Toxicity Begets Toxicity: Unraveling Conversational Chains in Political Podcasts

Animesh Mukherjee; Kiran Garimella; Naquee Rizwan; Nayandeep Deb; Sarthak Roy; Vishwajeet Singh Solanki

REVIEW 4 major objections 7 minor 37 references

Toxicity Begets Toxicity: Unraveling Conversational Chains in Political Podcasts

T0 review · 4 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read In political podcasts, toxicity builds through reply chains to a peak 'anchor' segment and then decays, and the paper argues this structure makes real-time moderation feasible.

desk verdict A useful podcast toxicity dataset and change-point benchmark wrapped around a headline finding that is confounded by the max-over-chunks anchor definition; the descriptive claims need a controlled re-run before they can be trusted. read the letter →

arxiv 2501.12640 v2 pith:IUMGODGK submitted 2025-01-22 cs.CL cs.AI

Naquee Rizwan , Nayandeep Deb , Sarthak Roy , Vishwajeet Singh Solanki , Kiran Garimella , Animesh Mukherjee This is my paper · ORCID

classification cs.CLcs.AI

keywords toxicconversationchainspodcastspoliticaltranscriptschangepointdetectiontoxicitybegetscontentmoderationconversationaldynamics

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that toxicity in political podcasts follows a reproducible conversational arc: it builds across reply chains, peaks in a highly toxic 'anchor' segment, and then fades, so that toxic talk begets toxic talk rather than appearing as an isolated outburst. To test this, the authors transcribe and diarize 12,322 episodes from 52 popular U.S. political podcasts, construct 15,758 toxic conversation chains of 21 roughly one-minute segments each, and compare the anchor segment with its ten predecessors and ten successors. They report that anchor segments are longer, more repetitive, less lexically diverse, richer in metaphors and hyperboles, and emotionally dominated by anger, fury, and annoyance. They also manually annotate change points in 200 chains and show that an audio-capable large language model locates those shifts more precisely than classical change-point algorithms, which they read as evidence that real-time moderation of podcasts is feasible.

What carries the argument

The central object is the toxic conversation chain, a 21-segment window around an anchor segment: the anchor is a speaker-turn chunk, split into one to four sub-chunks of about 17 seconds each, whose segment-level toxicity score, defined as the maximum across its chunks, is at least 0.7. The ten preceding and ten following segments provide the conversational context. This construction turns continuous audio into comparable units, letting the paper ask what rises before a toxic peak, what happens at the peak, and what decays afterward. A second mechanism is change point detection on the toxicity time series, with human majority-voted annotations of tone shifts, topical shifts, and toxicity changes used as ground truth to evaluate both classical algorithms and large language models.

What would settle it

Re-run the chain analysis with segment toxicity defined as the mean or per-token toxicity instead of the maximum, and with anchor segments matched to non-anchor segments of equal duration and token count; if anchors no longer show higher repetition, more hyperbole, and stronger anger relative to their neighbors, the central characterization fails. A second falsifier is behavioral: if human annotators shown only the pre-anchor segments cannot predict where toxicity will peak above chance, then 'toxicity begets toxicity' fails as a predictive claim about conversational escalation.

Watch

Extended reading notes

Core claim

The authors introduce toxic conversation chains, built by splitting transcribed, diarized podcast speech into segments and marking as anchors those segments whose toxicity score from a commercial toxicity API is at least 0.7, then taking the ten segments before and after each anchor. The central discovery is that the anchor is not merely a random spike: it is the most linguistically distinctive point of the chain, where speakers talk longer, repeat themselves more, produce more metaphors and roughly double the proportion of hyperboles, and where emotion terms shift toward anger, fury, annoyance, and disgust while positive and anticipatory emotions drop. Manual annotation of 200 highly toxic chains shows that most contain two to four change points, and the best-performing audio-capable large language model detects those points more accurately than classical change-point detectors, particularly in precision and Hausdorff distance. The authors conclude that these patterns indicate toxic talk escalates through conversational structure and that automated tools can monitor the trajectory in real time.

Load-bearing premise

The entire comparison rests on the definition that a segment's toxicity is the maximum toxicity of its chunks, so a longer segment has more chances to contain one highly toxic chunk, which makes 'anchor segments are longer' partly a consequence of the measurement and the later comparisons of repetition, entropy, and emotion inherit the same length bias.

