Scoring summaries using recurrent neural networks

Stefan Ruseti, Mihai Dascalu, Amy Johnson, Danielle McNamara, Renu Balyan, Kathryn S. McCarthy, Stefan Trausan-Matu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

12 Scopus citations


Summarization enhances comprehension and is considered an effective strategy to promote and enhance learning and deep understanding of texts. However, summarization is seldom implemented by teachers in classrooms because the manual evaluation requires a lot of effort and time. Although the need for automated support is stringent, there are only a few shallow systems available, most of which rely on basic word/n-gram overlaps. In this paper, we introduce a hybrid model that uses state-of-the-art recurrent neural networks and textual complexity indices to score summaries. Our best model achieves over 55% accuracy for a 3-way classification that measures the degree to which the main ideas from the original text are covered by the summary. Our experiments show that the writing style, represented by the textual complexity indices, together with the semantic content grasped within the summary are the best predictors, when combined. To the best of our knowledge, this is the first work of its kind that uses RNNs for scoring and evaluating summaries.

Original languageEnglish (US)
Title of host publicationIntelligent Tutoring Systems - 14th International Conference, ITS 2018, Proceedings
EditorsJulita Vassileva, Roger Nkambou, Roger Azevedo
PublisherSpringer Verlag
Number of pages11
ISBN (Print)9783319914633
StatePublished - 2018
Event14th International Conference on Intelligent Tutoring Systems, ITS 2018 - Montreal, Canada
Duration: Jun 11 2018Jun 15 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10858 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Other14th International Conference on Intelligent Tutoring Systems, ITS 2018


  • Automated summary evaluation
  • Recurrent neural network
  • Semantic models
  • Word embeddings

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Scoring summaries using recurrent neural networks'. Together they form a unique fingerprint.

Cite this