haikubot

Live on Twitter

What I Do:

Give me the name of an artist, and I'll try to come up with a random haiku using their lyrics.

Algorithm:

• Search genius.com for the artist
• Scrape top songs links from that artist
• Scrape song pages for lyrics
• Process lyrics line by line:
  • Inclusion criteria: 5 or 7 syllables
  • Exclusion criteria: verse header
• Generate a random haiku

Problems:

• Natural language processing is difficult, and the counting of syllables is very hard given the diversity of words in song lyrics. I used a combination of NLTK's corpus dictionary and textstat as a backup. Sometimes the syllable count is off.
• Does not store the lyrics locally, so is slowed down by the HTML requests
• Only handles requests via DMs right now, because I don't want the bot to be banned for returing haikus with profanity or other inappropriate words publically.

How Do I Run This?

If you want to run this yourself locally, you need your Twitter OAuth keys in secrets.py, then run python bot.py

Dependencies:

• tweepy for Twitter API handling
• textstat and NLTK for syllable processing
• BeautifulSoup4 for HTML parsing
• urllib2 for web requests