ParaCotta: Synthetic Multilingual Paraphrase Corpora from the Most Diverse Translation Sample Pair

Alham Fikri Aji; Tirana Noor Fatyanosa; Radityo Eko Prasojo; Philip Arthur; Suci Fitriany; Salma Qonitah; Nadhifa Zulfa; Tomi Santoso; Mahendra Data

ParaCotta: Synthetic Multilingual Paraphrase Corpora from the Most Diverse Translation Sample Pair

Alham Fikri Aji^*
, Tirana Noor Fatyanosa^*
, Radityo Eko Prasojo^*
, Philip Arthur
, Suci Fitriany^*
, Salma Qonitah^*
, Nadhifa Zulfa^*
, Tomi Santoso^*
, Mahendra Data

^*Corresponding author for this work

Faculty of Computer Science

Research output: Contribution to conference › Paper › peer-review

5 Citations (Scopus)

Abstract

We release our synthetic parallel paraphrase corpus across 17 languages: Arabic, Catalan, Czech, German, English, Spanish, Estonian, French, Hindi, Indonesian, Italian, Dutch, Romanian, Russian, Swedish, Vietnamese, and Chinese. Our method relies only on monolingual data and a neural machine translation system to generate paraphrases, hence simple to apply. We generate multiple translation samples using beam search and choose the most lexically diverse pair according to their sentence BLEU. We compare our generated corpus with the ParaBank2. According to our evaluation, our synthetic paraphrase pairs are semantically similar and lexically diverse.

Original language	English
Pages	666-675
Number of pages	10
Publication status	Published - 2021
Event	35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021 - Shanghai, China Duration: 5 Nov 2021 → 7 Nov 2021

Conference

Conference	35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021
Country/Territory	China
City	Shanghai
Period	5/11/21 → 7/11/21

Cite this

@conference{09a07d3019cc4af3a38874d2116d520f,

title = "ParaCotta: Synthetic Multilingual Paraphrase Corpora from the Most Diverse Translation Sample Pair",

abstract = "We release our synthetic parallel paraphrase corpus across 17 languages: Arabic, Catalan, Czech, German, English, Spanish, Estonian, French, Hindi, Indonesian, Italian, Dutch, Romanian, Russian, Swedish, Vietnamese, and Chinese. Our method relies only on monolingual data and a neural machine translation system to generate paraphrases, hence simple to apply. We generate multiple translation samples using beam search and choose the most lexically diverse pair according to their sentence BLEU. We compare our generated corpus with the ParaBank2. According to our evaluation, our synthetic paraphrase pairs are semantically similar and lexically diverse.",

author = "Aji, \{Alham Fikri\} and Fatyanosa, \{Tirana Noor\} and Prasojo, \{Radityo Eko\} and Philip Arthur and Suci Fitriany and Salma Qonitah and Nadhifa Zulfa and Tomi Santoso and Mahendra Data",

note = "Publisher Copyright: {\textcopyright} Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021.; 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021 ; Conference date: 05-11-2021 Through 07-11-2021",

year = "2021",

language = "English",

pages = "666--675",

}

ParaCotta: Synthetic Multilingual Paraphrase Corpora from the Most Diverse Translation Sample Pair. / Aji, Alham Fikri; Fatyanosa, Tirana Noor; Prasojo, Radityo Eko et al.
2021. 666-675 Paper presented at 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021, Shanghai, China.

Research output: Contribution to conference › Paper › peer-review

TY - CONF

T1 - ParaCotta

T2 - 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021

AU - Aji, Alham Fikri

AU - Fatyanosa, Tirana Noor

AU - Prasojo, Radityo Eko

AU - Arthur, Philip

AU - Fitriany, Suci

AU - Qonitah, Salma

AU - Zulfa, Nadhifa

AU - Santoso, Tomi

AU - Data, Mahendra

N1 - Publisher Copyright: © Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation, PACLIC 2021.

PY - 2021

Y1 - 2021

N2 - We release our synthetic parallel paraphrase corpus across 17 languages: Arabic, Catalan, Czech, German, English, Spanish, Estonian, French, Hindi, Indonesian, Italian, Dutch, Romanian, Russian, Swedish, Vietnamese, and Chinese. Our method relies only on monolingual data and a neural machine translation system to generate paraphrases, hence simple to apply. We generate multiple translation samples using beam search and choose the most lexically diverse pair according to their sentence BLEU. We compare our generated corpus with the ParaBank2. According to our evaluation, our synthetic paraphrase pairs are semantically similar and lexically diverse.

AB - We release our synthetic parallel paraphrase corpus across 17 languages: Arabic, Catalan, Czech, German, English, Spanish, Estonian, French, Hindi, Indonesian, Italian, Dutch, Romanian, Russian, Swedish, Vietnamese, and Chinese. Our method relies only on monolingual data and a neural machine translation system to generate paraphrases, hence simple to apply. We generate multiple translation samples using beam search and choose the most lexically diverse pair according to their sentence BLEU. We compare our generated corpus with the ParaBank2. According to our evaluation, our synthetic paraphrase pairs are semantically similar and lexically diverse.

UR - https://www.scopus.com/pages/publications/85127479373

M3 - Paper

AN - SCOPUS:85127479373

SP - 666

EP - 675

Y2 - 5 November 2021 through 7 November 2021

ER -

ParaCotta: Synthetic Multilingual Paraphrase Corpora from the Most Diverse Translation Sample Pair

Abstract

Conference

Other files and links

Fingerprint

Cite this