The reviewed record of science sign in
Pith

arxiv: 2605.30668 · v1 · pith:WMFXZCVO · submitted 2026-05-29 · cs.CL · cs.AI

CobSeg: Coherence Boundary Modeling for Dialogue Topic Segmentation

Sijin Sun , Liangbin Zhao , Jiaxiang Cai , Ming Deng , Mingyu Luo , Xiuju Fu This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-06-28 23:12 UTCgrok-4.3pith:WMFXZCVOrecord.jsonopen to challenge →

classification cs.CL cs.AI
keywords dialogue topic segmentationcoherence boundary modelingmulti-branch architecturelexical transitionssemantic continuityboundary predictionP_k metricW_d metric
0
0 comments X

The pith

CobSeg improves dialogue topic segmentation by separating coherence-level semantic continuity from lexical boundary transitions via directional prediction and weighting.

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

The paper establishes a method for dialogue topic segmentation that explicitly models two distinct kinds of boundary signals: semantic continuity across utterances and lexical shifts near utterance edges. It does so with a multi-branch architecture that predicts boundaries directionally, weights positions by their informativeness, and blends in a corpus-derived coherence cue through learned weights. The approach is tested both with gold boundaries and with automatically induced ones, showing metric gains on five benchmarks without requiring large language models at inference time. A sympathetic reader would care because many human-AI systems rely on accurate topic tracking in conversations, and the method keeps the model compact and trainable.

Core claim

CobSeg is a compact trainable segmenter that recovers both semantic discontinuities and lexical transitions through a multi-branch architecture with directional boundary prediction, boundary informativeness weighting, and a corpus-derived topic coherence cue, achieving lower P_k and W_d errors on five benchmarks under both gold and induced boundary supervision.

What carries the argument

Multi-branch architecture that separates coherence-level semantic continuity from lexical boundary transitions and recovers both through directional boundary prediction plus boundary informativeness weighting.

If this is right

  • Reduces P_k by 0.7 points and W_d by 0.6 points on VHF under gold supervision.
  • Reaches P_k of 1.0 on DialSeg711 with gold boundaries.
  • With induced boundaries, reduces P_k by 14.8 points on VHF, 1.5 on DialSeg711, and 1.1 on TIAGE.
  • Outperforms prior non-LLM approaches across the five benchmarks.
  • Enables enhanced boundary prediction without LLM calls during inference.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same separation of boundary types could be tested on document-level segmentation where lexical and semantic signals also compete.
  • Avoiding LLM calls at inference time could allow real-time topic tracking on edge devices.
  • The informativeness weighting may generalize to other sequence labeling tasks where some positions carry more boundary information than others.
  • Replacing the corpus-derived cue with signals from smaller models could be checked to see whether the gains persist.

Load-bearing premise

That separating coherence-level semantic continuity from lexical boundary transitions and recovering both via directional boundary prediction plus boundary informativeness weighting is what produces the reported gains.

What would settle it

Retraining the model on VHF after removing the directional boundary prediction branch while keeping all other components yields equal or higher P_k scores than the full model.

Figures

Figures reproduced from arXiv: 2605.30668 by Jiaxiang Cai, Liangbin Zhao, Ming Deng, Mingyu Luo, Sijin Sun, Xiuju Fu.

Figure 1
Figure 1. Figure 1: Overview Structure of CobSeg. in Vessel Traffic Services (VTS) and Air Traffic Control (ATC) systems. Recent work spans super￾vised models trained on gold boundaries (Koshorek et al., 2018; Jiang et al., 2023), unsupervised methods that induce boundaries from coherence patterns (Gao et al., 2023; Xing and Carenini, 2021), pseudo-label training with auxiliary signals (Artemiev et al., 2024), and LLM-based r… view at source ↗
Figure 2
Figure 2. Figure 2: Detailed architecture of CobSeg. et al., 2018), the Lexical Boundary Detector ap￾plies a temporally sensitive structure for token se￾quence modeling within the utterance. Bidirec￾tional context modeling allows the encoder to cap￾ture dependencies from both past and future tokens (Schuster and Paliwal, 1997). Let Enctok denote a bidirectional LSTM encoder that processes token sequences within each utterance… view at source ↗
Figure 3
Figure 3. Figure 3: Attribution at the token level and informativeness at the utterance level on a VHF example. Lexical cues [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of utterance embeddings with [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
read the original abstract

