Improving unsupervised neural aspect extraction for online discussions using out-of-domain classification
This work addresses a domain-specific issue for researchers and practitioners using aspect extraction on non-user-generated texts, but it is incremental as it builds on existing methods without modifying core mechanisms.
The paper tackled the problem of low topic coherence in unsupervised neural aspect extraction when applied to newsgroup documents by introducing a sentence filtering approach based on out-of-domain classification, resulting in improved topic coherence compared to models trained on unfiltered texts.
Deep learning architectures based on self-attention have recently achieved and surpassed state of the art results in the task of unsupervised aspect extraction and topic modeling. While models such as neural attention-based aspect extraction (ABAE) have been successfully applied to user-generated texts, they are less coherent when applied to traditional data sources such as news articles and newsgroup documents. In this work, we introduce a simple approach based on sentence filtering in order to improve topical aspects learned from newsgroups-based content without modifying the basic mechanism of ABAE. We train a probabilistic classifier to distinguish between out-of-domain texts (outer dataset) and in-domain texts (target dataset). Then, during data preparation we filter out sentences that have a low probability of being in-domain and train the neural model on the remaining sentences. The positive effect of sentence filtering on topic coherence is demonstrated in comparison to aspect extraction models trained on unfiltered texts.