Editorial extensions

If this is right

Podcast platforms could automatically transcribe, score, and track toxicity per segment, flagging anchor peaks in near real time rather than reviewing hours of audio.
Hosts or moderators could be alerted during the pre-anchor buildup, when segment duration and repetition begin to rise, allowing intervention before toxicity peaks.
The 200-chain annotated change-point dataset provides a benchmark for future methods that detect tone, topic, and toxicity shifts in long-form spoken conversation.
Because the pattern appears in both right- and left-leaning channels, the rise-peak-decay structure may be a general property of political talk rather than an artifact of one ideology.
The finding that anchor speech is more repetitive and less organized suggests textual redundancy itself could serve as a cheap signal in automatic toxicity monitoring.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If toxicity is genuinely a chain-level phenomenon, then moderation policies that score individual utterances in isolation will systematically miss the buildup phase; effective intervention should target the pre-anchor trajectory, not just the peak.
The max-over-chunks definition of segment toxicity means longer segments are likelier to contain one highly toxic chunk, so the 'anchors are longer' result may be partly definitional; a natural test is to recompute the analysis with mean toxicity or with duration-matched anchor and non-anchor segments.
The success of audio-capable language models suggests the transcribed words are not the only carrier of escalation; prosody, volume, or speech overlap may signal a coming toxic peak earlier than the text does, a testable extension of the change-point setup.
Because the dataset comes from publicly available RSS feeds and automatic transcription, prevalence numbers likely underestimate toxicity delivered through tone, sarcasm, or culturally specific references, so a multimodal re-analysis could reveal additional anchors.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Two things worth knowing. First, the paper ships a genuinely useful resource: 12,322 transcribed and diarized episodes from 52 political podcasts, plus a human-annotated change-point dataset on 200 toxic chains, with code and data linked. The CPD benchmark is also well executed: they compare classical algorithms, LLMs, and audio-LLMs with zero-shot and chain-of-thought prompting, report means and standard deviations across three runs, and repeat the key comparisons with alternate toxicity scorers. That part earns its keep.

Second, the headline descriptive finding—that the anchor segment is longer, more repetitive, less coherent, and emotionally hotter—is not established. The problem is in Section 4: segment toxicity is the maximum over the 1–4 chunks in the segment. A four-chunk segment has four draws from the chunk toxicity distribution; a one-chunk segment has one. So longer segments are more likely to clear the 0.7 anchor threshold even if speakers do not actually speak longer when toxic. The observed longer anchor duration is partly a selection artifact, and token count, type-token ratio, entropy, and perplexity are all length-sensitive. The control group in Appendix E.5 matches channel distribution but not chunk count or duration, and its "anchors" are not selected by the same max rule, so it cannot break the confound. The paper even concedes in E.1 that token count "naturally follows" from the longer duration. Contribution (iv)(a), (iv)(b), and the length-sensitive parts of (iv)(c) therefore need a re-analysis that matches chunk count or segment duration, or uses a permutation baseline.

The "toxicity begets toxicity" framing has a related problem. Selecting segments above 0.7 and then showing that neighbors are lower is partly definitional. The V-shaped pattern could be real escalation and decay, but the current design cannot distinguish that from regression to the mean around a max-selected point. They should compare against chains selected at random or with the same max rule on non-toxic scores.

What is solid: the human validation that anchors are actually toxic (86% agreement, Cohen's kappa 0.69) and the audio-LLM change-point results, which are reproducible and benchmarked against classical methods. The citation pattern is appropriate, and the limitations appendix is honest about what is not answered.

Who should read it: anyone working on podcast moderation, toxicity dynamics, or audio-LLM evaluation. It deserves a serious referee—but the referee should ask for the controlled re-run before the descriptive claims are accepted. If the authors cannot control for the confound, the paper should stand on the dataset and CPD benchmark alone.

Referee Report

4 major / 7 minor

Summary. The paper presents a dataset of 12,322 episodes from 52 US political podcasts, transcribed with Whisper and diarized with Pyannote. It defines a 'toxic conversation chain' as a 21-segment window around an 'anchor segment,' where a segment is a 1-4 chunk speaker turn and its toxicity is the maximum Perspective API score over its chunks. An anchor is any segment with toxicity at least 0.7. The authors then compare anchor segments with the 10 preceding and 10 following segments on duration, token count, type-token ratio, entropy, perplexity, figurative language (metaphor/hyperbole), and Empath emotion features. They also manually annotate change points in the top 200 chains, and evaluate classical CPD algorithms, LLMs, and audio-LLMs against these annotations. The headline claims are that anchors are longer, more repetitive, less coherent, more figurative, and more anger-laden, and that toxicity 'begets toxicity' across the chain.

