Jean Carletta

Virtual team meetings: An analysis of communication and context

We report a simulation study of virtual team meetings. Participants role-played companies collaborating on a design problem while supported by a range of IT tools, such as videoconferencing and shared applications. Meetings were analysed to investigate how sharing computing facilities, operating the technology, and company status, influenced communications. Significantly more talk occurred in larger teams where participants shared I.T. facilities BUT this extra talk was restricted to talk within a single location. No extra talk was shared across the virtual team via the communications link. Where facilities were shared, technology controllers dominated cross-site talk. To encourage free communication across distributed virtual teams we recommend providing each participant with their own communications facility even if this is technologically less advanced than if technology support were shared.

An annotation scheme for discourse-level argumentation in research articles

In order to build robust automatic abstracting systems, there is a need for better training resources than are currently available. In this paper, we introduce an annotation scheme for scientific articles which can be used to build such a resource in a consistent way. The seven categories of the scheme are based on rhetorical moves of argumentation. Our experimental results show that the scheme is stable, reproducible and intuitive to use.

Group Discussion as Interactive Dialogue or as Serial Monologue: The Influence of Group Size

Current models draw a broad distinction between communication as dialogue and communication as monologue. The two kinds of models have different implications for who influences whom in a group discussion. If the discussion is like interactive dialogue, group members should be influenced most by those with whom they interact in the discussion; if it is like serial monologue, they should be influenced most by the dominant speaker. The experiments reported here show that in small, 5-person groups, the communication is like dialogue and members are influenced most by those with whom they interact in the discussion. However, in large, 10-person groups, the communication is like monologue and members are influenced most by the dominant speaker. The difference in mode of communication is explained in terms of how speakers in the two sizes of groups design their utterances for different audiences.

The NITE XML Toolkit: Flexible annotation for multimodal language data

Multimodal corpora that show humans interacting via language are now relatively easy to collect. Current tools allow one either to apply sets of time-stamped codes to the data and consider their timing and sequencing or to describe some specific linguistic structure that is present in the data, built over the top of some form of transcription. To further our understanding of human communication, the research community needs code sets with both timings and structure, designed flexibly to address the research questions at hand. The NITE XML Toolkit offers library support that software developers can call upon when writing tools for such code sets and, thus, enables richer analyses than have previously been possible. It includes data handling, a query language containing both structural and temporal constructs, components that can be used to build graphical interfaces, sample programs that demonstrate how to use the libraries, a tool for running queries, and an experimental engine that builds interfaces on the basis of declarative specifications.

A shallow model of backchannel continuers in spoken dialogue

Spoken dialogue systems would be more acceptable if they were able to produce backchannel continuers such as mm-hmm in naturalistic locations during the users utterances. Using the HCRC Map Task Corpus as our data source, we describe models for predicting these locations using only limited processing and features of the users speech that are commonly available, and which therefore could be used as a low-cost improvement for current systems. The baseline model inserts continuers after a predetermined number of words. One further model correlates back-channel continuers with pause duration, while a second predicts their occurrence using trigram POS frequencies. Combining these two models gives the best results.

Incorporating Speaker and Discourse Features into Speech Summarization

We have explored the usefulness of incorporating speech and discourse features in an automatic speech summarization system applied to meeting recordings from the ICSI Meetings corpus. By analyzing speaker activity, turn-taking and discourse cues, we hypothesize that such a system can outperform solely text-based methods inherited from the field of text summarization. The summarization methods are described, two evaluation methods are applied and compared, and the results clearly show that utilizing such features is advantageous and efficient. Even simple methods relying on discourse cues and speaker activity can outperform text summarization approaches.

Animacy encoding in English: why and how

We report on two recent medium-scale initiatives annotating present day English corpora for animacy distinctions. We discuss the relevance of animacy for computational linguistics, specifically generation, the annotation categories used in the two studies and the interannotator reliability for one of the studies.

Risk-taking and recovery in task-oriented dialogue

The Principle of Parsimony states that people usually try to complete tasks with the least effort that will produce a satisfactory solution. In task-oriented dialogue, this produces a tension between conveying information carefully to the partner and leaving it to be inferred, risking a misunderstanding and the need for recovery. Using natural dialogue examples, primarily from the HCRC Map Task, we apply the Principle of Parsimony to a range of information types and identify a set of applicable recovery strategies. We argue that risk-taking and recovery are crucial for efficient dialogue because they pinpoint which information must be transferred and allow control of the interaction to switch to the participant who can best guide the course of the dialogue.

The effects of multimedia communication technology on non-collocated teams: a case study

Collaborative teams are becoming increasingly important for industry, both within and across companies. There is a need for communication technology to support teams because many teams are non-collocated, or ‘virtual’. Two automotive supply chain teams were observed while they were experimenting with multimedia conferencing in order to determine what support non-collocated teams need and the potential effects of introducing technologies on their group processes. The observations included meeting recordings and other sources that show the organizational factors affecting teams. Working in teams requires very close collaboration. Communication technology can help teams if it is used to foster close and relatively informal person-to-person interaction. Organizational constraints on how the technology is introduced favour high-technology, specialpurpose installations, but teams can best be supported using relatively modest equipment with desktop access.

Reliability measurement without limits

In computational linguistics, a reliability measurement of 0.8 on some statistic such as is widely thought to guarantee that hand-coded data is fit for purpose, with 0.67 to 0.8 tolerable, and lower values suspect. We demonstrate that the main use of such data, machine learning, can tolerate data with low reliability as long as any disagreement among human coders looks like random noise. When the disagreement introduces patterns, however, the machine learner can pick these up just like it picks up the real patterns in the data, making the performance figures look better than they really are. For the range of reliability measures that the field currently accepts, disagreement can appreciably inflate performance figures, and even a measure of 0.8 does not guarantee that what looks like good performance really is. Although this is a commonsense result, it has implications for how we work. At the very least, computational linguists should look for any patterns in the disagreement among coders and assess what impact they will have.