Sentence Embeddings as an intermediate target in end-to-end summarisation

Maciej Zembrzuski, Saad Mahamood · 2025

Read Sentence Embeddings as an intermediate target in end-to-end summarisation by Maciej Zembrzuski, Saad Mahamood (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-03481.txt.

Description

Maciej Zembrzuski and Saad Mahamood summarize piles of hotel reviews by predicting what the summary's own sentences should mean, in this 2025 paper. Neural summarizers choke on very long inputs. Their fix grafts a picking step onto a writing model, guided by ready-made sentence vectors.

The system first scores source sentences with those vectors, then writes the summary. It also learns to predict the vector of the target summary itself. On loosely matched source and target texts, that extra target beat the usual sentence-picking scores. The result is better summaries of large review sets with less strain.

Catalog record

CREATORMaciej Zembrzuski, Saad Mahamood
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-03481

Browse the full library · About the Tower of Babel