick in your extension HTML files. Outside of that, we had trouble getting rate-limited by several search libraries we tried before finally setting up a custom search engine.

Accomplishments that we're proud of

We're proud that we managed to get the auth flow up and working, as that was something we struggled with for a while--it seemed like there was always a new error every time.

What's next for Reviewddit

One of the biggest priorities is better parsing of Reddit posts to help with accuracy. Occasionally if you enter a somewhat vague product name, you might get some posts unrelated to the product itself which can throw off the results. Additionally, we plan to summarize all the different comments into one bite-sized blurb so you can get a quick overview of what everybody is saying.

Built With

• ai
• auth0
• css
• fastapi
• firefox
• google-custom-search
• html
• javascript
• mongodb
• natural-language-processing
• praw
• python
• spacy