CoChat

Rhetorical Figure Detection: the Case of Chiasmus

Open access
Dubremetz, Marie, Nivre, Joakim · 2015 · Proceedings of the Fourth Workshop on Computational Linguistics for Literature

The DOI matches its Crossref record; 14 of the 15 references it cites were found in a registry or an index; none of the 17 sources its claims cite is checked for support yet.

Abstract

We propose an approach to detecting the rhetorical figure called chiasmus, which involves the repetition of a pair of words in reverse order, as in "all for one, one for all".Although repetitions of words are common in natural language, true instances of chiasmus are rare, and the question is therefore whether a computer can effectively distinguish a chiasmus from a random criss-cross pattern.We argue that chiasmus should be treated as a graded phenomenon, which leads to the design of an engine that extracts all criss-cross patterns and ranks them on a scale from prototypical chiasmi to less and less likely instances.Using an evaluation inspired by information retrieval, we demonstrate that our system achieves an average precision of 61%.As a by-product of the evaluation we also construct the first annotated corpus of chiasmi.

Overview

CoChat Summary

TL;DR

The paper proposes treating chiasmus detection as a graded ranking problem rather than binary classification, using a linear model over four feature groups to rank criss-cross word repetitions. On a Europarl test set, the best system reaches 61% average precision and 90% recall among 19 manually identified true positives.

Key findings

  • The best model (“+Lexical clues”) achieves average precision of 61%, with precision at top 10, 50, 100, and 200 candidates of 70%, 28%, 17%, and 9%, respectively.
  • Adding size features finds 14 of 19 chiasmi within the top 200 hits (74% recall); adding N-gram similarity raises this to 17 of 19 (90% recall), while lexical clues improve precision slightly without increasing recall.
  • At candidate rank 9, the proposed system has precision 78%, recall 37%, and F1 59%, compared with Hromada (2011) at precision 67%, recall 32%, F1 43%, and Gawryjolek (2009) at 0% precision and recall.
  • The system processes one million words in 214 seconds, about three times slower than Hromada’s 78 seconds per million words, and the work also produces 1200 manually annotated criss-cross patterns as true, false, or borderline cases.

Methods

  • Experiments use English Europarl, split into two two-million-word extracts: one tuning corpus and one held-out test corpus.
  • The system tokenises and lemmatises with TreeTagger, extracts every pair of repeated and inverted lemmas within a 30-token window, and scores candidates with a linear model using 20 features in four groups: basic stopword/punctuation, size, N-gram similarity, and lexical clues.
  • Weights are set empirically by observing successive results on the tuning set; evaluation manually annotates the top 200 hits from each experiment and reports precision-recall graphs and average precision.

Limitations

  • The paper states that because of the small number of positive examples, only tentative conclusions can be drawn and the results might not be statistically significant.
  • The paper states that manual tuning means not every possible weight combination was tried, and that no existing annotated data prevented supervised machine learning or absolute recall estimation.
  • Evident from the text: evaluation is limited to English political discourse from Europarl, recall is estimated from only 19 true positives found through top-200 annotations across four experiments, borderline cases are excluded from results, and annotation relies on human intuition because no clear discriminative test for chiasmus is stated.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3115/v1/w15-0703

Cites

14 of 14 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Schema of a chiasmus.
Figure 1 Schema of a chiasmus.From the PDF
Figure 3 Precision-Recall graphs for the top two hundred candidates in each experiment. (The 'basic' experiment is not included because of too low precision.)
Figure 3 Precision-Recall graphs for the top two hundred candidates in each experiment. (The 'basic' experiment is not included because of too low precision.)From the PDF
Image 3
From the PDF

Verification

Claims the paper makes

0 Valid · 2 Not valid · 14 Not checked · 24 Own results and interpretations

10 more claims

Create a free account to see all 40 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 16
  • Probably not a substantial claim. 11

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 14 references

Research works printed but not linked: 3

  1. Marie Dubremetz. 2013. Vers une identification automatique du chiasme de mots. In Actes de la 15e Rencontres des ´ Etudiants Chercheurs en Informatique pour le Traitement Automatique des Langues (RECITAL'2013) , pages 150-163, Les Sables d'Olonne, France.
  2. Randy Harris and Chrysanne DiMarco. 2009. Constructing a Rhetorical Figuration Ontology. In Persuasive Technology and Digital Behaviour Intervention Symposium , pages 47-52, Edinburgh, Scotland.
  3. Helmut Schmid. 1994. Probabilistic Part-of-Speech Tagging Using Decision Trees. In Proceedings of International Conference on New Methods in Language Processing , pages 44-49, Manchester, Great Britain.

Cited by

See all 17

Metadata

Authors
Dubremetz, M., & Nivre, J.
Affiliations
Uppsala University (SE)
Venue
Proceedings of the Fourth Workshop on Computational Linguistics for Literature
Published
2015
Pages
23-31
OpenAlex
W2250667488
Cited by
16
References
15
Citations by year
2024: 1 · 2023: 1 · 2021: 1 · 2020: 1 · 2018: 2 · 2017: 6 · 2016: 3 · 2015: 1
FWCI
2.15
Type
conference paper
Language
en
Open access
Licence
cc-by
Locations
Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature (LaTeCH-CLfL) (publishedVersion, cc-by)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Physical Sciences › Computer Science › Artificial Intelligence › Advanced Text Analysis Techniques
Keywords
chiasmus, rhetorical devices, word repetition, annotated corpus, information retrieval evaluation

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.