Dialogue topic segmentation is critical in many human-AI collaborative applications which requires identifying heterogeneous boundary cues, including lexical transitions near utterance edges and semantic discontinuities across utterances. Existing utterance models often dilute these local lexical signals. We propose CobSeg, a novel multi-branch architecture that separates coherence-level semantic continuity from lexical boundary transitions and recovers both through directional boundary prediction. CobSeg further uses boundary informativeness weighting to emphasize high-utility utterance positions, and incorporates a corpus-derived topic coherence cue with learned combination weights. While CobSeg is evaluated as a compact trainable segmenter under supervised gold-boundary training and a pseudo-label setting with automatically induced boundaries, it performs enhanced boundary prediction without LLM calls during inference. Across five benchmarks, it improves $P_k$ and $W_d$ particularly when local lexical cues are prominent: under gold supervision, it reduces $P_k$ by 0.7 points and $W_d$ by 0.6 points on VHF, and reaches $P_k$ of 1.0 on DialSeg711; with induced boundaries, it reduces $P_k$ by 14.8 points on VHF, by 1.5 points on DialSeg711, and by 1.1 points on TIAGE, outperforming prior non-LLM approaches.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The paper proposes CobSeg, a multi-branch architecture for dialogue topic segmentation that separates coherence-level semantic continuity from lexical boundary transitions via directional boundary prediction, incorporates boundary informativeness weighting, and combines it with a corpus-derived topic coherence cue using learned weights. It is evaluated as a compact trainable segmenter under both gold-boundary supervision and pseudo-label settings with induced boundaries, claiming to avoid LLM calls at inference while improving P_k and W_d on five benchmarks (particularly VHF, DialSeg711, and TIAGE) when local lexical cues are prominent.

Significance. If the reported gains hold under proper controls, the work would demonstrate a practical non-LLM alternative for modeling heterogeneous boundary cues in dialogue, with potential utility in efficient human-AI collaborative systems. The explicit separation of cue types and the induced-boundary results (e.g., large P_k reductions on VHF) address a known limitation of utterance-level models, though the necessity of the multi-branch design remains unverified in the provided text.

major comments (1)
  1. [Abstract] Abstract: the central claim that directional boundary prediction plus informativeness weighting recovers both cue types and drives the reported gains (14.8 P_k reduction on VHF, 1.5 on DialSeg711 under induced boundaries) cannot be evaluated, as the abstract supplies no experimental details, baseline descriptions, ablation results, or statistical tests; this renders the modeling premise (separation of coherence-level continuity from lexical transitions) untestable and load-bearing for attribution of improvements.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment on the abstract. We address the concern point-by-point below and agree that revisions are warranted to improve transparency.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that directional boundary prediction plus informativeness weighting recovers both cue types and drives the reported gains (14.8 P_k reduction on VHF, 1.5 on DialSeg711 under induced boundaries) cannot be evaluated, as the abstract supplies no experimental details, baseline descriptions, ablation results, or statistical tests; this renders the modeling premise (separation of coherence-level continuity from lexical transitions) untestable and load-bearing for attribution of improvements.

    Authors: We acknowledge that the abstract, as a high-level summary, omits specific experimental details such as baseline names, ablation configurations, and statistical significance tests, which limits immediate verifiability of the attribution. The full manuscript provides these in Sections 4 (experimental setup, five benchmarks, gold vs. induced boundary protocols) and 5 (results tables comparing against prior non-LLM methods, component ablations isolating directional prediction and informativeness weighting). The reported P_k reductions are from direct metric comparisons on VHF, DialSeg711, and TIAGE. To strengthen the abstract without exceeding length constraints, we will revise it to briefly note the evaluation settings (gold and induced boundaries), the benchmarks used, and the key gains under induced boundaries. We will also add a short clause referencing the multi-branch separation of cues. This addresses the load-bearing concern while preserving the abstract's conciseness. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical architecture with metric-reported gains, no derivations or self-referential reductions

full rationale

The paper proposes a multi-branch neural architecture for dialogue segmentation and reports empirical improvements on five benchmarks under gold and induced boundary settings. No equations, first-principles derivations, or predictions appear in the provided text. Model components (directional boundary prediction, informativeness weighting, corpus-derived cue) are presented as design choices whose value is demonstrated by P_k and W_d deltas rather than by any reduction to fitted inputs or self-citations. The central modeling premise is therefore not shown to be equivalent to its inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the effectiveness of the multi-branch separation and weighting scheme introduced in the abstract; no free parameters are numerically specified, no new entities are postulated, and the only domain assumption is that heterogeneous boundary cues exist and can be disentangled.

free parameters (1)
  • learned combination weights
    The abstract states that the model incorporates a corpus-derived topic coherence cue with learned combination weights.
axioms (1)
  • domain assumption Dialogue topic segmentation requires identifying heterogeneous boundary cues including lexical transitions near utterance edges and semantic discontinuities across utterances.
    Explicitly stated as the motivation for the architecture in the abstract.

