Summarization Beyond News: The Automatically Acquired Fandom Corpora

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 6700–6708 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 6700 Summarization Beyond News: The Automatically Acquired Fandom Corpora Benjamin H¨attasch, Nadja Geisler, Christian M. Meyer, Carsten Binnig TU Darmstadt {benjamin.haettasch, nadja.geisler, carsten.binnig}@cs.tu-darmstadt.de Abstract Large state-of-the-art corpora for training neural networks to create abstractive summaries are mostly limited to the news genre, as it is expensive to acquire human-written summaries for other types of text at a large scale. In this paper, we present a novel automatic corpus construction approach to tackle this issue as well as three new large open-licensed summarization corpora based on our approach that can be used for training abstractive summarization models. Our constructed corpora contain fictional narratives, descriptive texts, and summaries about movies, television, and book series from different domains. All sources use a creative commons (CC) license, hence we can provide the corpora for download. In addition, we also provide a ready-to-use framework that implements our automatic construction approach to create custom corpora with desired parameters like the length of the target summary and the number of source documents from which to create the summary. The main idea behind our automatic construction approach is to use existing large text collections (e.g., thematic wikis) and automatically classify whether the texts can be used as (query-focused) multi-document summaries and align them with potential source texts. As a final contribution, we show the usefulness of our automatic construction approach by running state-of-the-art summarizers on the corpora and through a manual evaluation with human annotators. Keywords: corpus construction, multi-document summarization, query-focused summarization 1. Introduction Motivation: Abstractive summaries help users to under- stand new text collections efficiently but writing these sum- maries is time-consuming and complex. Automatic sum- marization aims to eliminate or reduce the manual process. Since the advent of deep learning, automatic summariza- tion methods require huge corpora to properly train neu- ral architectures. The CNN/DailyMail dataset (Hermann et al., 2015), Gigaword (Napoles et al., 2012), and the New York Times (NYT) corpus (Paulus et al., 2018) are cur- rently the largest summarization corpora. They have been used successfully for training a wide range of neural archi- tectures, including recurrent neural networks (Nallapati et al., 2017), pointer-generator networks (See et al., 2017), attention mechanisms (Paulus et al., 2018; Gehrmann et al., 2018), and approaches based on reinforcement learning (Narayan et al., 2018; Gao et al., 2018). All of these corpora are limited to the news genre where texts are typically too short to qualify as general-purpose summaries. For example, CNN/DailyMail provides only bullet-point summaries, Gigaword contains headlines as the summary of an article’s first sentence, and the NYT cor- pus pairs news articles with their abstracts. To break new ground in the automatic summarization of other genres, we require new corpora that can cover other text genres and summary types on the one side but are large enough to train neural networks on the other side. However, construct- ing summarization …

Content truncated to 500 words