REVIEW 3 major objections 5 minor 29 references
Contrastive Reasons Detection and Clustering from Online Polarized Debate
T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read This paper claims that an unsupervised pipeline can extract and cluster the reasons given by opposing sides in online debates, producing a two-column contrastive digest table without stance labels.
desk verdict The pipeline is a real advance, but the paper's 'significant improvement' claim rests on tables with no error bars or significance tests. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is PhAITV, the Phrase Author Interaction Topic-Viewpoint model: a generative model in which an author first chooses a viewpoint, and each multiword phrase in a post is either drawn from a background distribution or assigned a topic and drawn from a topic-viewpoint word distribution. Reply structure enters through equation (1), which makes the probability that a reply is a rebuttal depend on whether the reply's viewpoint differs from the viewpoints of the parent author's posts; the model fixes the rebuttal indicator $\mathrm{Rb}=1$ for all cross-author replies, so the viewpoint sampler is pushed to choose viewpoints opposed to a post's parents and children. This adversarial pressure, combined with topic-viewpoint assignment of phrases, is what lets the pipeline separate the reasons of the two sides without stance labels.
What would settle it
Run the same pipeline on a debate corpus in which a manual rating shows that fewer than half of replies attack the previous post, then compare the inferred viewpoint clusters with the authors' known stances. If clustering accuracy and relevance rates fall to the level of the unigram or random baselines, the reply-as-rebuttal assumption is doing the work and the claimed advantage does not transfer beyond debate forums with high rebuttal rates.
Extended reading notes
Core claim
The central discovery is that a topic-viewpoint model operating on phrases instead of unigrams, augmented with the assumption that a reply usually rebuts the author it answers, can jointly detect argument facets and assign posts to viewpoints. In the PhAITV model, each phrase is assigned a topic and a viewpoint; phrases assigned to the same topic-viewpoint cell are then grouped by word overlap into clusters treated as argument facets, and the most representative phrase becomes a facet label. A final extraction step picks a short sentence from the posts that realizes each label, filling the digest table. The paper reports that this phrase-based pipeline yields higher relevance rates and better viewpoint clustering than TAM and LAM_LEX on all tested datasets, and that its ROUGE-2 summary scores approach those of correct reference summaries on one dataset. The authors present the result as evidence that facet-guided phrase modeling plus reply interactions is a productive route to contrastive summarization of polarized text.
Load-bearing premise
The method assumes that almost every reply in a debate attacks the post it replies to, so it treats all cross-author replies as rebuttals; if a forum's replies are mostly supportive or neutral, the inferred viewpoints and the clustering built on them become unreliable.
Editorial extensions
If this is right
- A purely unsupervised pipeline can output pro/con digest tables with reasons relevant enough for human annotators to infer the stance of each sentence.
- Multiword phrase modeling carries much of the gain: PhAITV beats its unigram counterpart AITV on relevance and summary informativeness, so facet detection benefits from phrase-level units.
- Author reply interactions are a usable stance signal: even without true viewpoints, the model approaches the performance of PhAITV_view, where true viewpoints are given, especially on negative predictive value.
- The pipeline extends to datasets without any reason or stance annotations: it produced digest tables for 4Forums and Reddit datasets where only text and reply structure are available.
Reading between the lines
- A natural extension is to replace the fixed rebuttal indicator with a learned or lexicon-based disagreement score, which could let the same model work on social media platforms where supportive or neutral replies are common.
- The grouping stage could be retested with noun-phrase-aware or hyponym-based similarity instead of verb-expression preference, which might recover general facets like 'birth control' more consistently.
- Because the paper omits a separate evaluation of post-level viewpoint assignments, a testable next step is to compare PhAITV's inferred post stances against known author stances on the 4Forums data to see whether the viewpoint clustering is accurate beyond the final digest.
- The ROUGE comparison on CreateDebate is built from reference summaries that combine one sentence per annotated reason label, so a direct gold-standard evaluation on 4Forums would test whether the summary informativeness advantage generalizes beyond reason-annotated data.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper presents an unsupervised pipeline for extracting contrastive sentential reasons from online debate forums. The pipeline consists of a phrase-mining stage, a new Phrase Author Interaction Topic-Viewpoint (PhAITV) model with collapsed Gibbs sampling, a grouping and facet-labeling stage, and an extraction stage that produces a two-column digest table of reasons supporting opposing viewpoints. The authors evaluate the pipeline on CreateDebate, 4Forums, and Reddit datasets, comparing against TAM, LAM_LEX, and a unigram AITV variant. Evaluation covers ROUGE-based informativeness, human-judged relevance, and viewpoint clustering accuracy. The abstract claims that the pipeline 'shows a significant improvement over state-of-the-art methods,' but the reported evaluation does not include the statistical machinery needed to support that claim.
