TY - GEN

T1 - A new corpus of the russian social network news feed paraphrases

T2 - 16th Mexican International Conference on Artificial Intelligence, MICAI 2017

AU - Pronoza, Ekaterina

AU - Yagunova, Elena

AU - Pronoza, Anton

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2018. Copyright: Copyright 2019 Elsevier B.V., All rights reserved.

PY - 2018

Y1 - 2018

N2 - In this paper we present a new Russian paraphrase corpus derived from the news feed of the social network and conduct its primary analysis. Most media agencies post their news reports on their pages in social networks, and the headlines of the messages are often the same as those of the corresponding news articles from the official websites of the agencies. However, sometimes these pairs of headlines differ, and in such cases a headline from the social network can be considered a compression or a paraphrase of the original headline. In other words, such news feed from social networks is a rich resource of textual entailment, and, as it is shown in this paper, various linguistic phenomena, e.g., irony, presupposition and attention attracting markers. We collect the described pairs of headlines and construct the Russian social network news feed paraphrase corpus based on them. We test the paraphrase detection model trained on the other existing Russian paraphrase corpus, ParaPhraser.ru, collected from official news headlines only, against the constructed dataset, and explore its linguistic and pragmatic features.

AB - In this paper we present a new Russian paraphrase corpus derived from the news feed of the social network and conduct its primary analysis. Most media agencies post their news reports on their pages in social networks, and the headlines of the messages are often the same as those of the corresponding news articles from the official websites of the agencies. However, sometimes these pairs of headlines differ, and in such cases a headline from the social network can be considered a compression or a paraphrase of the original headline. In other words, such news feed from social networks is a rich resource of textual entailment, and, as it is shown in this paper, various linguistic phenomena, e.g., irony, presupposition and attention attracting markers. We collect the described pairs of headlines and construct the Russian social network news feed paraphrase corpus based on them. We test the paraphrase detection model trained on the other existing Russian paraphrase corpus, ParaPhraser.ru, collected from official news headlines only, against the constructed dataset, and explore its linguistic and pragmatic features.

KW - Linguistic phenomena

KW - Loose paraphrase

KW - News headlines

KW - Paraphrase corpus

KW - Social network news feed

KW - Text compression

KW - Textual entailment

UR - http://www.scopus.com/inward/record.url?scp=85059937316&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-02840-4_11

DO - 10.1007/978-3-030-02840-4_11

M3 - Conference contribution

AN - SCOPUS:85059937316

SN - 9783030028398

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 133

EP - 145

BT - Advances in Computational Intelligence - 16th Mexican International Conference on Artificial Intelligence, MICAI 2017, Proceedings

A2 - González-Mendoza, Miguel

A2 - Castro, Félix

A2 - Miranda-Jiménez, Sabino

PB - Springer Nature

Y2 - 23 October 2017 through 28 October 2017

ER -