Significance. If the descriptive findings held, the paper would contribute a new podcast toxicity dataset, a new unit of analysis (toxic conversation chains), and a reproducible evaluation of LLM/audio-LLM change point detection that is relevant for real-time content moderation. The paper has concrete strengths: the dataset and code are released, the CPD experiments use three independent runs at temperature 0.001, the annotation process is described with inter-annotator agreement, and robustness checks include Detoxify, GPT-4o classification, and human evaluation. However, the central descriptive claims are currently undermined by a definitional confound in Section 4: since a segment's toxicity is the max over 1-4 chunks, longer segments are mechanically more likely to become anchors, and the reported length and length-sensitive linguistic differences are therefore partly artifacts of the selection rule. The change point detection part is less affected by this confound and retains value, but the paper's main narrative needs a reanalysis with proper controls.

major comments (4)

[Section 4, segment toxicity definition] The definition 'The toxicity score of a segment is set to the maximum of the toxicity score across these chunks within the segment' introduces a direct selection bias: a 4-chunk segment has up to four chances to contain a high-scoring chunk, while a 1-chunk segment has only one. Since anchors are segments with max toxicity >= 0.7, longer segments are more likely to be selected as anchors regardless of whether speakers actually speak longer when toxic. The finding in Section 5.1(i) and Figure 5 that anchor segments have the highest mean duration is therefore partly a mechanical consequence of this rule, and the same bias propagates to token count (Appendix E.1, Figure 11) and to any downstream comparison of anchors with their neighbors. The paper needs a reanalysis that controls for the number of chunks per segment or uses a fixed-length window for anchor detection.
[Section 5.1(ii) and Appendix E.1(ii)] The reported differences in type-token ratio, unigram entropy, and perplexity are all length-sensitive. TTR mechanically decreases with text length even for randomly sampled text, and entropy and perplexity estimates are affected by the amount of text available in a segment. Since the anchor segments are longer on average (and longer by construction through the max-over-chunks rule), the observed lower TTR and higher entropy/perplexity in the anchor segment (Figures 5 and 11) are confounded. The analysis should be redone on segments matched for duration or chunk count, or with length-normalized statistics, before the claims in contribution (iv)(a)-(b) can be accepted.
[Appendix E.5, control group] The non-toxic control group in Appendix E.5 matches the distribution of podcast channels but does not match chunk count or segment duration, and its 'anchor' segments are not selected by the same max-over-chunks rule. Consequently, Figure 13 cannot disentangle the effect of toxicity from the mechanical effect of the anchor-selection procedure. A valid control should apply the same selection rule to toxicity-shuffled or randomly placed pseudo-anchors, and should match the number of chunks per segment, so that any remaining differences can be attributed to toxicity rather than to selection artifacts.
[Title, abstract, and chain construction] The framing 'toxicity begets toxicity' implies an escalation toward the anchor and a decay afterward, but the chain construction selects anchors exactly at local maxima (toxicity >= 0.7), which guarantees that neighboring segments will have lower toxicity on average even under pure noise. The paper does not report the mean toxicity trajectory across chain positions, nor does it compare it with a null model using randomly placed pseudo-anchors. As a result, the escalation/decay narrative is not currently supported by the evidence presented; this is load-bearing for the paper's central claim and should be addressed explicitly, either by providing the missing trajectory analysis or by softening the claim.

minor comments (7)

[Section 2] The novelty claim 'we are the first to look at toxicity in podcasts' contradicts the immediately preceding discussion citing Wirtschafter (2021) and Litterer et al. (2024) as prior work on toxicity in podcasts; please rephrase or clarify the precise novelty.
[Section 4] The sentence 'We manually tune to ten previous and next segments' introduces a free parameter without sensitivity analysis; the paper should either report results for other context lengths or justify why the findings are robust to this choice.
[Section 5.2] The metaphor and hyperbole percentages in Figure 6 are reported without confidence intervals or significance tests; given the large sample sizes, bootstrap confidence intervals or a paired test across chain positions would strengthen the claim.
[Section 6.1] The statement that 'only rbf as the cost function coupled with Pelt, KernelCPD, BottomUp & Binseg search methods successfully identifies at least one change point' should state explicitly that all other cost functions (cosine, l2, linear) failed to identify any change point, to avoid ambiguity.
[Section 6.5] The text refers to 'toxic-bert-original and toxic-bert-biased', but the models are described in Section 4 as 'original & unbiased'; the label 'biased' appears to be a typo and should be corrected to 'unbiased' in the text and in Table 2.
[Appendix A] The limitations section does not mention the max-over-chunks selection confound discussed above; the authors should add a explicit limitation and describe the planned reanalysis to address it.
[Throughout] There are several typos, including 'usestate-of-the-art' in Section 4 and 'lead to' in Section 7 (should be 'led to'); a careful proofread is recommended.

