liaoandi
commented
6 years ago Thanks for the review and advice! The NLP part might become the focus of my thesis project :)
Open LeosonH opened 6 years ago
liaoandi
commented
6 years ago Thanks for the review and advice! The NLP part might become the focus of my thesis project :)
I appreciate the well thought out methodology in this project, especially with the comparison between the various iterations of models used, and the eventual optimization of your classification accuracy in the final model. I was able to clearly understand the steps involved, because a lot of effort was put into explaining the data preparation process and the models deployed – for example, the differences between the baseline, weighted, and gridsearch models.
The poster is easy on the eyes with a light and consistent color palette. The graphs are naturally emphasized without feeling out of place, and are of an appropriate size such that no one has to squint at the labels and values. The layout is balanced with three distinct and equally sized columns, clearly demarcating where each section begins and ends.
The research question is interesting and picks up on the nuanced distinction between “expert users” and “clever/popular users” on Zhihu. As a parallel follow-up, I would suggest performing an study in a lab setting where participants’ perception of a user’s’ “expertness” are studied for their correlatedness with how the model classifies these users. More text methods could also have been considered for the task of detecting user expertise, such as topic modelling or natural language processing techniques.