Significance. If the empirical claims held, this would be a useful contribution to unsupervised contrastive summarization and argument mining. The model is clearly specified, the pipeline is concrete and reproducible in structure, and the evaluation uses multiple datasets, external baselines, and human annotations. The phrase-level topic-viewpoint modeling and the use of author-reply interactions are plausible design choices. However, the strongest claim in the abstract is currently not supported by the evidence: no confidence intervals, variance measures, or hypothesis tests are reported for the headline comparisons, and model selection is performed on the evaluation data. These issues are load-bearing because the paper's contribution is explicitly framed as a significant empirical improvement. The Rb=1 assumption is a limitation that the authors acknowledge and that is partially supported by dataset statistics, so I view it as secondary to the missing statistical support.
major comments (3)
- [Section 4.4, Table 3 and Section 4.5, Table 4] The abstract's claim of 'significant improvement' is not supported by the reported statistics. Table 3 reports only averages over 10 runs without standard deviations or confidence intervals, and Table 4 reports medians over 5 runs without any measure of dispersion. No paired significance test is reported for any comparison between PhAITV and the baselines. This matters in practice: on Gay Rights, PhAITV's ROUGE-2 values (2.7 recall, 2.8 F-measure) are identical to AITV's, so the headline claim cannot be assessed from the numbers as presented. Please provide per-run results, confidence intervals, or paired tests (e.g., Wilcoxon signed-rank or bootstrap) for the comparisons underlying the abstract's claim.
- [Section 4.2] Model selection is performed on the same data used for evaluation. The authors report choosing K by measuring NPMI coherence on the Topic-Viewpoint clusters of words and selecting hyperparameters by testing combinations that give a 'satisfying overall performance' on the datasets. Since the baselines are run with default parameters while PhAITV is tuned on the evaluation data, the comparison may be biased in PhAITV's favor. Please use a held-out validation split or a nested cross-validation procedure for K and hyperparameter selection, or report a sensitivity analysis showing that the headline conclusions are stable across reasonable choices of K and the Dirichlet parameters.
- [Section 3.2, Eq. (1) and Table 2] The rebuttal variable Rb is fixed to 1 for every reply, and the viewpoint sampling in Eq. (2) depends on this assumption. While the CreateDebate and 4Forums columns of Table 2 report rebuttal rates between 66% and 78%, no rebuttal rate is reported for the Reddit dataset, and the model is applied to Reddit under the same fixed-Rb assumption. Please report the rebuttal rate for the Reddit data or include a sensitivity analysis that relaxes the Rb=1 assumption, and state how the relevance and clustering results change under alternative settings.
minor comments (5)
- [Abstract] The abstract contains a typo ('conv eyed'), which should read 'conveyed'.
- [Section 3.2, Eqs. (3) and (4)] The product indices in the Gibbs sampling equations are written as starting at j=0, but the text describes sampling words within a phrase of length M; please clarify the index bounds and define all count variables explicitly at first use.
- [Algorithm 1] The phrase 'grp.cumulatif NbAssign' mixes French and English and is not defined; it should be 'grp.cumulativeNbAssign' with a formal definition. Also, the containment step in lines 21-30 may transfer assignments from a phrase to multiple containing phrases; please specify how conflicts are resolved when more than one containing phrase is a Verbal Expression.
- [Section 4.3] Report inter-annotator agreement for the argument-facet annotation task. The paper reports 'average Kappa agreement' of 0.66 only for the relevance task in Section 4.5, but no agreement measure is given for the two annotators in Section 4.3.
- [Table 4] The phrase 'notwithstanding PhAITV view' in the table note is unclear; please rephrase to state that PhAITVview is an oracle variant with true viewpoint labels, and that bold-face comparison excludes it.