Circularity Check

3 steps flagged · score 6.0 of 10

Anchor 'longer duration' and related textual findings are partly forced by the max-over-chunks anchor definition; change-point and figurative-language analyses are independent.

self definitional [Section 4 (anchor definition) and Section 5.1(i) (Time coverage)]
"We club 1-4 chunks (approximately one minute) into a segment... The toxicity score of a segment is set to the maximum of the toxicity score across these chunks within the segment... A segment is identified as an anchor segment if its toxicity score is 0.7 or higher... Interestingly, the anchor segment has the highest mean implying that speakers tend to speak for longer duration when their speech is most toxic."

Chunks have roughly fixed length, so segment duration is a proxy for the number of chunks in the segment. The segment toxicity is the maximum over those chunks, and the probability that a maximum of k draws crosses the 0.7 threshold stochastically increases with k. Longer segments (more chunks) therefore become anchors more often even if per-chunk toxicity is unrelated to duration. The 'anchor segments are longer' result in Figure 5 is thus partly an artifact of the anchor definition, and no chunk-count or duration-matched control is provided.
other [Appendix E.1 (Token count and TTR) and contribution (iv)(b)]
"Figure 11 shows that the mean token count is the largest for the anchor segment which naturally follows from the earlier observation that these segments have the longest mean duration... Figure 11 shows that the mean TTR for anchor segment is lower than all other segments in the chain, indicating that there is more repetition (possibly of the same hateful remark) in the anchor segment."

Token count is a deterministic increasing function of segment duration, so the largest-token-count claim inherits the max-rule selection bias from the anchor definition. TTR mechanically declines as token count increases for any natural text, so the lower TTR of anchor segments is a length confound rather than independent evidence of toxic repetition. The 'anchor is more repetitive' part of contribution (iv) therefore reduces to the same construction unless token count or duration is controlled.

1 more flagged steps

other [Appendix E.5 (Right-leaning control group)]
"we choose such conversations that have the toxicity value of the anchor segment lower than 0.3 (as per the recommendation of Perspective API) for right-leaning channels. We also take care that the randomly selected chains have a distribution in line with the distribution of podcast channels in toxic conversation chains."

The control group matches the podcast-channel distribution but does not match segment duration or chunk count, and its 'anchor' is selected by the opposite max-below-0.3 rule. Since the confound is precisely that the max-based anchor rule favors longer segments with more chunks, this control cannot disentangle duration or chunk count from toxicity, so it does not establish that the time-coverage, TTR, entropy, or perplexity differences are due to toxicity rather than segment length.

full rationale

The central 'anchor characteristics' claims are partially circular. Section 4 defines segment toxicity as the maximum of Perspective chunk scores over 1-4 chunks, so longer segments have more chances to contain a high-scoring chunk and are more likely to be selected as anchors. Section 5.1 then reports that anchor segments have the longest mean duration, which is a direct consequence of this selection rule rather than an independent observation about speaker behavior. Appendix E.1's token-count and TTR findings inherit the same length confound: token count grows with duration, and TTR falls with token count even for random text. The control group in Appendix E.5 matches channel distribution but not duration or chunk count, so it cannot remove the bias. The remaining analyses are largely independent: metaphor/hyperbole detection (Section 5.2) and empath emotion features (Section 5.3) compare segment content without using the max rule in their measurement, and the change-point evaluation (Section 6) is anchored to three human annotators, providing external ground truth. Self-citations in the paper are not load-bearing for the derivation. Overall, the paper's central claim about anchor length and repetition is partly forced by the anchor definition, giving a partial circularity score of 6; the change-point and figurative-language contributions retain independent content.

Assumptions & free parameters 3 free parameters · 5 assumptions · 1 invented entities

The central findings depend on a small number of hand-picked parameters (chain length, chunk duration, chunks per segment) and on the choice of max aggregation for segment toxicity. These choices are not derived from theory and directly shape the observed patterns.

free parameters (3)

Chain context length (previous/next segments) = 10
Manually tuned in Section 4: 'We manually tune to ten previous and next segments, as this gives a perfect amount of back-and-forth context around the anchor point.'
Chunk duration = 17 seconds
Set to the median conversation time of a turn in the dataset, Section 4: 'we split the diarized text corresponding to a speaker turn into chunks of 17 seconds since this is the median conversation time of a turn in our dataset.'
Maximum chunks per segment = 4 chunks
Chosen by hand, Section 4: 'We club 1-4 chunks (approximately one minute) into a segment.' This bounds the max-aggregation effect.

