Forum Representation for Cross-Domain Recommendation

An internet forum is an online site for people to have conversations. It contains threads to hold discussions between users. Recommending appropriate threads to forum users is one of the main goals of an internet forum. To provide positive user experience, cross-domain thread recommendation is required, which can be benefited greatly from the help of forum representations. This research project aims to use two different approaches to create forum representation. One approach is to use the content-based method that utilizes textual data in each subforum and build a topic model to generate subforum embedding vectors. Another approach is the user-based method. It generates subforum embeddings by using a modified skip-gram model, which uses the subforum to predict its user contexts. Lastly, the research project will explore the possibility for a hybrid user-content based approach to further increase the embedding performance.

Faculty Supervisor:

Gerald Penn

Student:

Yizhan Jiang

Partner:

VerticalScope Inc.

Discipline:

Computer science

Sector:

University:

University of Toronto

Program: