Paraphrase generation and evaluation: a view from the trenches
Published in Proceedings of SPIE: Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 2019, 2019
In this paper we evaluate the current state of the art in natural language paraphrase generation using deep learning methods. The focus is put on the entire modeling pipeline from data gathering up to model evaluation. Specifically, we list the publicly available datasets suitable for this task, assess their quality and discuss procedures connected with data preparation and model training. Finally, we discuss problems related to the currently used evaluation approaches.
Recommended citation: Franus, W., Twardowski, B., Zawistowski, P., & Nowak, R. M. (2019). "Paraphrase generation and evaluation: a view from the trenches." W R. Romaniuk & M. G. Linczuk (Redaktorzy), Proceedings of SPIE: Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 2019.
Download Paper