assumptions (5)

domain assumption Perspective API toxicity scores are a valid operationalization of toxicity for podcast transcripts.
Used to define anchors and to prompt CPD models; validated only indirectly against Detoxify, GPT-4o, and human labels on a subset.
ad hoc to paper Segment-level toxicity is best summarized by the maximum chunk toxicity.
Section 4: 'The toxicity score of a segment is set to the maximum of the toxicity score across these chunks within the segment.' This choice creates the length confound.
domain assumption Whisper transcription and Pyannote diarization are accurate enough for chain-level analysis.
Section 3 cites roughly 90% diarization accuracy from prior work and applies no downstream verification of transcript fidelity.
ad hoc to paper Ten segments before and after an anchor form a meaningful conversational unit.
Section 4: manually tuned; no external validation that this window captures complete conversational moves.
domain assumption Manual annotations of change points by three experts are ground truth for evaluation.
Section 6.2; inter-annotator agreement is not reported for change point annotations, only that a majority voting scheme was used.

invented entities (1)

Toxic conversation chain
purpose: Analytical unit consisting of an anchor segment (toxicity >= 0.7) plus ten preceding and ten following segments, used to study the rise and fall of toxicity.
This is a definitional construct introduced by the paper; no independent evidence shows it corresponds to a natural conversational boundary.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Toxicity Begets Toxicity: Unraveling Conversational Chains in Political Podcasts." pith.science (2026). https://pith.science/paper/IUMGODGK

@misc{pith2026250112640,
  author       = {Pith},
  title        = {Pith review of: Toxicity Begets Toxicity: Unraveling Conversational Chains in Political Podcasts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IUMGODGK}},
  note         = {Machine review of arXiv:2501.12640}
}

read the original abstract

Tackling toxic behavior in digital communication continues to be a pressing concern for both academics and industry professionals. While significant research has explored toxicity on platforms like social networks and discussion boards, podcasts despite their rapid rise in popularity remain relatively understudied in this context. This work seeks to fill that gap by curating a dataset of political podcast transcripts and analyzing them with a focus on conversational structure. Specifically, we investigate how toxicity surfaces and intensifies through sequences of replies within these dialogues, shedding light on the organic patterns by which harmful language can escalate across conversational turns. Warning: Contains potentially abusive/toxic contents.

Figures

Figures reproduced from arXiv: 2501.12640 by the authors.

**Figure 1.** Top: Computation of segments from chunks and their contents. Bottom: Schema for toxic conversation chains. The segment marked in red color represents the anchored segment with toxicity above a threshold of 0.7. to a speaker turn into chunks of 17 seconds since this is the median conversation time of a turn in our dataset (see [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 3.** Distribution of toxic conversation chains across the podcast channels for each leaning, i.e. right and left . Percentage contribution for Top 10 podcast channels are shown. Distribution of the toxic chains across podcasts: We compute the distribution of the toxic chains across the different podcast channels [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 2.** Top 10 podcast shows with most amount of toxic content for each leaning, i.e. right and left . The bars show the percentage of episodes containing at least one toxic conversation. Most toxic channels [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Figures from the paper (9 more)

**Figure 4.** Figure 4: Example: Since it is not feasible to illustrate all segments, one among the previous and next segments is shown along with the anchor segment from the toxic conversation chain. Toxic texts in the anchor segment are marked in red. Note: start and end times are in second…

**Figure 5.** Figure 5: shows that the anchor segment has a higher average entropy compared to the other segments indicating that the text in the anchor segment is more random and less well-formed. Further, we also calculate the perplexity of the diarized text assuming uni-gram probabilities…

**Figure 7.** Figure 7: Empath Features: Plot on the left presents the features those which increased for anchor segment. Plot on the right illustrates the ones with decreased value. In each of these plots, anchor point is ‘0’; left & right semicircles represent previous and next segments res…

**Figure 8.** Figure 8: Cpd Examples: Plots representing two samples comparing human annotation with GPT-4o-Audio zero-shot setup’s detected change points. Correctly predicted points are marked with a red hexagon and incorrect predictions are marked with red cross. 1 and 7, respectively. The …

**Figure 9.** Figure 9: Employed Prompts: Textual prompts used on Qwen-2 and GPT-4o are in sub-figures (a) and (b). Similarly, sub-figures (c) and (d) are employed with Qwen-2-Audio and GPT-4o-Audio. Different components of an input prompt are separated within individual boxes. Each of the su…