Circularity Check
No significant circularity: the pipeline's outputs are evaluated against external annotations and baselines, and self-citations are not load-bearing.
full rationale
The paper's derivation chain is self-contained at the level of the claimed contribution. PhAITV is defined by a generative process and Gibbs-sampling equations (Eqs. 1-4) whose inputs are raw posts, author identities, reply structure, and fixed hyperparameters; none of these inputs is the evaluation target. The Rb=1 rebuttal assumption is an explicit modeling choice supported by dataset statistics (Table 2) and is not derived from the ROUGE, relevance, or clustering metrics. Hyperparameter and topic-count selection use NPMI coherence on the model's own output, not the reported evaluation labels, so no reported advantage reduces to a fit to the benchmark. Evaluation uses external resources: Hasan and Ng's CreateDebate reason annotations, the Internet Argument Corpus stance labels, Reddit data, and human annotation, so the reported gains over TAM, LAM_LEX, and AITV are not forced by construction. The authors' prior work appears only as terminology citations and as the AITV ablation baseline; there is no load-bearing self-citation, no uniqueness claim, and no predicted quantity that reduces to a fitted parameter. The abstract's word 'significant' lacks supporting significance tests, but that is a statistical-evidence concern, not a circularity concern under the stated criteria.
Assumptions & free parameters
free parameters (8)
- topic count K =
30, 10, 10, 50 for PhAITV, LAM_LEX, TAM, AITV respectively
- Dirichlet hyperparameter alpha =
0.1
- Dirichlet hyperparameter beta =
1
- viewpoint prior gamma =
1
- background word prior beta_B =
0.1
- rebuttal smoothing parameter eta =
0.01
- route variable prior omega =
10
- phrase significance threshold =
not specified, set empirically
assumptions (6)
- domain assumption Each author draws a viewpoint from a fixed per-author distribution, and each post draws a topic distribution, so latent viewpoint and topic variables explain word choice.
- domain assumption Every reply to another author's post is modeled as a rebuttal (Rb=1), and the probability a rebuttal targets a viewpoint depends on opposition counts in Equation 1.
- ad hoc to paper Lexical overlap between phrases groups them into argument facets, and Verbal Expressions are preferred as facet labels.
- ad hoc to paper The most frequent overlapping word set across a label's sentences, combined with the shortest sentence, yields a representative sentential reason.
- domain assumption NPMI coherence is a valid criterion for selecting K.
- domain assumption Human annotations of whether a sentence contains a stance-inferable reason are reliable enough to score relevance.
Cite this review
Pith. "Pith review of Contrastive Reasons Detection and Clustering from Online Polarized Debate." pith.science (2026). https://pith.science/paper/LJ7MOH7Z
@misc{pith2026190800648,
author = {Pith},
title = {Pith review of: Contrastive Reasons Detection and Clustering from Online Polarized Debate},
year = {2026},
howpublished = {\url{https://pith.science/paper/LJ7MOH7Z}},
note = {Machine review of arXiv:1908.00648}
}
read the original abstract
This work tackles the problem of unsupervised modeling and extraction of the main contrastive sentential reasons conveyed by divergent viewpoints on polarized issues. It proposes a pipeline approach centered around the detection and clustering of phrases, assimilated to argument facets using a novel Phrase Author Interaction Topic-Viewpoint model. The evaluation is based on the informativeness, the relevance and the clustering accuracy of extracted reasons. The pipeline approach shows a significant improvement over state-of-the-art methods in contrastive summarization on online debate datasets.
