Twitter AUD Analysis

Analysis code for examinning the prevalance of AUD-related terms on Twitter throughout the pandemic. Although the data used was non-identifying to begin with, any non-summary data is left out as an added measure of privacy.

File descriptions

keyword_extract.py (year, month, keyword)

Extracts all tweets containing passed keyword for specified month of tweet data.

keyword_remove.py (year, month, keyword)

Removes all isntances of keyword from data collected in keyword_extract script. This is useful for cases such as ALCOHOL minus SANITIZER related tweets.

us_states_extract.py (keyword)

Further subsets the data from keyword_extract to identify the state in which each tweet was posted. Pattern matches the profile_location column with state name strings.

count_states.py (year, month)

Counts the total tweets per identifiable state for a given date. The data generated is generally used as a reference for total Covid-related tweet volume.

Data Generation Process

Given a keyword such as 'BEER' to search for:

• keyword_extract parses all tweets in available data and writes one csv file per month containing tweets with keyterm
• us_states_extract uses these data to do another round of parsing, keeping tweets where the US state is possible to identify via profile_location data
• keyword_remove is used to remove corner cases that are not relevant to the study such as ALCOHOL tweets that are related to SANITIZErs
• count_states is used to collect the total number of tweets produced by identifiable US state tweets