pith-pipeline@v0.9.1-grok · 5766 in / 1338 out tokens · 23917 ms · 2026-06-28T23:12:16.313505+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

59 extracted references · 9 canonical work pages · 3 internal anchors

  1. [1]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Aleksei Artemiev, Daniil Parinov, Alexey Grishanov, Ivan Borisov, Alexey Vasilev, Daniil Muravetskii, Aleksey Rezvykh, Aleksei Goncharov, and Andrey Savchenko. 2024. Leveraging summarization for unsupervised dialogue topic segmentation. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 4697--4704

    2024

  4. [4]

    Pinkesh Badjatiya, Litton J Kurisinkel, Manish Gupta, and Vasudeva Varma. 2018. Attention-based neural text segmentation. In European Conference on Information Retrieval, pages 180--193. Springer

    2018

  5. [5]

    Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2014. Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  6. [6]

    Doug Beeferman, Adam Berger, and John Lafferty. 1999. Statistical models for text segmentation. Machine learning, 34(1):177--210

    1999

  7. [7]

    Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Yunxuan Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, and 1 others. 2024. Scaling instruction-finetuned language models. Journal of Machine Learning Research, 25(70):1--53

    2024

  8. [8]

    Michael H Coen. 2025. When f1 fails: Granularity-aware evaluation for dialogue topic segmentation. arXiv preprint arXiv:2512.17083

    work page arXiv 2025

  9. [9]

    Sarkar Snigdha Sarathi Das, Chirag Shah, Mengting Wan, Jennifer Neville, Longqi Yang, Reid Andersen, Georg Buscher, and Tara Safavi. 2024. S3-dst: Structured open-domain dialogue segmentation and state tracking in the era of llms. In Findings of the Association for Computational Linguistics: ACL 2024, pages 14996--15014

    2024

  10. [10]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), pages 4171--4186

    2019

  11. [11]

    Lan Du, Wray Buntine, and Mark Johnson. 2013. Topic segmentation with a structured topic model. In Proceedings of the 2013 conference of the North American chapter of the Association for Computational Linguistics: Human language technologies, pages 190--200

    2013

  12. [12]

    Jacob Eisenstein and Regina Barzilay. 2008. Bayesian unsupervised topic segmentation. In Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, pages 334--343

    2008

  13. [13]

    Yaxin Fan, Feng Jiang, Peifeng Li, and Haizhou Li. 2024. Uncovering the potential of chatgpt for discourse analysis in dialogue: An empirical study. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 16998--17010

    2024

  14. [14]

    Song Feng, Hui Wan, Chulaka Gunasekara, Siva Patel, Sachindra Joshi, and Luis Lastras. 2020. doc2dial: A goal-oriented document-grounded dialogue dataset. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 8118--8128

    2020

  15. [15]

    Haoyu Gao, Rui Wang, Ting-En Lin, Yuchuan Wu, Min Yang, Fei Huang, and Yongbin Li. 2023. Unsupervised dialogue topic segmentation with topic-aware contrastive learning. In SIGIR '23, SIGIR '23, page 2481–2485, New York, NY, USA. Association for Computing Machinery

    2023

  16. [16]

    Tianyu Gao, Xingcheng Yao, and Danqi Chen. 2021. Simcse: Simple contrastive learning of sentence embeddings. In Proceedings of the 2021 conference on empirical methods in natural language processing, pages 6894--6910

    2021

  17. [17]

    Goran Glava s , Federico Nanni, and Simone Paolo Ponzetto. 2016. Unsupervised text segmentation using semantic relatedness graphs. In Proceedings of the Fifth Joint Conference on Lexical and Computational Semantics, pages 125--130

    2016

  18. [18]

    Zheng Gong, Shiwei Tong, Han Wu, Qi Liu, Hanqing Tao, Wei Huang, and Runlong Yu. 2022. Tipster: A topic-guided language model for topic-aware text segmentation. In International Conference on Database Systems for Advanced Applications, pages 213--221. Springer

    2022

  19. [19]

    Marti A Hearst. 1997. Text tiling: Segmenting text into multi-paragraph subtopic passages. Computational linguistics, 23(1):33--64

    1997

  20. [20]

    Xia Hou, Qifeng Li, and Tongliang Li. 2024. An unsupervised dialogue topic segmentation model based on utterance rewriting. arXiv preprint arXiv:2409.07672

    work page arXiv 2024

  21. [21]

