Disentangling Chat

Elsner, Micha; Charniak, Eugene

doi:10.1162/coli_a_00003

Cited by 67 publications

(76 citation statements)

References 18 publications

(38 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent work on synchronous conversations has been focusing on disentangling multi-party chats, which have a linear structure. For example, several studies propose models to disentangle multi-party chat (Elsner & Charniak, 2010Wang & Oard, 2009;Mayfield, Adamson, & Rosé, 2012). On the other hand, asynchronous conversations like email and social media services (e.g., Gmail, Twitter) generally organize comments into tree-structured threads using headers.…”

Section: Conversational Structure Extractionmentioning

confidence: 99%

“…Our supervised model is built on the graph-theoretic framework which has been used in many NLP tasks, including coreference resolution (Soon, Ng, & Lim, 2001) and chat disentanglement (Elsner & Charniak, 2010). This method works in two steps.…”

Section: Proposed Supervised Modelmentioning

confidence: 99%

“…If y's comment is same as or reply to x's comment, then it is likely that the two sentences talk about the same topic. Participants sometimes mention each other's name in multi-party conversations to make disentanglement easier (Elsner & Charniak, 2010). We also use this as a feature in our supervised segmentation model.…”

Section: Gapmentioning

confidence: 99%

“…Therefore, the standard metrics widely used in sequential topic segmentation in monolog and synchronous dialog, such as the P k (Beeferman, Berger, & Lafferty, 1999) and W indowDif f (W D) (Pevzner & Hearst, 2002), are also not applicable. Rather, the one-to-one and local agreement metrics described by Elsner and Charniak (2010) are more appropriate for our segmentation task. The one-to-one metric measures global agreement between two annotations by pairing up topical segments from the two annotations in a way (i.e., by computing the optimal max-weight bipartite matching) that maximizes the total overlap, and then reports the percentage of overlap.…”

Section: Metrics For Topic Segmentationmentioning

confidence: 99%

“…In the naturally occurring ICSI multi-party meetings (Janin et al 2003), Galley et al (2003) report an average of 7.5 topical segments per conversation. In multi-party chat, Elsner and Charniak (2010) report an average of 2.75 discussions active at a time. In the email and blog corpora, that we present in this article, annotators found an average of 2.5 and 10.77 topics per email and blog conversation, respectively.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Topic Segmentation and Labeling in Asynchronous Conversations

Joty

Carenini

2013

jair

View full text Add to dashboard Cite

Topic segmentation and labeling is often considered a prerequisite for higher-level conversation analysis and has been shown to be useful in many Natural Language Processing (NLP) applications. We present two new corpora of email and blog conversations annotated with topics, and evaluate annotator reliability for the segmentation and labeling tasks in these asynchronous conversations. We propose a complete computational framework for topic segmentation and labeling in asynchronous conversations. Our approach extends state-of-the-art methods by considering a fine-grained structure of an asynchronous conversation, along with other conversational features by applying recent graph-based methods for NLP. For topic segmentation, we propose two novel unsupervised models that exploit the fine-grained conversational structure, and a novel graph-theoretic supervised model that combines lexical, conversational and topic features. For topic labeling, we propose two novel (unsupervised) random walk models that respectively capture conversation specific clues from two different sources: the leading sentences and the fine-grained conversational structure. Empirical evaluation shows that the segmentation and the labeling performed by our best models beat the state-of-the-art, and are highly correlated with human annotations.

show abstract

Section: Conversational Structure Extractionmentioning

confidence: 99%

Section: Proposed Supervised Modelmentioning

confidence: 99%

Section: Gapmentioning

confidence: 99%

Section: Metrics For Topic Segmentationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Topic Segmentation and Labeling in Asynchronous Conversations

Joty

Carenini

2013

jair

View full text Add to dashboard Cite

show abstract

Mass Collaboration on the Web: Textual Content Analysis by Means of Natural Language Processing

Habernal

Daxenberger

Gurevych

2016

Mass Collaboration and Education

View full text Add to dashboard Cite

This chapter describes perspectives for utilizing natural language processing (NLP) to analyze artifacts arising from mass collaboration on the web. In recent years, the amount of user-generated content on the web has grown drastically. This content is typically noisy, un-or at best semi-structured, so that traditional analysis tools cannot properly handle it. To discover linguistic structures in this data, manual analysis is not feasible due to the large quantities of data. In this chapter, we explain and analyze web-based resources of mass collaboration, namely wikis, web forums, debate platforms and blog comments. We introduce recent advances and ongoing efforts to analyze textual content in two of these resources with the help of NLP. This includes an approach to discover flows of knowledge in online mass collaboration as well as methods to mine argumentative structures in natural language text. Finally, we outline application scenarios of the previously discussed techniques and resources within the domain of education.

show abstract