Scholarlink

<p>As part of the eSPERTo paraphrasing project, we developed a novel linguistic and computational resource called “paraphrasary” (<i>Parafrasário</i>, in Portuguese). This article presents its creation and analysis, focusing on a dataset of paraphrastic unit pairs at the multiword, expression, and phrasal levels. The current version of <i>Parafrasário</i> includes paraphrases from both Portuguese from Portugal and Portuguese from Brazil, capturing variety-based differences across the two language varieties. The paraphrastic pairs were aligned using the e-PACT parallel corpus, derived from English literary texts translated into both Portuguese varieties. <i>Parafrasário</i> also encodes recurring paraphrasing patterns as formal rules and classifies the pairs of paraphrastic units into five categories based on two dimensions: surface variation (lexical and morphosyntactic) and semantic difference. We analyzed these categories with attention to morphosyntactic variation, lexical choice, word order, and degrees of semantic similarity. The methodology behind <i>Parafrasário</i> involves identifying expressions with distinctive usage across the aligned sentences and marking differences where variation is clear and widely recognized. A long-term goal is to extend coverage to Lusophone African varieties. The version described here includes approximately 1500 entries and is freely available via the <a href="https://multi3generation.inesc-id.pt/">Multi3Generation COST Action (CA18231)</a> website for research use.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Parafrasário: a variety-based paraphrasary for Portuguese

  • 期刊论文

  • 发表日期：

    2026年3月2日

查看全文
  • Anabela Barreiro,
  • Ida Rebelo-Arnold,
  • Cristina Mota

摘要

As part of the eSPERTo paraphrasing project, we developed a novel linguistic and computational resource called “paraphrasary” (Parafrasário, in Portuguese). This article presents its creation and analysis, focusing on a dataset of paraphrastic unit pairs at the multiword, expression, and phrasal levels. The current version of Parafrasário includes paraphrases from both Portuguese from Portugal and Portuguese from Brazil, capturing variety-based differences across the two language varieties. The paraphrastic pairs were aligned using the e-PACT parallel corpus, derived from English literary texts translated into both Portuguese varieties. Parafrasário also encodes recurring paraphrasing patterns as formal rules and classifies the pairs of paraphrastic units into five categories based on two dimensions: surface variation (lexical and morphosyntactic) and semantic difference. We analyzed these categories with attention to morphosyntactic variation, lexical choice, word order, and degrees of semantic similarity. The methodology behind Parafrasário involves identifying expressions with distinctive usage across the aligned sentences and marking differences where variation is clear and widely recognized. A long-term goal is to extend coverage to Lusophone African varieties. The version described here includes approximately 1500 entries and is freely available via the Multi3Generation COST Action (CA18231) website for research use.