Conversation Modeling on Reddit using a Graph-Structured LSTM
This work addresses the problem of conversation modeling for social media platforms like Reddit, offering an incremental improvement over existing methods.
The paper tackles modeling threaded discussions on social media by proposing a graph-structured bidirectional LSTM that captures hierarchical and temporal structures, resulting in improved prediction of comment popularity on Reddit, with analyses showing benefits across early and late stages and in identifying controversial comments.
This paper presents a novel approach for modeling threaded discussions on social media using a graph-structured bidirectional LSTM which represents both hierarchical and temporal conversation structure. In experiments with a task of predicting popularity of comments in Reddit discussions, the proposed model outperforms a node-independent architecture for different sets of input features. Analyses show a benefit to the model over the full course of the discussion, improving detection in both early and late stages. Further, the use of language cues with the bidirectional tree state updates helps with identifying controversial comments.