7 Hits in 2.5 sec

Dialogue Session Segmentation by Embedding-Enhanced TextTiling [article]

Yiping Song, Lili Mou, Rui Yan, Li Yi, Zinan Zhu, Xiaohua Hu, Ming Zhang
2016 arXiv   pre-print
We propose an embedding-enhanced TextTiling approach, inspired by the observation that conversation utterances are highly noisy, and that word embeddings provide a robust way of capturing semantics.  ...  However, it is unwise to track all previous utterances in the current session as not all of them are equally important. In this paper, we address the problem of session segmentation.  ...  In our study, we prefer the simple yet effective TextTiling approach for open-domain dialogue session segmentation, but enhance it with modern advances of word embeddings, which are robust in capturing  ... 
arXiv:1610.03955v1 fatcat:khvjezh2fzezlmhhlzbeecs3wa

Improving Unsupervised Dialogue Topic Segmentation with Utterance-Pair Coherence Scoring [article]

Linzi Xing, Giuseppe Carenini
2021 arXiv   pre-print
Dialogue topic segmentation is critical in several dialogue modeling problems. However, popular unsupervised approaches only exploit surface features in assessing topical coherence among utterances.  ...  In this work, we address this limitation by leveraging supervisory signals from the utterance-pair coherence scoring task.  ...  This research was supported by the Language & Speech Innovation Lab of Cloud BU, Huawei Technologies Co., Ltd.  ... 
arXiv:2106.06719v1 fatcat:fwutddbpw5ej3pvuhzliegkumm

A Weakly Supervised Method for Topic Segmentation and Labeling in Goal-oriented Dialogues via Reinforcement Learning

Ryuichi Takanobu, Minlie Huang, Zhongzhou Zhao, Fenglin Li, Haiqing Chen, Xiaoyan Zhu, Liqiang Nie
2018 Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence  
We propose a reinforcement learning (RL) method for topic segmentation and labeling in goal-oriented dialogues, which aims to detect topic boundaries among dialogue utterances and assign topic labels to  ...  Topic structure analysis plays a pivotal role in dialogue understanding.  ...  Embedding enhanced TextTiling works much better since word embedding can well capture the semantic similarity.  ... 
doi:10.24963/ijcai.2018/612 dblp:conf/ijcai/TakanobuHZLCZN18 fatcat:xh7266632jfrjbsfoa67pdfimu

Experiments with interactive question-answering

Sanda Harabagiu, Andrew Hickl, John Lehmann, Dan Moldovan
2005 Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics - ACL '05  
questions into the context of a Q/A dialogue.  ...  We present experimental results from large user studies (featuring a fully-implemented interactive Q/A system named FERRET) that demonstrates that surprising performance is achieved by integrating predictive  ...  As with Approach 2, segments were produced using the TextTiling algorithm.  ... 
doi:10.3115/1219840.1219866 dblp:conf/acl/HarabagiuHLM05 fatcat:7fcbldwgszfahpcnitjley2dfe

Message from the general chair

Benjamin C. Lee
2015 2015 IEEE International Symposium on Performance Analysis of Systems and Software (ISPASS)  
Our end system outperforms the state-of-the-art baseline by 2 B 3 F1 points on non-transcript portion of the ACE 2004 dataset.  ...  and polysemy by learning multiple embeddings per word.  ...  This algorithm is based on the well-known TextTiling algorithm, and segments documents using the Latent Dirichlet Allocation (LDA) topic model.  ... 
doi:10.1109/ispass.2015.7095776 dblp:conf/ispass/Lee15 fatcat:ehbed6nl6barfgs6pzwcvwxria

ACHI 2017 COMMITTEE ACHI Steering Committee

Leslie Miller, Alma Culén, Christos Bouras, Jan Broer, Marcin Butlewski, Jacek Chmielewski, Rolf Drechsler, Carlos Duarte, Sascha Fagel, Peter Forbrig, Toni Granollers, Maki Habib (+36 others)
ACKNOWLEDGMENTS ACKNOWLEDGMENT The authors would like to thank Günes ¸Karababa, C ¸agatay Koc ¸and Ege Sarıoglu for the fruitful discussion sessions and their invaluable ideas.  ...  This work was supported by JSPS KAKENHI Grant Number JP15K12093.  ...  In this system, we emphasized familiarity by embedding context, but it is not perfect.  ... 

Modeling Users' Information Needs in a Document Recommender for Meetings

Maryam Habibi
We also provided an example of the results obtained by our method and the baselines.  ...  We used sentence-aligned parallel corpora for training the model used by the Moses and document-aligned parallel corpora for learning multilingual topic models.  ...  The ASR transcripts or the subtitles of each lecture of the dataset are segmented using the TextTiling algorithm implemented in the NLTK toolkit (Bird, 2006) .  ... 
doi:10.5075/epfl-thesis-6760 fatcat:pq5bkum2hffrrp52frgdcxhwzu