As mentioned in the paper "On Identifying Disaster-Related Tweets: Matching-Based or Learning-Based?" by Hien To et al., we apply machine learning to retrieve disaster tweets and compare it with our matching based algorithm. We find that extracting tweets using a list of most effective keywords discovered in the wild is not sufficient. In addition to it, capturing the most populous hashtags from the retrieved tweets and extracting additional tweets from the hashtags improves the retrieval drastically. We also demonstrate that learning based method works well when used with effective feature extraction techniques.

Report: Download Here

Venkat, N., Gupta, A., Sharma, S. (2018). Comparing Learning-Based and Matching-Based Methods for Identifying Disaster Related Tweets. DOI: 10.13140/RG.2.2.28614.32325

This is a python3 project. Install the required packages given in requirements.txt. This can be done using pip3.

python3 -m pip install -r requirements.txt

Once all requirements are installed, the results can be seen by executing the main.py:

python3 main.py

Naveen Venkat (2015A7PS0078P) Aman Gupta (2015B3A70640P) Sanskriti Sharma (2015B3A70553P)

BITS Pilani