Figures
Reference graph
Works this paper leans on
-
[1]
Abbott, R., Ecker, B., Anand, P., Walker, M.A.: Internet a rgument corpus 2.0: An sql schema for dialogic social media and the corpora to go w ith it. In: LREC (2016)
work page 2016
-
[2]
Journal of Machine Learning Research 3, 993–1022 (2003)
Blei, D.M., Ng, A.Y., Jordan, M.I.: Latent dirichlet allo cation. Journal of Machine Learning Research 3, 993–1022 (2003)
work page 2003
-
[3]
In: Proceedings of the First Workshop on Argument ation Mining
Boltuˇ zi´ c, F.,ˇSnajder, J.: Back up your stance: Recognizing arguments in o nline dis- cussions. In: Proceedings of the First Workshop on Argument ation Mining. pp. 49–
-
[4]
In: Proceedings of the 2nd Workshop on Argumentation Mining
Boltuˇ zi´ c, F.,ˇSnajder, J.: Identifying prominent arguments in online deb ates using semantic textual similarity. In: Proceedings of the 2nd Workshop on Argumentation Mining. pp. 110–115. Association for Computational Lingui stics, Denver, CO (June 2015), http://www.aclweb.org/anthology/W15-0514
work page 2015
-
[5]
Bouma, G.: Normalized (pointwise) mutual information in collocation extraction. Proceedings of GSCL pp. 31–40 (2009)
work page 2009
- [6]
-
[7]
Erkan, G., Radev, D.R.: Lexrank: Graph-based lexical cen trality as salience in text summarization. J. Artif. Intell. Res.(JAIR) 22(1), 457–479 (2004)
work page 2004
-
[8]
Computational Linguistics 43(1), 125–179 (2017)
Habernal, I., Gurevych, I.: Argumentation mining in user-generated web discourse. Computational Linguistics 43(1), 125–179 (2017). https://doi.org/10.1162/COLI a 00276, https://doi.org/10.1162/COLI_a_00276
doi:10.1162/coli 2017
Show all 29 references
-
[9]
In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP )
Hasan, K.S., Ng, V.: Why are you taking this stance? identi fying and clas- sifying reasons in ideological debates. In: Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP ). pp. 751–
2014
-
[10]
In: Proceedings of the AAAI Conference on Artifici al Intelligence (2015), https://www.aaai.org/ocs/index.php/AAAI/AAAI15/paper/view/9398
Li, H., Mukherjee, A., Si, J., Liu, B.: Extracting verb ex pressions implying negative opinions. In: Proceedings of the AAAI Conference on Artifici al Intelligence (2015), https://www.aaai.org/ocs/index.php/AAAI/AAAI15/paper/view/9398
2015
-
[11]
In: Marie- Francine Moens, S.S
Lin, C.Y.: Rouge: A package for automatic evaluation of s ummaries. In: Marie- Francine Moens, S.S. (ed.) Text Summarization Branches Out : Proceedings of the ACL-04 Workshop. pp. 74–81. Association for Computatio nal Linguistics, Barcelona, Spain (July 2004)
2004
-
[12]
I n: Proceedings of the 2015 Conference of the North American Chapter of the Asso ciation for Computational Linguistics: Human Language Technologies
Misra, A., Anand, P., Fox Tree, J.E., Walker, M.: Using su mmarization to discover argument facets in online idealogical dialog. I n: Proceedings of the 2015 Conference of the North American Chapter of the Asso ciation for Computational Linguistics: Human Language Technologies...
2015
-
[13]
In: Proceedings of th e 21th Workshop on the Semantics and Pragmatics of Dialogue (SemDial 2017)
Misra, A., Oraby, S., Tandon, S., TS, S., Anand, P., Walke r, M.A.: Summarizing dialogic arguments from social media. In: Proceedings of th e 21th Workshop on the Semantics and Pragmatics of Dialogue (SemDial 2017). pp . 126–136 (August 2017)
2017
-
[14]
ACM Trans
Mohammad, S.M., Sobhani, P., Kiritchenko, S.: Stance an d sentiment in tweets. ACM Trans. Internet Technol. 17(3), 26:1–26:23 (Jun 2017). https://doi.org/10.1145/3003433, http://doi.acm.org/10.1145/3003433
2017 doi
-
[15]
In: Proceedings of the First Workshop on Argu mentation Mining
Park, J., Cardie, C.: Identifying appropriate support f or propositions in online user comments. In: Proceedings of the First Workshop on Argu mentation Mining. pp. 29–38. Association for Computational Linguistics, Bal timore, Maryland (June 2014), http://www.aclweb.org/anthol...
2014
-
[16]
In: Proceedings of the 2010 Conference on Empirical Me thods in Natural Language Processing
Paul, M., Zhai, C., Girju, R.: Summarizing contrastive v iewpoints in opinionated text. In: Proceedings of the 2010 Conference on Empirical Me thods in Natural Language Processing. pp. 66–76. Association for Computati onal Linguistics, Cam- bridge, MA (October 2010), http://ww...
2010
-
[17]
In: Proceedings of the 2013 Conference of the No rth American Chap- ter of the Association for Computational Linguistics: Huma n Language Technolo- gies
Qiu, M., Jiang, J.: A latent variable model for viewpoint discovery from threaded forum posts. In: Proceedings of the 2013 Conference of the No rth American Chap- ter of the Association for Computational Linguistics: Huma n Language Technolo- gies. pp. 1031–1040. Association f...