    Yerin Hwang, Yongil Kim, Yunah Jang, Jeesoo Bang, Hyunkyung Bae, and Kyomin Jung. 2024. Mp2d: An automated topic shift dialogue generation framework leveraging knowledge graphs. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17682--17702

    2024

  22. [22]

    Hakan Inan, Rashi Rungta, and Yashar Mehdad. 2022. Structured summarization: Unified text segmentation and segment labeling as a generation task. arXiv preprint arXiv:2209.13759

    work page arXiv 2022

  23. [23]

    Junfeng Jiang, Chengzhang Dong, Sadao Kurohashi, and Akiko Aizawa. 2023. Superdialseg: A large-scale dataset for supervised dialogue segmentation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 4086--4101

    2023

  24. [24]

    Omri Koshorek, Adir Cohen, Noam Mor, Michael Rotman, and Jonathan Berant. 2018. Text segmentation as a supervised learning task. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers), pages 469--473

    2018

  25. [25]

    Seungmin Lee, Yongsang Yoo, Minhwa Jung, and Min Song. 2025. Def-dts: Deductive reasoning for open-domain dialogue topic segmentation. In Findings of the Association for Computational Linguistics: ACL 2025, pages 20736--20753

    2025

  26. [26]

    Jing Li, Aixin Sun, and Shafiq R Joty. 2018. Segbot: A generic neural text segmentation model with pointer network. In IJCAI, pages 4166--4172

    2018

  27. [27]

    Jiangyi Lin, Yaxin Fan, Xiaomin Chu, Peifeng Li, and Qiaoming Zhu. 2023. Multi-granularity prompts for topic shift detection in dialogue. In International Conference on Intelligent Computing, pages 511--522. Springer

    2023

  28. [28]

    Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  29. [29]

    Zhengyuan Liu, Siti Umairah Md Salleh, Hong Choon Oh, Pavitra Krishnaswamy, and Nancy Chen. 2023. Joint dialogue topic segmentation and categorization: A case study on clinical spoken conversations. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing: Industry Track, pages 185--193

    2023

  30. [30]

    Kelvin Lo, Yuan Jin, Weicong Tan, Ming Liu, Lan Du, and Wray Buntine. 2021. Transformer over pre-trained transformer for neural text segmentation with enhanced topic coherence. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 3334--3340

    2021

  31. [31]

    Yantong Lv, Wanjie Tao, Qun Dai, Zulong Chen, Quan Lu, and Ning Jiang. 2025. Dynamic topic segmentation in dialogues: Enhancing boundaries with topic-aware propagation. In Companion Proceedings of the ACM on Web Conference 2025, pages 1172--1175

    2025

  32. [32]

    Xinbei Ma, Yi Xu, Hai Zhao, and Zhuosheng Zhang. 2024. Multi-turn dialogue comprehension from a topic-aware perspective. Neurocomputing, 578:127385

    2024

  33. [33]

    Inderjeet Nair, Aparna Garimella, Balaji Vasan Srinivasan, Natwar Modani, Niyati Chhaya, Srikrishna Karanam, and Sumit Shekhar. 2023. A neural crf-based hierarchical approach for linear text segmentation. In Findings of the Association for Computational Linguistics: EACL 2023, pages 883--893

    2023

  34. [34]

    Seongmin Park, Jinkyu Seo, and Jihwa Lee. 2023. Unsupervised dialogue topic segmentation in hyperdimensional space. arXiv preprint arXiv:2308.10464

    work page arXiv 2023

  35. [35]

    Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. Glove: Global vectors for word representation. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pages 1532--1543

    2014

  36. [36]

    Lev Pevzner and Marti A Hearst. 2002. A critique and improvement of an evaluation metric for text segmentation. Computational Linguistics, 28(1):19--36

    2002

  37. [37]

    Hengfeng Pu and Liqing Wang. 2023. Dialogue segmentation based on dynamic context coherence. In Proceedings of the 2023 7th International Conference on Natural Language Processing and Information Retrieval, pages 190--195

    2023

  38. [38]

    MengNan Qi, Hao Liu, YuZhuo Fu, and Ting Liu. 2021. Improving abstractive dialogue summarization with hierarchical pretraining and topic segment. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 1121--1130

    2021

  39. [39]

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research, 21(140):1--67

    2020

  40. [40]

    Mike Schuster and Kuldip K Paliwal. 1997. Bidirectional recurrent neural networks. IEEE transactions on Signal Processing, 45(11):2673--2681

    1997

  41. [41]

    Karin Sevegnani, David M Howcroft, Ioannis Konstas, and Verena Rieser. 2021. Otters: One-turn topic transitions for open-domain dialogue. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 2492--2504

    2021

  42. [42]