**Figure 10.** Figure 10: Conversation Examples: One among previous and next segments are plotted along with anchor segment since it is not feasible to plot all segments. Toxic contents in the anchor segment are marked in red color. Note: start and end times are in seconds. Anchor Previous Nex…

**Figure 11.** Figure 11: Token count (left) and type token ratio (right) across segments at 95% confidence intervals. E.4 Number of speakers in toxic chains We present the number of speaker turns and its distribution over 8,634 toxic chains from right-leaning channels presented as (#speakers…

**Figure 13.** Figure 13: Control Group Statistics: Mean at 95% confidence intervals for time coverage, entropy, perplexity, token count & ttr for control group across segments. 10 8 6 4 2 0 2 4 6 8 10 26 28 30 32 34 36 38 Time coverage (s) 10 8 6 4 2 0 2 4 6 8 10 4.8 4.9 5.0 5.1 5.2 5.3 5.4 5…

**Figure 14.** Figure 14: Left-leaning Statistics: Mean at 95% confidence intervals for time coverage, entropy, perplexity, token count & ttr for left-leaning podcast channels across segments [PITH_FULL_IMAGE:figures/full_fig_p014_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 15 canonical work pages

[1]

Qwen2 Technical Report

2024. Qwen2 Technical Report. (2024)

2024
[2]

Naveen Badathala, Abisek Rajakumar Kalarani, Tejpalsingh Siledar, and Pushpak Bhattacharyya. 2023. A Match Made in Heaven: A Multi-task Framework for Hyperbole and Metaphor Detection. arXiv:2305.17480 [cs.CL] https://arxiv.org/ abs/2305.17480

arXiv 2023
[3]

Utkucan Balci, Jay Patel, Berkan Balci, and Jeremy Blackburn. 2024. Podcast Outcasts: Understanding Rumble’s Podcast Dynamics. arXiv:2406.14460 [cs.SI] https://arxiv.org/abs/2406.14460

work page Pith review arXiv 2024
[4]

Doug Beeferman, William Brannon, and Deb Roy. 2019. Radiotalk: A large-scale corpus of talk radio transcripts. arXiv preprint arXiv:1907.07073 (2019)

work page Pith review arXiv 2019
[5]

Hervé Bredin and Antoine Laurent. 2021. End-to-end speaker segmentation for overlap-aware resegmentation. arXiv preprint arXiv:2104.04045 (2021)

arXiv 2021
[6]

Konijn, and Gerard J

Christian Burgers, Elly A. Konijn, and Gerard J. Steen. 2016. Fig- urative Framing: Shaping Public Discourse Through Metaphor, Hy- perbole, and Irony. Communication Theory 26, 4 (04 2016), 410–

work page 2016
[7]

Yunfei Chu, Jin Xu, Qian Yang, Haojie Wei, Xipin Wei, Zhifang Guo, Yichong Leng, Yuanjun Lv, Jinzheng He, Junyang Lin, et al. 2024. Qwen2-audio technical report. arXiv preprint arXiv:2407.10759 (2024)

arXiv 2024
[8]

Ann Clifton, Sravana Reddy, Yongze Yu, Aasish Pappu, Rezvaneh Rezapour, Hamed Bonab, Maria Eskevich, Gareth Jones, Jussi Karlgren, Ben Carterette, and Rosie Jones. 2020. 100,000 Podcasts: A Spoken English Document Corpus. In Proceedings of the 28th International Conference on Computational Linguistics , Donia Scott, Nuria Bel, and Chengqing Zong (Eds.). I...

work page 2020

Show all 37 references

[9]

Daryna Dementieva, Daniil Moskovskiy, Nikolay Babakov, Abinew Ali Ayele, Naquee Rizwan, Frolian Schneider, Xintog Wang, Seid Muhie Yimam, Dmitry Ustalov, Elisei Stakovskii, Alisa Smirnova, Ashraf Elnagar, Animesh Mukherjee, and Alexander Panchenko. 2024. Overview of the Multil...

2024
[10]

Manqing Dong, Hao Huang, and Longbing Cao. 2024. Can LLMs Serve As Time Series Anomaly Detectors? arXiv:2408.03475 [cs.LG] https://arxiv.org/abs/2408. 03475

2024 arXiv
[11]

Lei Gao and Ruihong Huang. 2017. Detecting Online Hate Speech Using Context Aware Models. In Proceedings of the International Conference Recent Advances in Natural Language Processing, RANLP 2017 , Ruslan Mitkov and Galia Angelova (Eds.). INCOMA Ltd., Varna, Bulgaria, 260–266....

