Specialized Document Embeddings for Aspect-based Similarity of Research Papers

Abstract

Document embeddings and similarity measures underpin contentbased recommender systems, whereby a document is commonly represented as a single generic embedding. However, similarity computed on single vector representations provides only one perspective on document similarity that ignores which aspects make two documents alike. To address this limitation, aspect-based similarity measures have been developed using document segmentation or pairwise multi-class document classification. While segmentation harms the document coherence, the pairwise classification approach scales poorly to large scale corpora. In this paper, we treat aspect-based similarity as a classical vector similarity problem in aspect-specific embedding spaces. We represent a document not as a single generic embedding but as multiple specialized embeddings. Our approach avoids document segmentation and scales linearly w.r.t. the corpus size. In an empirical study, we use the Papers with Code corpus containing 157, 606 research papers and consider the task, method, and dataset of the respective research papers as their aspects. We compare and analyze three generic document embeddings, six specialized document embeddings and a pairwise classification baseline in the context of research paper recommendations. As generic document embeddings, we consider FastText, SciBERT, and SPECTER. To compute the specialized document embeddings, we compare three alternative methods inspired by retrofitting, fine-tuning, and Siamese networks. In our experiments, Siamese SciBERT achieved the highest scores. Additional analyses indicate an implicit bias of the generic document embeddings towards the dataset aspect and against the method aspect of each research paper. Our approach of aspect-based document embeddings mitigates potential risks arising from implicit biases by making them explicit. This can, for example, be used for more diverse and explainable recommendations.

A domain-adaptive pre-training approach for language bias detection in news

“…Recent media bias studies have progressed from manually generated linguistic features [37,38] to state-of-the-art NLP models yielding internal word representations by unsupervised or supervised training on massive text corpora. The Transformer architecture [48] has shown superior performance in several downstream tasks, such as, text classification [26][27][28], plagiarism detection [50,51], word sense disambiguation [52] and fake news detection on the health domain [49]. However, the use of neural language models, such as BERT [8] and RoBERTa [24] in the media bias domain is still incipient [41,42].…”