    Alessandro Solbiati, Kevin Heffernan, Georgios Damaskinos, Shivani Poddar, Shubham Modi, and Jacques Cali. 2021. Unsupervised topic segmentation of meetings with bert embeddings. arXiv preprint arXiv:2106.12978

    work page Pith review arXiv 2021

  43. [43]

    Swapna Somasundaran and 1 others. 2020. Two-level transformer and auxiliary coherence modeling for improved text segmentation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 7797--7804

    2020

  44. [44]

    Yiping Song, Lili Mou, Rui Yan, Li Yi, Zinan Zhu, Xiaohua Hu, and Ming Zhang. 2016. Dialogue session segmentation by embedding-enhanced texttiling. arXiv preprint arXiv:1610.03955

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  45. [45]

    Sijin Sun, Liangbin Zhao, Ming Deng, and Xiuju Fu. 2025. Dash: Dialogue-aware similarity and handshake recognition for topic segmentation in public-channel conversations. arXiv preprint arXiv:2512.15042

    work page arXiv 2025

  46. [46]

    Ryuichi Takanobu, Minlie Huang, Zhongzhou Zhao, Feng-Lin Li, Haiqing Chen, Xiaoyan Zhu, Liqiang Nie, and 1 others. 2018. A weakly supervised method for topic segmentation and labeling in goal-oriented dialogues via reinforcement learning. In Ijcai, pages 4403--4410

    2018

  47. [47]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems, 30

    2017

  48. [48]

    Anvesh Rao Vijjini, Hanieh Deilamsalehy, Franck Dernoncourt, and Snigdha Chaturvedi. 2023. Curricular next conversation prediction pretraining for transcript segmentation. In Findings of the Association for Computational Linguistics: EACL 2023, pages 2597--2607

    2023

  49. [49]

    Weishi Wang, Steven CH Hoi, and Shafiq Joty. 2020. Response selection for multi-party conversations with dynamic topic tracking. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6581--6591

    2020

  50. [50]

    Chien-Sheng Wu, Steven CH Hoi, Richard Socher, and Caiming Xiong. 2020. Tod-bert: Pre-trained natural language understanding for task-oriented dialogue. In Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP), pages 917--929

    2020

  51. [51]

    Jinxiong Xia, Cao Liu, Jiansong Chen, Yuchen Li, Fan Yang, Xunliang Cai, Guanglu Wan, and Houfeng Wang. 2022. Dialogue topic segmentation via parallel extraction network with neighbor smoothing. In Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 2126--2131

    2022

  52. [52]

    Huiyuan Xie, Zhenghao Liu, Chenyan Xiong, Zhiyuan Liu, and Ann Copestake. 2021. Tiage: A benchmark for topic-shift aware dialog modeling. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 1684--1690

    2021

  53. [53]

    Linzi Xing and Giuseppe Carenini. 2021. Improving unsupervised dialogue topic segmentation with utterance-pair coherence scoring. In Proceedings of the 22nd annual meeting of the Special Interest Group on discourse and dialogue, pages 167--177

    2021

  54. [54]

    Jiahui Xu, Feng Jiang, Anningzhe Gao, Luis Fernando D'Haro, and Haizhou Li. 2025. Unsupervised mutual learning of discourse parsing and topic segmentation in dialogue. IEEE Transactions on Audio, Speech and Language Processing

    2025

  55. [55]

    Jun Xu, Zeyang Lei, Haifeng Wang, Zheng-Yu Niu, Hua Wu, and Wanxiang Che. 2021 a . Discovering dialog structure graph for coherent dialog generation. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1726--1739

    2021

  56. [56]

    Peng Xu, Hamidreza Saghir, Jin Sung Kang, Teng Long, Avishek Joey Bose, Yanshuai Cao, and Jackie Chi Kit Cheung. 2019. A cross-domain transferable neural coherence model. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 678--687

    2019

  57. [57]

    Yi Xu, Hai Zhao, and Zhuosheng Zhang. 2021 b . Topic-aware multi-turn dialogue modeling. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 14176--14184

    2021

  58. [58]

    Chenxu Yang, Zheng Lin, Jiangnan Li, Fandong Meng, Weiping Wang, Lanrui Wang, and Jie Zhou. 2022. Take: Topic-shift aware knowledge selection for dialogue generation. In Proceedings of the 29th International Conference on Computational Linguistics, pages 253--265

    2022

  59. [59]

    Shihao Yang, Ziyi Zhang, Yue Jiang, Chunsheng Qin, and Shuhua Liu. 2025. A unified supervised and unsupervised dialogue topic segmentation framework based on utterance pair modeling. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Pap...

    2025