About a week or so back, a few Northeastern Dorms caught fire, and Northeastern has an emergency message notification service that basically sends alerts related to any emergencies near Northeastern University.
Taking inspiration from this idea, we came up with TwiNotify.
TwiNotify is a messaging service that broadcasts trending alerts.
To accomplish this, we performed the following steps:
Classification of Tweets
Selection of Tweets
In order to get tweets, we used the Twitter API, and also imported data from a publicly available dataset.
Classification of Tweets:
We classified tweets as Relevant and Not Relevant using Logistic Regression.
Selection of Tweets:
We select the tweets based on their relevance using a custom algorithm that is explained separately.
We convert the list of words with arranged by their weights over all the tweets in the test data, into a dictionary. The key is the word itself, while the rank is the rank of the word in the input list. We then iterate over list of words in all the tweets and pick out the tweets with the greatest score. Score for each of the tweets is the sum of all the words found in the tweet.
We use Twilio’s SMS API to then broadcast messages that we have selected.
The scope of the project is currently limited to disaster-related information, but later can be broadened to other topics of interest.