Understanding social interaction within groups is key to analyzing online communities. Most current work focuses on structural properties: who talks to whom, and how such interactions form larger network structures. The interactions themselves, however, generally take place in the form of natural language — either spoken or written — and one could reasonably suppose that signals manifested in language might also provide information about roles, status, and other aspects of the group’s dynamics. To date, however, finding domain-independent language-based signals has been a challenge. Here, we show that in group discussions, power differentials between participants are subtly revealed by how much one individual immediately echoes the linguistic style of the person they are responding to. Starting from this observation, we propose an analysis framework based on linguistic coordination that can be used to shed light on power relationships and that works consistently across multiple types of power — including a more “static ” form of power based on status differences, and a more “situational ” form of power in which one individual experiences a type of dependence on another. Using this framework, we study how conversational behavior can reveal power relationships in two very different settings: discussions among Wikipedians and arguments before the

When multiple conversations occur simultaneously, a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately. We refer to this task as disentanglement. We present a corpus of Internet Relay Chat dialogue in which the various conversations have been manually disentangled, and evaluate annotator reliability. We propose a graph-based clustering model for disentanglement, using lexical, timing, and discourse-based features. The model’s predicted disentanglements are highly correlated with manual annotations. We conclude by discussing two extensions to the model, specificity tuning and conversation start detection, both of which are promising but do not currently yield practical improvements. 1.

A well-written text follows an overall structure, with each sentence following naturally from the ones before and leading into the ones which come afterwards. We call this structure “coherence”; without it, a document becomes a confusing series of non sequiturs. Understanding the principles that make a text coherent is an important goal of natural language processing. These principles can be applied to the design of systems that create new documents, like summaries, or make changes to existing documents. Coherence is a universal principle of language, but typical approaches to evaluation focus on the application of multidocument summarization. We test the generality of our models by applying them to a new task, chat disentanglement, in which we distinguish independent conversational threads in a crowded chat room. To study this task, we create our own corpus and evaluation metrics, propose a baseline model with basic coherence features, and then test the performance of our own and others' more sophisticated models of local coherence. We present evidence that despite the significant differences between this task setting and conventional summarization-inspired evaluations, many of these models generalize fairly well, improving over the baseline. Problems with lexicalized models are mostly the fault of insufficient in-domain training data, rather than representing weaknesses in the models themselves. Thus we conclude that many of the same basic principles