Papers by Kristjan Arumae

4 papers
Towards Annotating and Creating Summary Highlights at Sub-sentence Level (D19-54)

Copied to clipboard

Challenge: Creating summary highlights at the sub-sentence level is desirable because sub-entrances are more concise than whole sentences.
Approach: They propose to annotate summary-worthy sub-sentences and teach classifiers to do the same . they frame task as selecting important sentences and identifying a single most informative textual unit from each sentence.
Outcome: The proposed method reduces the complexity involved in sentence compression by reducing the number of words and phrases deleted.
An Empirical Investigation Towards Efficient Multi-Domain Language Model Pre-training (2020.emnlp-main)

Copied to clipboard

Challenge: Pre-training large language models is a standard practice in the natural language processing community.
Approach: They propose to use elastic weight consolidation to mitigate catastrophic forgetting when pre-trained large language models are evaluated on generic benchmarks.
Outcome: The proposed model achieves state-of-the-art on out-of domain tasks with minimal pre-training . elastic weight consolidation provides best overall scores yielding only a 0.33% drop in performance across seven generic tasks while remaining competitive in bio-medical tasks.
Guiding Extractive Summarization with Question-Answering Rewards (N19-1)

Copied to clipboard

Challenge: a primary challenge faced by extractive summarization systems is the lack of annotated data.
Approach: They propose a supervised extractive summarization system that rewards question-answering by identifying salient sequences of words from a document and highlighting them in the text.
Outcome: The proposed system compares with baselines of strong summarization and human assessors on question-answering.
Reinforced Extractive Summarization with Question-Focused Rewards (P18-3)

Copied to clipboard

Challenge: Existing methods for extractive summarization use human abstracts to create annotations for extraction units.
Approach: They propose a method where abstracts are converted to Cloze-style comprehension questions to generate extractive summarization.
Outcome: The proposed method surpasses state-of-the-art systems on the standard summarization dataset.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations