Semantic search as extractive paraphrase span detection: Semantic search as extractive...: J. Kanerva et al.
Saved in:
| Title: | Semantic search as extractive paraphrase span detection: Semantic search as extractive...: J. Kanerva et al. |
|---|---|
| Authors: | Kanerva, Jenna1 (AUTHOR) jmnybl@utu.fi, Kitti, Hanna1 (AUTHOR), Chang, Li-Hsin1 (AUTHOR), Vahtola, Teemu2 (AUTHOR), Creutz, Mathias2 (AUTHOR), Ginter, Filip1 (AUTHOR) |
| Source: | Language Resources & Evaluation. Mar2025, Vol. 59 Issue 1, p257-276. 20p. |
| Subjects: | Heather, Paraphrase, Corpora, Terms & phrases, Possibility |
| Abstract: | In this paper, we approach the problem of semantic search by introducing a task of paraphrase span detection, i.e. given a segment of text as a query phrase, the task is to identify its paraphrase in a given document, the same modelling setup as typically used in extractive question answering. While current work in paraphrasing has almost uniquely focused on sentence-level approaches, the novel span detection approach gives a possibility to retrieve a segment of arbitrary length. On the Turku Paraphrase Corpus of 100,000 manually extracted Finnish paraphrase pairs including their original document context, we find that by achieving an exact match of 88.73 our paraphrase span detection approach outperforms widely adopted sentence-level retrieval baselines (lexical similarity as well as BERT and SBERT sentence embeddings) by more than 20pp in terms of exact match, and 11pp in terms of token-level F-score. This demonstrates a strong advantage of modelling the paraphrase retrieval in terms of span extraction rather than commonly used sentence similarity, the sentence-level approaches being clearly suboptimal for applications where the retrieval targets are not guaranteed to be full sentences. Even when limiting the evaluation to sentence-level retrieval targets only, the span detection model still outperforms the sentence-level baselines by more than 4 pp in terms of exact match, and almost 6pp F-score. Additionally, we introduce a method for creating artificial paraphrase data through back-translation, suitable for languages where manually annotated paraphrase resources for training the span detection model are not available. [ABSTRACT FROM AUTHOR] |
| Copyright of Language Resources & Evaluation is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Text: Availability: 1 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 183750667 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: Semantic search as extractive paraphrase span detection: Semantic search as extractive...: J. Kanerva et al. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Kanerva%2C+Jenna%22">Kanerva, Jenna</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> jmnybl@utu.fi</i><br /><searchLink fieldCode="AR" term="%22Kitti%2C+Hanna%22">Kitti, Hanna</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Chang%2C+Li-Hsin%22">Chang, Li-Hsin</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Vahtola%2C+Teemu%22">Vahtola, Teemu</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Creutz%2C+Mathias%22">Creutz, Mathias</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Ginter%2C+Filip%22">Ginter, Filip</searchLink><relatesTo>1</relatesTo> (AUTHOR) – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Language+Resources+%26+Evaluation%22">Language Resources & Evaluation</searchLink>. Mar2025, Vol. 59 Issue 1, p257-276. 20p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22Heather%22">Heather</searchLink><br /><searchLink fieldCode="DE" term="%22Paraphrase%22">Paraphrase</searchLink><br /><searchLink fieldCode="DE" term="%22Corpora%22">Corpora</searchLink><br /><searchLink fieldCode="DE" term="%22Terms+%26+phrases%22">Terms & phrases</searchLink><br /><searchLink fieldCode="DE" term="%22Possibility%22">Possibility</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: In this paper, we approach the problem of semantic search by introducing a task of paraphrase span detection, i.e. given a segment of text as a query phrase, the task is to identify its paraphrase in a given document, the same modelling setup as typically used in extractive question answering. While current work in paraphrasing has almost uniquely focused on sentence-level approaches, the novel span detection approach gives a possibility to retrieve a segment of arbitrary length. On the Turku Paraphrase Corpus of 100,000 manually extracted Finnish paraphrase pairs including their original document context, we find that by achieving an exact match of 88.73 our paraphrase span detection approach outperforms widely adopted sentence-level retrieval baselines (lexical similarity as well as BERT and SBERT sentence embeddings) by more than 20pp in terms of exact match, and 11pp in terms of token-level F-score. This demonstrates a strong advantage of modelling the paraphrase retrieval in terms of span extraction rather than commonly used sentence similarity, the sentence-level approaches being clearly suboptimal for applications where the retrieval targets are not guaranteed to be full sentences. Even when limiting the evaluation to sentence-level retrieval targets only, the span detection model still outperforms the sentence-level baselines by more than 4 pp in terms of exact match, and almost 6pp F-score. Additionally, we introduce a method for creating artificial paraphrase data through back-translation, suitable for languages where manually annotated paraphrase resources for training the span detection model are not available. [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Language Resources & Evaluation is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=183750667 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1007/s10579-023-09715-7 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 20 StartPage: 257 Subjects: – SubjectFull: Heather Type: general – SubjectFull: Paraphrase Type: general – SubjectFull: Corpora Type: general – SubjectFull: Terms & phrases Type: general – SubjectFull: Possibility Type: general Titles: – TitleFull: Semantic search as extractive paraphrase span detection: Semantic search as extractive...: J. Kanerva et al. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Kanerva, Jenna – PersonEntity: Name: NameFull: Kitti, Hanna – PersonEntity: Name: NameFull: Chang, Li-Hsin – PersonEntity: Name: NameFull: Vahtola, Teemu – PersonEntity: Name: NameFull: Creutz, Mathias – PersonEntity: Name: NameFull: Ginter, Filip IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 03 Text: Mar2025 Type: published Y: 2025 Identifiers: – Type: issn-print Value: 1574020X Numbering: – Type: volume Value: 59 – Type: issue Value: 1 Titles: – TitleFull: Language Resources & Evaluation Type: main |
| ResultId | 1 |