mail2vec

Landing Page
Application Doing Work

Inspiration

Email has often been categorized manually, ex. the common Gmail folders. Traditional email search is very narrow and limited, usually boiling down to term matching. Semantic analysis and automated categorization could be implemented by the use of neural networks to automatically sort and categorize using vector analysis.

What it does

Lets you select a term and determines which emails, utilizing a massive vector representation of common terms, should be categorized under that term.

How we built it

word2vec python library running on a python server and nodeJS server as middleware between our web application and the python server.

Challenges we ran into

Using Gmail API on a Node.js server was a pain, so we had to port it over to the client side and restructure our entire application flow upon this.

Accomplishments that we're proud of

Everything, from running a python server to successfully using Google's Gmail API for Javascript.

What we learned

Even with extensive planning, there were a lot of unexpected challenges with the logistics and problems with API calls.

What's next for mail2vec

For mail2vec specifically, we want to create vector models from runtime that incorporates email text and expand the vocabulary to words that may be specific to one's email. In addition, mail2vec can easily be applied beyond emails such as large databases of any form of text. There is an untapped potential of NLP commercial based applications and vector-based NLP such as word2vec can fill this niche.