2013
-
[18]
In: Proceedings of the 2014 Conference o n Empiri- cal Methods in Natural Language Processing (EMNLP)
Stab, C., Gurevych, I.: Identifying argumentative disc ourse structures in persuasive essays. In: Proceedings of the 2014 Conference o n Empiri- cal Methods in Natural Language Processing (EMNLP). pp. 46– 56. As- sociation for Computational Linguistics, Doha, Qatar (Oct ober 2...
2014
-
[19]
In: Proceedings of the 16th Annu al Meeting of the Special Interest Group on Discourse and Dialogue
Swanson, R., Ecker, B., Walker, M.: Argument mining: Ext racting argu- ments from online dialogue. In: Proceedings of the 16th Annu al Meeting of the Special Interest Group on Discourse and Dialogue. pp. 21 7–226. Associa- tion for Computational Linguistics, Prague, Czech Repu...
2015
-
[20]
Thonet, T., Cabanac, G., Boughanem, M., Pinel-Sauvagna t, K.: VODUM: A Topic Model Unifying Viewpoint, Topic and Opinion Discovery, pp. 533–545. Springer International Publishing (2016)
2016
-
[21]
In: Proceedings of the 5th Workshop on Lan guage Analysis for Social Media (LASM)
Trabelsi, A., Zaiane, O.R.: Finding arguing expression s of divergent viewpoints in online debates. In: Proceedings of the 5th Workshop on Lan guage Analysis for Social Media (LASM). pp. 35–43. Association for Computa tional Linguistics, Gothenburg, Sweden (April 2014), http:/...
2014
-
[22]
In: M´ etais, E., Roche, M., Teisseire, M
Trabelsi, A., Za ¨ ıane, O.R.: A joint topic viewpoint mod el for contention analysis. In: M´ etais, E., Roche, M., Teisseire, M. (eds.) Natural Lan guage Processing and Information Systems. pp. 114–125. Springer International Publishing, Cham (2014)
2014
-
[23]
In: Proceedings of the 2014 IEEE International Con- ference on Data Mining
Trabelsi, A., Zaiane, O.R.: Mining contentious documen ts using an unsupervised topic model based approach. In: Proceedings of the 2014 IEEE International Con- ference on Data Mining. pp. 550–559 (December 2014)
2014
-
[24]
Data & Knowledge Engineering 100, 226 – 239 (2015)
Trabelsi, A., Za ¨ ıane, O.R.: Extraction and clustering of arguing expres- sions in contentious text. Data & Knowledge Engineering 100, 226 – 239 (2015). https://doi.org/https://doi.org/10.1016/j .datak.2015.05.004, http://www.sciencedirect.com/science/article/pii/S0169023X15000324
2015 doi
-
[25]
Knowledge and Information Systems 48(3), 537– 560 (Sep 2016)
Trabelsi, A., Za ¨ ıane, O.R.: Mining contentious docu- ments. Knowledge and Information Systems 48(3), 537– 560 (Sep 2016). https://doi.org/10.1007/s10115-015-088 8-6, https://doi.org/10.1007/s10115-015-0888-6
2016 doi
-
[26]
In: Procee dings of the AAAI Inter- national Conference on Web and Social Media (ICWSM)
Trabelsi, A., Za ¨ ıane, O.R.: Unsupervised model for top ic viewpoint discovery in online debates leveraging author interactions. In: Procee dings of the AAAI Inter- national Conference on Web and Social Media (ICWSM). pp. 425 –433. Association for the Advancement of Artificia...
2018
-
[27]
In: Proceedings of the 2017 Conference on Empirical Methods in Natural Langu age Processing
Vilares, D., He, Y.: Detecting perspectives in politica l debates. In: Proceedings of the 2017 Conference on Empirical Methods in Natural Langu age Processing. pp. 1573–1582. Association for Computational Linguistics , Copenhagen, Denmark (September 2017), https://www.aclweb.o...
2017
-
[58]
Association for Computational Linguistics, Baltimore , Maryland (June 2014), http://www.aclweb.org/anthology/W14-2107
2014
-
[762]
Association for Computational Linguistics, Doha, Qat ar (October 2014), http://www.aclweb.org/anthology/D14-1083 Contrastive Reasons Detection and Clustering 17
2014
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.