article-suggest

Inspiration

We wanted to be able to take more information into account than just the title of an article when suggesting similar articles in order to be able to get semantically similar text.

What it does

When given any text, It looks through a large database of news articles and finds the article most similar with context to the uploaded text.

How we built it

We used BERT, a natural language processing library capable of understanding the context of text and text vectorization to quicly search the databse and find a match.

Challenges we ran into

The database was very large and difficult to parse through the NLP library.

Accomplishments that we're proud of

We increased the NLP parsing speed by using massively parallel processing capabilities.

What we learned

We learned state-of-the-art NLP libraries and techniques.

What's next for article-suggest

Including a chrome extension, direct file upload, and optical character recognition as ways of inputting text to the system. Additionally, increasing database size and scope.

Built With

• artificial-intelligence
• bert
• csv
• javascript
• json
• natural-language-processing
• neural-networks
• node.js
• python
• slurm
• spacy