word embedding
explain word embeddings via automatic rule learning in text classification
In this project, we introduce a novel methodology to find out task-related dimensions within word embeddings. By harnessing the power of automatic rule learning, we effectively extract the critical dimensions relevant to particular tasks.
Rule-based Representation Learner (RRL) is a classifier that automatically learns interpretable non-fuzzy rules for data representation and classification.
In this project, the word embeddings of text are initially fed into the RRL for gender classification, then we can derive the gender-related dimensions (identified as the “red dimensions”) from RRL learned rules. These gender-related dimension values will then be removed for subsequent tasks.
