Mitigating Data Scarceness through Data Synthesis, Augmentation and Curriculum for Abstractive Summarization
This paper explores three simple data manipulation techniques (synthesis, augmentation, curriculum) for improving abstractive summarization models without the need for any additional data. We introduce a method of data synthesis with paraphrasing, a data augmentation technique with sample mixing, and curriculum learning with two new difficulty metrics based on specificity and abstractiveness. We conduct experiments to show that these three techniques can help improve abstractive summarization across two summarization models and two different small datasets. Furthermore, we show that these techniques can improve performance when applied in isolation and when combined.
Code (0)
등록된 구현이 없습니다.
Tasks
Abstractive Text SummarizationData AugmentationSpecificitySimilar Papers 제목 키워드 기반
GigaPevt: Multimodal Medical Assistant
Building an intelligent and efficient medical assistant is still a challenging AI problem. The major limitation comes from the data modality scarceness, which reduces comprehensive patient perception. This demo paper pre…
Question AnsweringInTeX: Interactive Text-to-texture Synthesis via Unified Depth-aware Inpainting
Text-to-texture synthesis has become a new frontier in 3D content creation thanks to the recent advances in text-to-image models. Existing methods primarily adopt a combination of pretrained depth-aware diffusion and inp…
Texture SynthesisThe TALP-UPC Machine Translation Systems for WMT18 News Shared Translation Task
In this article we describe the TALP-UPC research group participation in the WMT18 news shared translation task for Finnish-English and Estonian-English within the multi-lingual subtrack. All of our primary submissions i…
Machine TranslationTranslationAfrican Wordnet: facilitating language learning in African languages
The development of the African Wordnet (AWN) has reached a stage of maturity where the first steps towards an application can be attempted. The AWN is based on the expand method, and to compensate for the general resourc…
Mitigating Latent Mismatch in cVAE-Based Singing Voice Synthesis via Flow Matching
Singing voice synthesis (SVS) aims to generate natural and expressive singing waveforms from symbolic musical scores. In cVAE-based SVS, however, a mismatch arises because the decoder is trained with latent representatio…