2017 doi
[12]

Maarten Grootendorst. 2020. KeyBERT: Minimal keyword extraction with BERT. doi:10.5281/zenodo.4461265

2020 doi
[13]

Maarten Grootendorst. 2022. BERTopic: Neural topic modeling with a class-based TF-IDF procedure. arXiv:2203.05794 [cs.CL] https://arxiv.org/abs/2203.05794

2022 arXiv
[14]

Laura Hanu and Unitary team. 2020. Detoxify. Github. https://github.com/unitaryai/detoxify

2020
[15]

Prince Jha, Raghav Jain, Konika Mandal, Aman Chadha, Sriparna Saha, and Push- pak Bhattacharyya. 2024. MemeGuard: An LLM and VLM-based Framework for Advancing Content Moderation via Meme Intervention. arXiv:2406.05344 [cs.CL] https://arxiv.org/abs/2406.05344

2024 arXiv
[16]

Paddy Leerssen. 2023. An end to shadow banning? Transparency rights in the Digital Services Act between content moderation and curation. Computer Law & Security Review 48 (2023), 105790

2023
[17]

Tran, Yi Tay, Jeffrey Sorensen, Jai Gupta, Donald Metzler, and Lucy Vasserman

Alyssa Lees, Vinh Q. Tran, Yi Tay, Jeffrey Sorensen, Jai Gupta, Donald Metzler, and Lucy Vasserman. 2022. A New Generation of Perspective API: Efficient Multilingual Character-level Transformers. arXiv:2202.11176 [cs.CL] https: //arxiv.org/abs/2202.11176

2022 arXiv
[18]

Benjamin Litterer, David Jurgens, and Dallas Card. 2024. Mapping the Pod- cast Ecosystem with the Structured Podcast Research Corpus. arXiv preprint arXiv:2411.07892 (2024)

2024
[19]

Binny Mathew, Punyajoy Saha, Hardik Tharad, Subham Rajgaria, Prajwal Sing- hania, Suman Kalyan Maity, Pawan Goyal, and Animesh Mukherjee. 2019. Thou Shalt Not Hate: Countering Online Hate Speech. Proceedings of the Interna- tional AAAI Conference on Web and Social Media 13, 01...

2019 doi
[20]

Binny Mathew, Punyajoy Saha, Seid Muhie Yimam, Chris Biemann, Pawan Goyal, and Animesh Mukherjee. 2022. HateXplain: A Benchmark Dataset for Explainable Hate Speech Detection. arXiv:2012.10289 [cs.CL] https://arxiv.org/abs/2012.10289

2022 arXiv
[21]

Shervin Minaee, Tomas Mikolov, Narjes Nikzad, Meysam Chenaghlu, Richard Socher, Xavier Amatriain, and Jianfeng Gao. 2024. Large Language Models: A Survey. arXiv:2402.06196 [cs.CL] https://arxiv.org/abs/2402.06196

2024 arXiv
[22]

OpenAI. 2024. GPT-4o — openai.com. https://platform.openai.com/docs/guides/ audio/. [Accessed 15-02-2024]

2024
[23]

Sachin Pathiyan Cherumanal, Ujwal Gadiraju, and Damiano Spina. 2024. Every- thing We Hear: Towards Tackling Misinformation in Podcasts. InInternational Conference on Multimodel Interaction (ICMI ’24) . ACM, 596–601. doi:10.1145/ 3678957.3678959

2024
[24]

María Antonia Paz, Julio Montero-Díaz, and Alicia Moreno-Delgado. 2020. Hate speech: A systematized review. Sage Open 10, 4 (2020), 2158244020973022

2020
[25]

Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. Robust Speech Recognition via Large-Scale Weak Supervision. doi:10.48550/ARXIV.2212.04356

2022 doi arXiv
[26]

Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust speech recognition via large-scale weak supervision. In International conference on machine learning . PMLR, 28492–28518

2023
[27]

Hannah Rashkin, Eric Michael Smith, Margaret Li, and Y-Lan Boureau. 2019. Towards Empathetic Open-domain Conversation Models: a New Benchmark and Dataset. arXiv:1811.00207 [cs.CL] https://arxiv.org/abs/1811.00207

2019 arXiv
[28]

Qwen Technical Report. 2023. Qwen Technical Report. arXiv preprint arXiv:2309.16609 (2023)

2023 arXiv
[29]

Naquee Rizwan, Paramananda Bhaskar, Mithun Das, Swadhin Satyaprakash Majhi, Punyajoy Saha, and Animesh Mukherjee. 2025. Exploring the Limits of Zero Shot Vision Language Models for Hate Meme Detection: The Vulnerabilities and their Interpretations. In Proceedings of the Intern...

