This document overviews the strategy, effort and aftermath of the MultiLing 2013 multilingual summarization data collection. We describe how the Data Contributors of MultiLing collected and generated a multilingual multi-document summarization corpus on 10 different languages: Arabic, Chinese, Czech, English, French, Greek, Hebrew, Hindi, Romanian and Spanish. We discuss the rationale behind the main decisions of the collection, the methodology used to generate the multilingual corpus, as well as challenges and problems faced per language. This paper overviews the work on Czech, Hebrew and Spanish languages.

Towards Language-Independent News Summarization

Multi-document multilingual summarization corpus preparation, Part 2: Czech, Hebrew and Spanish

The paper describes our participation in the Multi-document summarization task of Multiling-2013. The community initiative was born as a pilot task for the Text Analysis Conference in 2011. This year the corpus was extended by new three languages and another five topics, covering in total 15 topics in 10 languages. Our summariser is based on latent semantic analysis and it is in principle language independent. Its results on the Multiling-2011 corpus were promising. The generated summaries were ranked first in several languages based on various metrics. The summariser with minor changes was run on the updated 2013 corpus. Although we do not have the manual evaluation results yet the ROUGE-2 score indicates good results again. The summariser produced best summaries in 6 from 10 considered languages according to the ROUGE-2 metric.

background

The UWB Summariser at Multiling-2013

In this book chapter, we discuss several pertinent aspects of an automatic system that generates summaries in multiple languages for sets of topic-related news articles (multilingual multi-document summarisation), gathered by news aggregation systems. The discussion follows a framework based on Latent Semantic Analysis (LSA) because LSA was shown to be a high-performing method across many different languages. Starting from a sentence-extractive approach we show how domain-specific aspects can be used and how a compression and paraphrasing method can be plugged in. We also discuss the challenging problem of summarisation evaluation in different languages. In particular, we describe two approaches: the first uses a parallel corpus and the second statistical machine translation.

background or methods

/pdf/aspects-of-multilingual-news-summarisation-1qzuueomjw.pdf

Aspects of Multilingual News Summarisation

We are presenting a method for the evaluation of multilingual multi-document summarisation that allows saving precious annotation time and that makes the evaluation results across languages directly comparable. The approach is based on the manual selection of the most important sentences in a cluster of documents from a sentence-aligned parallel corpus, and by projecting the sentence selection to various target languages. We also present two ways of exploiting inter-annotator agreement levels, apply them both to a baseline sentence extraction summariser in seven languages, and discuss the result differences between the two evaluation versions, as well as a preliminary analysis between languages. The same method can in principle be used to evaluate single-document summarisers or information extraction tools.

Towards Language-Independent News Summarization

Citations

Multi-document multilingual summarization corpus preparation, Part 2: Czech, Hebrew and Spanish

The UWB Summariser at Multiling-2013

Aspects of Multilingual News Summarisation

References

Using parallel corpora for multilingual (multi-document) summarisation evaluation

Related Papers (5)

Language Independent Summarization Approaches

Abstractive Summarization of Russian News Learning on Quality Media

Information extraction by an abstractive text summarization for an Indian regional language

NewsGist: a multilingual statistical news summarizer

The NLP Techniques for Automatic Multi-article News Summarization Based on Abstract Meaning Representation