.ooooo.  ooo. .oo.  .oo.   ooo. .oo.  .oo.    .oooo.
d88' `88b `888P"Y88bP"Y88b  `888P"Y88bP"Y88b  `P  )88b
888ooo888  888   888   888   888   888   888   .oP"888
888    .,  888   888   888   888   888   888  d8(  888
`Y8bod8P' o888o o888o o888o o888o o888o o888o `Y888""8o

        ELECTRONIC MODEL of MAPPED ASSOCIATIONS

      Written by Ellie Cochran & Alexander Howard

Emma is a computer program that generates rough concepts of associations by reading input. She uses these associations to generate a reply (consequently, Emma is not a run-of-the-mill Markov bot. She's much more interesting than that~). She was created by Digital Media student, programmer, & computer artist Ellie Cochran with help from Computer Science & Mathematics student Alexander Howard.

Talk to Emma

You can talk to Emma using Tumblr Asks at @emma@botsin.space.

How Does Emma Work?

1. Emma reads a message from Mastodon

• The message is first prepared to be read, which involves screening for banned words or vulgar language, as well as expanding common abbreviations and adding punctuation if none exists
• The positive or negative sentiment of the message is recorded and used along with other sentiments from other messages to calculate Emma's mood
• The message Emma has chosen to respond to is parsed using pattern.en
  • This gives us all kinds of information about the language used, including lemata, chunks, chunk relations, and parts of speech
• We look through the new metadata-tagged sentences to fix up any remaining things that could hinder Emma's understanding, such as evaluating pronouns and posessive references
• Emma reads through the sentences and records any new words she finds, along with their parts of speech and a (currently unused) affinity score
• Emma uses a pattern matching strategy to find key phrase structures in the message that indicate a relationship between two objects, and records the objects and their relationship
  • This is what makes up Emma's Association Model
  • Current association types are IS-A, HAS-PROPERTY, HAS-ABILITY-TO, HAS, and HAS-OBJECT
    • As a side note, HAS-OBJECT refers not to a noun that another noun has in its posession, but to an object of a sentence that is valid for a noun to use
  • Associations are also weighted

2. Emma replies to the message and posts the response to Mastodon

• Emma looks for important words in the message to figure out what it's about
• Emma decides the number and types of sentences to generate
  • This is based on the types of associations that exist for a given object. For example, if plenty of IS-A associations exist, a DECLARATIVE sentence could be generated. If no or few associations exist, an INTERROGATIVE sentence could be generated.
  • This is influenced by Emma's mood
• Emma creates rough outlines of the sentence, which include placeholders for articles, conjunctions, punctuation, and words that can change based on noun plurality
• The placeholders are conjugated based on their surrounding words, and capitalization and punctuation are added
• The reply is posted to Mastodon

Contact the Developers

Ellie and Alex are on social media! Ask us about Emma! Ellie is @deersyrup@yiff.life on Mastodon, and Alex is @ale303sh on Twitter.

Special Thanks

• Omri Barak
• Alexander Lozada