Master's Thesis

Abstractive Multi-document Summarization using Topical Simplicial Curves

João Rui Serote Nunes Martins Cruz — 2021

Key information

Authors:

João Rui Serote Nunes Martins Cruz (João Rui Serote Nunes Martins Cruz)

Supervisors:

David Manuel Martins de Matos (David Manuel Martins de Matos); Ricardo Daniel Santos Faro Marques Ribeiro

Published in

January 20, 2021

Abstract

We explore the effectiveness of simplicial curves, a word-representation method that is context- sensitive, motivated by its intrinsic mathematical properties (e.g., differentiation and ease of combining representations), in the multi-document summarization task. To this effect, we adapt the simplicial curves framework to use dense word-representations as its basis matrix representation, and we develop an algebra over objects in the simplex. We use the DUC 2006 and DUC 2007 multi-document summarization corpora. The generated summaries are compared with the reference summaries using the ROUGE-1, ROUGE-2 and ROUGE-L evaluation metrics. Compared to the ROUGE-1 score of 0.29 of the simplest chosen baseline, our method achieves a ROUGE-1 score of 0.04, falling below our expectations. We conclude with an exploration of the obtained results and suggest other applications of the simplicial curves method.

Publication details

Authors in the community:

Supervisors of this institution:

Fields of Science and Technology (FOS)

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

Publication language (ISO code)

eng - English

Rights type:

Embargo lifted

Date available:

December 11, 2021

Institution name

Instituto Superior Técnico