tear 0.0.2

TExt Analytics for Reconnaissance.

This version of code is refactored by Vedant Mathur

TAR Software Package Repository

Code combining preprocessing, data collection, training and inference to generate automated disaster reports.

Key Files

• tar_main.py - File that consolidates relevant functions to produce a report
• date2template* - Files that do different collectiong/processing of USGIS data to be added to the briefings
• classifiers.py - Calls classifiers (regression, SVN, GAN, CNN) and runs a majority vote to determine the final classification for sentences according to 4 categories (buildings, infrastructure, resilience, other)
• resilience_curve.py - Generates resilience curves, and calculates t0 and t1 (to calculate recovery time for disaster)
• config.ini - Set of parameters to control briefing generation
• data - Folder containing log of earthquakes, tweets and news articles

Usage

Generating a report

To generate a report, run

python -m tear

This will iterate through earthquakes listed in the earthquake log and output a report to the "reports" directory.

Generating a resilience curve

To do this, call the generateResilience function in resilience_curve.py. It takes the following parameters -

• ruptureTime - Reference time to when the earthquake happened (e.g. 2021-02-24 02:05:59)
• twitterFile - CSV with tweets for earthquake
• keywords - keywords to filter tweets by

For example:

generateResilience("2021-02-24 02:05:59", "data/tweets/ArgentinaTweets.csv", ["electricity", "lights"])