2025
[30]

Naquee Rizwan, Seid Muhie Yimam, Daryna Dementieva, Dr. Florian Skupin, Tim Fischer, Daniil Moskovskiy, Aarushi Ajay Borkar, Robert Geislinger, Pun- yajoy Saha, Sarthak Roy, Martin Semmann, Alexander Panchenko, Chris Bie- mann, and Animesh Mukherjee. 2025. HatePRISM: Policies,...

2025
[31]

Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2020. DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. arXiv:1910.01108 [cs.CL] https://arxiv.org/abs/1910.01108

2020 arXiv
[32]

Charles Truong, Laurent Oudre, and Nicolas Vayatis. 2020. Selective review of offline change point detection methods. Signal Processing 167 (2020), 107299. doi:10.1016/j.sigpro.2019.107299

2020
[33]

Gomez, Lukasz Kaiser, and Illia Polosukhin

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2023. Attention Is All You Need. arXiv:1706.03762 [cs.CL] https://arxiv.org/abs/1706.03762

2023 arXiv
[34]

Valerie Wirtschafter. 2021. The challenge of detecting misinformation in pod- casting. Brookings (25 August 2021). https://www.brookings.edu/articles/the- challenge-of-detecting-misinformation-in-podcasting/ Accessed: 2024-12-02

2021
[35]

Valerie Wirtschafter. 2023. Audible reckoning: How top political pod- casters spread unsubstantiated and false claims. Brookings (March 2023). https://www.brookings.edu/articles/audible-reckoning-how-top- political-podcasters-spread-unsubstantiated-and-false-claims/ Accessed: ...

2023
[36]

role" : “system

Xinghua Zhang, Haiyang Yu, Yongbin Li, Minzheng Wang, Longze Chen, and Fei Huang. 2024. The Imperative of Conversation Analysis in the Era of LLMs: A Survey of Tasks, Techniques, and Trends. arXiv preprint arXiv:2409.14195 (2024). A Limitations As future work, we plan to exten...

2024 arXiv
[430]

doi:10.1111/comt.12096 arXiv:https://academic.oup.com/ct/article- pdf/26/4/410/21973874/jcomthe0410.pdf

doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [1]

Qwen2 Technical Report

2024. Qwen2 Technical Report. (2024)

2024

[2] [2]

Naveen Badathala, Abisek Rajakumar Kalarani, Tejpalsingh Siledar, and Pushpak Bhattacharyya. 2023. A Match Made in Heaven: A Multi-task Framework for Hyperbole and Metaphor Detection. arXiv:2305.17480 [cs.CL] https://arxiv.org/ abs/2305.17480

arXiv 2023

[3] [3]

Utkucan Balci, Jay Patel, Berkan Balci, and Jeremy Blackburn. 2024. Podcast Outcasts: Understanding Rumble’s Podcast Dynamics. arXiv:2406.14460 [cs.SI] https://arxiv.org/abs/2406.14460

work page Pith review arXiv 2024

[4] [4]

Doug Beeferman, William Brannon, and Deb Roy. 2019. Radiotalk: A large-scale corpus of talk radio transcripts. arXiv preprint arXiv:1907.07073 (2019)

work page Pith review arXiv 2019

[5] [5]

Hervé Bredin and Antoine Laurent. 2021. End-to-end speaker segmentation for overlap-aware resegmentation. arXiv preprint arXiv:2104.04045 (2021)

arXiv 2021

[6] [6]

Konijn, and Gerard J

Christian Burgers, Elly A. Konijn, and Gerard J. Steen. 2016. Fig- urative Framing: Shaping Public Discourse Through Metaphor, Hy- perbole, and Irony. Communication Theory 26, 4 (04 2016), 410–

work page 2016

[7] [7]

Yunfei Chu, Jin Xu, Qian Yang, Haojie Wei, Xipin Wei, Zhifang Guo, Yichong Leng, Yuanjun Lv, Jinzheng He, Junyang Lin, et al. 2024. Qwen2-audio technical report. arXiv preprint arXiv:2407.10759 (2024)

arXiv 2024

[8] [8]

Ann Clifton, Sravana Reddy, Yongze Yu, Aasish Pappu, Rezvaneh Rezapour, Hamed Bonab, Maria Eskevich, Gareth Jones, Jussi Karlgren, Ben Carterette, and Rosie Jones. 2020. 100,000 Podcasts: A Spoken English Document Corpus. In Proceedings of the 28th International Conference on Computational Linguistics , Donia Scott, Nuria Bel, and Chengqing Zong (Eds.). I...

work page 2020