C.31 Forward Flow

1. Feature Name

C.31 Forward Flow/Lexical Cohesion

2. Literature Source (Serial Number, link)

[Paper] (https://psycnet-apa-org.proxy.library.upenn.edu/fulltext/2019-03039-001.pdf)

3. Description of how the feature is computed (In Layman’s terms)

Captures "stream of consciousness", i.e. the evolution of thoughts/chats over time. Overall, conversations that display higher levels of cohesion (determined via cosine similarity) amongst their chats have increased forward flow. At the chat level, this feature represents the cosine similarity of the current chat to the average of the all the previous chats.

Calculation:

1. Determine average cosine similarity the current chat and the "average chat" from all previous chats.
2. Compute the new "average chat", now incorporating the current chat, before moving on to the next chat.

4. Algorithms used (KNN, Logistic Regression etc.)

N/A

5. ML Inputs/Features

N/A

6. Statistical concepts used

N/A

7. Pages of the literature to be referred to for details

[Paper] (https://psycnet-apa-org.proxy.library.upenn.edu/fulltext/2019-03039-001.pdf)

8. Any tweaks/changes/adaptions made from the original source

For each chat, source computed the average of the distances between a given chat and EACH previous chat, as opposed to one overarching "average chat." Made this modification with runtime in mind.