Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification

Tim Schopf; Daniel Braun; Florian Matthes

doi:10.1007/978-3-031-24197-0_4

Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification

Tim Schopf, Daniel Braun, Florian Matthes

Industrial Engineering & Business Information Systems

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

6 Citations (Scopus)

Abstract

In this paper, we evaluate the Lbl2Vec approach for unsupervised text document classification. Lbl2Vec requires only a small number of keywords describing the respective classes to create semantic label representations. For classification, Lbl2Vec uses cosine similarities between label and document representations, but no annotation information. We show that Lbl2Vec significantly outperforms common unsupervised text classification approaches and a widely used zero-shot text classification approach. Furthermore, we show that using more precise keywords can significantly improve the classification results of similarity-based text classification approaches.

Original language	English
Title of host publication	Web Information Systems and Technologies
Subtitle of host publication	16th International Conference, WEBIST 2020, November 3–5, 2020, and 17th International Conference, WEBIST 2021, October 26–28, 2021, Virtual Events, Revised Selected Papers
Editors	Massimo Marchiori, Francisco José Domínguez Mayo, Joaquim Filipe
Place of Publication	Cham
Publisher	Springer
Pages	59-73
Number of pages	15
ISBN (Print)	978-3-031-24197-0
DOIs	https://doi.org/10.1007/978-3-031-24197-0_4
Publication status	Published - 18 Jan 2023

Publication series

Name	Lecture Notes in Business Information Processing

Keywords

NLA

Access to Document

10.1007/978-3-031-24197-0_4

Cite this

Schopf, T., Braun, D., & Matthes, F. (2023). Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification. In M. Marchiori, F. J. Domínguez Mayo, & J. Filipe (Eds.), Web Information Systems and Technologies: 16th International Conference, WEBIST 2020, November 3–5, 2020, and 17th International Conference, WEBIST 2021, October 26–28, 2021, Virtual Events, Revised Selected Papers (pp. 59-73). (Lecture Notes in Business Information Processing). Springer. https://doi.org/10.1007/978-3-031-24197-0_4

Schopf, Tim ; Braun, Daniel ; Matthes, Florian. / Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification. Web Information Systems and Technologies: 16th International Conference, WEBIST 2020, November 3–5, 2020, and 17th International Conference, WEBIST 2021, October 26–28, 2021, Virtual Events, Revised Selected Papers. editor / Massimo Marchiori ; Francisco José Domínguez Mayo ; Joaquim Filipe. Cham : Springer, 2023. pp. 59-73 (Lecture Notes in Business Information Processing).

@inproceedings{c6988df2117e4ab68c03bc711255df0e,

title = "Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification",

abstract = "In this paper, we evaluate the Lbl2Vec approach for unsupervised text document classification. Lbl2Vec requires only a small number of keywords describing the respective classes to create semantic label representations. For classification, Lbl2Vec uses cosine similarities between label and document representations, but no annotation information. We show that Lbl2Vec significantly outperforms common unsupervised text classification approaches and a widely used zero-shot text classification approach. Furthermore, we show that using more precise keywords can significantly improve the classification results of similarity-based text classification approaches.",

keywords = "NLA",

author = "Tim Schopf and Daniel Braun and Florian Matthes",

year = "2023",

month = jan,

day = "18",

doi = "10.1007/978-3-031-24197-0_4",

language = "English",

isbn = "978-3-031-24197-0",

series = "Lecture Notes in Business Information Processing",

publisher = "Springer",

pages = "59--73",

editor = "Massimo Marchiori and {Dom{\'i}nguez Mayo}, {Francisco Jos{\'e}} and Joaquim Filipe",

booktitle = "Web Information Systems and Technologies",

address = "Germany",

}

Schopf, T, Braun, D & Matthes, F 2023, Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification. in M Marchiori, FJ Domínguez Mayo & J Filipe (eds), Web Information Systems and Technologies: 16th International Conference, WEBIST 2020, November 3–5, 2020, and 17th International Conference, WEBIST 2021, October 26–28, 2021, Virtual Events, Revised Selected Papers. Lecture Notes in Business Information Processing, Springer, Cham, pp. 59-73. https://doi.org/10.1007/978-3-031-24197-0_4

Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification. / Schopf, Tim; Braun, Daniel; Matthes, Florian.
Web Information Systems and Technologies: 16th International Conference, WEBIST 2020, November 3–5, 2020, and 17th International Conference, WEBIST 2021, October 26–28, 2021, Virtual Events, Revised Selected Papers. ed. / Massimo Marchiori; Francisco José Domínguez Mayo; Joaquim Filipe. Cham: Springer, 2023. p. 59-73 (Lecture Notes in Business Information Processing).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification

AU - Schopf, Tim

AU - Braun, Daniel

AU - Matthes, Florian

PY - 2023/1/18

Y1 - 2023/1/18

N2 - In this paper, we evaluate the Lbl2Vec approach for unsupervised text document classification. Lbl2Vec requires only a small number of keywords describing the respective classes to create semantic label representations. For classification, Lbl2Vec uses cosine similarities between label and document representations, but no annotation information. We show that Lbl2Vec significantly outperforms common unsupervised text classification approaches and a widely used zero-shot text classification approach. Furthermore, we show that using more precise keywords can significantly improve the classification results of similarity-based text classification approaches.

AB - In this paper, we evaluate the Lbl2Vec approach for unsupervised text document classification. Lbl2Vec requires only a small number of keywords describing the respective classes to create semantic label representations. For classification, Lbl2Vec uses cosine similarities between label and document representations, but no annotation information. We show that Lbl2Vec significantly outperforms common unsupervised text classification approaches and a widely used zero-shot text classification approach. Furthermore, we show that using more precise keywords can significantly improve the classification results of similarity-based text classification approaches.

KW - NLA

U2 - 10.1007/978-3-031-24197-0_4

DO - 10.1007/978-3-031-24197-0_4

M3 - Conference contribution

SN - 978-3-031-24197-0

T3 - Lecture Notes in Business Information Processing

SP - 59

EP - 73

BT - Web Information Systems and Technologies

A2 - Marchiori, Massimo

A2 - Domínguez Mayo, Francisco José

A2 - Filipe, Joaquim

PB - Springer

CY - Cham

ER -

Schopf T, Braun D, Matthes F. Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification. In Marchiori M, Domínguez Mayo FJ, Filipe J, editors, Web Information Systems and Technologies: 16th International Conference, WEBIST 2020, November 3–5, 2020, and 17th International Conference, WEBIST 2021, October 26–28, 2021, Virtual Events, Revised Selected Papers. Cham: Springer. 2023. p. 59-73. (Lecture Notes in Business Information Processing). doi: 10.1007/978-3-031-24197-0_4

Semantic Label Representations with Lbl2Vec: A Similarity-Based Approach for Unsupervised Text Classification

Abstract

Publication series

Keywords

Access to Document

Fingerprint

Cite this