Abstract
Key Point Analysis (KPA) is one of the most essential tasks in building an Opinion Summarization system, which is capable of generating key points for a collection of arguments toward a particular topic. Furthermore, KPA allows quantifying the coverage of each summary by counting its matched arguments. With the aim of creating high-quality summaries, it is necessary to have an in-depth understanding of each individual argument as well as its universal semantic in a specified context. In this paper, we introduce a promising model, named Matching the Statements (MTS) that incorporates the discussed topic information into arguments/key points comprehension to fully understand their meanings, thus accurately performing ranking and retrieving best-match key points for an input argument. Our approach has achieved the 4th place in Track 1 of the Quantitative Summarization – Key Point Analysis Shared Task by IBM, yielding a competitive performance of 0.8956 (3rd) and 0.9632 (7th) strict and relaxed mean Average Precision, respectively.
Highlights
(2020) posed a question regarding the summarized ability of a small group of key points, and to some extent answered that question on their own by developing baseline models that could produce a concise bullet-like summary for the crowd-contributed arguments
Our approach1 has achieved the 4th place in Track 1 of the Quantitative Summarization – Key Point Analysis Shared Task by IBM, yielding a competitive performance of 0.8956 (3rd) and 0.9632 (7th) strict and relaxed mean Average Precision, respectively
ArgKP-2021 (Bar-Haim et al, 2020), the data set used in the Quantitative Summarization – Key Point Analysis Shared Task, is split into training and development sets with the ratio of 24 : 4. The training set is composed of 5583 arguments and 207 key points while those figures in the development set are 932 and 36
Summary
Reimers and Gurevych (2019) proposed a sentence embedding method via fine-tuning BERT models. A standard approach for key points and arguments on natural language inference (NLI) datasets. More analysis is properly extracting their meaningful se- recent studies in learning sentence representation mantics. Our model stems from recent literatures followed the contrastive learning paradigm and that are based on siamese neural networks (Reimers achieved state-of-the-art performance on numerand Gurevych, 2019; Gao et al, 2021) to measure ous of benchmark tasks
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have