Event-Driven News Stream Clustering using Entity-Aware Contextual Embeddings
This addresses the problem of organizing real-time news streams for applications like media monitoring, though it appears incremental as it builds on existing streaming K-means with transformer enhancements.
The paper tackles online news stream clustering by proposing a variant of non-parametric streaming K-means that uses entity-aware contextual embeddings and a neural classifier for clustering decisions. The method achieves a new state-of-the-art on a standard English document dataset.
We propose a method for online news stream clustering that is a variant of the non-parametric streaming K-means algorithm. Our model uses a combination of sparse and dense document representations, aggregates document-cluster similarity along these multiple representations and makes the clustering decision using a neural classifier. The weighted document-cluster similarity model is learned using a novel adaptation of the triplet loss into a linear classification objective. We show that the use of a suitable fine-tuning objective and external knowledge in pre-trained transformer models yields significant improvements in the effectiveness of contextual embeddings for clustering. Our model achieves a new state-of-the-art on a standard stream clustering dataset of English documents.