TweetReplyAnalyzer

Get counts of keywords that appear in the replies of a specific tweet.

Made on Python 2.7, but can easily be modified to work on Python 3.x

How to Use

1. Copy, clone, or download this repo.

2. Install packages

   1. pip install tweepy
   2. pip install nltk
   3. python -m nltk.downloader stopwords

3. On lines 26 through 31 of main.py, replace TwitterHandle and 000000000 with what appears in the URL of the tweet you want to analyze https://twitter.com/TwitterHandle/status/StatusID

4. Define your option buckets on lines 35 through 37 of main.py. Add as many as you need.

5. For each of your option buckets, add synonyms to the aliases dictionary on lines 41 through 56 of main.py. Add as many as you need, but make sure each synonym is a single, lowercase word with no punctuation (as these will be compared against filtered, split text).

6. Copy secret.sample.py and add your keys, tokens, and secrets according to the Twitter App you create. See Twitter Developer Docs for more info.

7. Run python start.py which will load your environment variables in secret.py and then run main.py. Check the output directory for results.

How it works

Generally, it collects all (up to TweetConfig.max_results most recent) replies to the given TweetConfig.tweet_id. For each reply, it removes punctuation, sets to lowercase, separates words, replaces synonyms with one of the bucket names (ex: one and 1 become Option_1), and removes duplicates within the same reply.

Finally it outputs the resulting counts in descending order, along with buckets for each voting option. The buckets contain every unmodified reply that was counted towards the voting option, so that it can be manually reviewed and results adjusted for errors. There is also a built-in bucket (named __Unmatched__, which you can change if desired) that collects any replies that fail to match any of the bucket synonyms (this is useful for catching typos).

Each bucket is output as its own output/buckets/Bucket_Name.json file. A full list of the reply texts is output in output/replies.txt, which is useful for comparing results from different days via diff. Newest results are on top. The top TweetConfig.max_top_count words are listed in output/count.txt