Unsupervised Extractive Dialogue Summarization in Hyperdimensional Space
This addresses the problem of efficient and accurate dialogue summarization for NLP applications, though it appears incremental as it combines existing concepts in a novel way.
The paper tackles unsupervised extractive dialogue summarization by introducing HyperSum, a framework that uses high-dimensional random vectors to create efficient sentence embeddings. The method outperforms state-of-the-art summarizers in accuracy and faithfulness while being 10 to 100 times faster.
We present HyperSum, an extractive summarization framework that captures both the efficiency of traditional lexical summarization and the accuracy of contemporary neural approaches. HyperSum exploits the pseudo-orthogonality that emerges when randomly initializing vectors at extremely high dimensions ("blessing of dimensionality") to construct representative and efficient sentence embeddings. Simply clustering the obtained embeddings and extracting their medoids yields competitive summaries. HyperSum often outperforms state-of-the-art summarizers -- in terms of both summary accuracy and faithfulness -- while being 10 to 100 times faster. We open-source HyperSum as a strong baseline for unsupervised extractive summarization.