ConvoKit version 2.2 update
Updates to various parts of ConvoKit:
Text processing
Added support for creating Transformers that compute utterance attributes. Also updated support for dependency-parsing text. An example of how this new functionality can be used is found here.
Corpus
Added some functionality to
- support loading and storage of auxiliary data
- handling of vector representations
- organizing users' activities within conversations
- build dataframes containing attributes of various objects
Prompt types
Updated the code used to compute prompt types and phrasing motifs, deprecating the old QuestionTypology module. An example of how the updated code is used can be found here and here.
User Conversation Diversity
Updated code used to compute linguistic divergence.
Other
Added support for pipelining, and some limited support for computing per-utterance attributes.