https://explici.tech

predictions visual
predictions 85 percent
Fun times (NLP keywords)
Fun times (NLP keywords) part 2
big data 1
big data 2
homepage

Inspiration

Ever find yourself occupied reading lengthy Terms & Conditions on websites, or worse agreeing to those conditions without reading them through?! Is there a better way?

Our project aims to solve this mundane, everyday nuisance through the use of intuitive (and eventually standardized) symbols.

What it does

Essentially, our application consists of a chrome extension (used in lieu of embedded HTML snippets, which would be used in practice however we don't have appropriate permissions from any websites to embed our code yet), which is activated on websites that are heuristically determined to contain either a privacy policy or terms of service. Then the extension embeds a small modal over the left corner of the website summarizing the lengthy legal document so that readers can understand the gist of the jargon without going through 5 years of law school.

This summarization works through Google's AutoML service, which is fed in line by line every sentence of the legal document. For each line, the NLP program will classify it according to our internal classifications, such as “Third-party data sharing”, “Irrelevant Jargon” etc.

For the purpose of our MVP demo, our extension combines these tags and gives the website a description of 3 categories which we have determined to be the most important to users; What data is the website collecting? Who has access to this data? If a user can create or upload intellectual property on a given website, who owns it?

How we built it

ReactJS, GrommetUI, GCP's AutoML / Entity & Classification models, keyword-search, pre-scraped data for NLP, mock embeddable HTML snippets (chrome extension used in lieu), Domain.com, GitHub Pages,

Challenges we ran into

We weren't able to find enough data to train the AutoML model, and given the lack of labeled data available on platforms like Kaggle for the purposes of our project, we, unfortunately, had to rely on some biased labels (i.e. hand-labeled tags).

Also, since we can't currently reach out to website owners such as Google/Facebook (as this hackathon takes place outside of normal business hours), we did not have any websites try out our embeddable button feature, and hence we had to show our demo using a Chrome extension instead.

Accomplishments that we're proud of

Making a boring task like reading jargon much more intuitive and much less painful.

Website deployed.

Pulled an all-nighter.

What we learned

Among the things we explored, we were most excited to learn about the different Google Cloud APIs and building embeddable HTML snippets as well as Chrome extensions!

What's next for https://explici.tech

We'd like to onboard companies to make the switch to a more intuitive and clear version of their policies rather than opting to use complex, inconvenient, and mundane jargon.