Tweet analysis for recruiters and applicants

• 

  result of CNN

• 

  Data labels

Inspiration

Over the past few years, Twitter has been increasingly been used by companies to check that the employees they will hire are good employees. In recent years, we have seen people lose their jobs and their respect for foolish things spoken on the internet many years prior. I am developing a tool which would raise awareness amongst job seekers about the importance of keeping their Twitter accounts civil, and also allow recruiters to screen applicants much more easily.

What it does

Allows recruiters and applicants to better analyze their own twitter accounts and see how offensive they may be to the general public, as well as build a picture of personality.

How I built it

Data acquired from github using pandas read_csv Data cleaned in a few steps, and all words undergo lemmatization (spacy) Using CNN to predict if text offensive (Keras)

Challenges I ran into

some versions of scapy require 4 cores but the system provided only had 2. Many features I attempted to use for the lemmatization had been discontinued and I had to find alternatives. Had to perform tasks in batches to avoid going over the memory limit.

Accomplishments that I'm proud of

Fixing the bugs with Scapy. Deploying the CNN with a 90% test accuracy.

What I learned

How to work with IBM Linux0NE, more practical experience with Keras, Matplotlib and Scapy than I previously had. How to deal with dependency issues and cross-platform non-compatibility

What's next for Tweet analysis

If having money is no issue, then getting a twitter api key and launching a web service via flask for easier use by the general public.

Built With

• github
• kaggle
• keras
• matplotlib
• pandas
• python
• scapy