The pilot corpus contains a mix of real meetings (which would have occurred anyway) and scenario-driven meetings to provide a broad coverage of different meeting styles as covered in the ontology below. This page contains information on the scenarii we employed to elicit various forms of naturally interactive speech in meeting settings. First we describe our ontology for meeting types and then describe the types of meetings we employed in collecting our pilot corpus and how they fit into the ontology.

Meeting Ontology

Meetings span a broad spectrum of informational (seminars, presentations) and collaborative (brainstorming, design, report-generation) activities and can range from formal (judicial, legislative proceedings) to informal (ad hoc group meetings). The different meeting types can be described by a set of characteristics describing the group and task dynamics.

A meeting scenario includes a definition of the characteristics of the group and describes what the participants should (try to) do (Methodology for evaluation of collaboration systems; John Cugini, et al.).

Meeting types are often broken down into their constituent work tasks. Work tasks (or collaborative activities) are the things people do, or wish to do, collaboratively. The following are work task descriptions (a subset of work tasks from group research by J.E. McGrath, 1984, "Groups: Interaction and Performance", Englewood Cliffs, N.J., Prentice Hall) which are appropriate for describing meeting scenarii:

Planning tasks:

Group members are given a goal and asked to develop a written plan for carrying out the steps needed to reach that goal.

Brainstorming and group creativity:

Members of a group are given a particular topic area and asked to brainstorm on ideas.

Negotiation:

Group members disagree on an issue, but an outcome has to be reached. Trade offs have to be made in multiple dimensions.

Decision making:

Group members are asked to develop consensus on issues that do not have correct answers, or make decisions from available information that is imperfectly correlated with criterion.

Competitive performance:

Groups are competing against each other. The goal of each group is to win over the other group.

Dissemination of information:

The task is to disseminate information among members of the group. Information can be distributed by the group members sharing information with each other, or a superior sharing information with others in the group. There may or may not be a question and answer session.

These tasks should be thought of as a continuum, not discrete tasks. So, a complete meeting can contain several of these types of tasks.

Pilot Corpus Meeting Types

We have selected a set of meeting types that involve activities and goals, and which will elicit natural interactions and discussions. We have also selected them so as to elicit different types of meeting behaviors as enumerated above.

Some meetings are simulated: participants are asked to act as if the goal was real but the acted out scenario is just a pretext to introduce diversity in the data. In order to maximize the naturalness and interactivity of the discussion, we have selected scenarii which will be challenging and interesting for the participants. As such, we have not employed some of the more formal meeting types. Other meetings are real: participants are really trying to accomplish the meeting goal as part of their everyday responsibilities.

The following list shows the meeting types that were used for the Pilot Corpus. Each meeting type has a title, an indication (next to the title) of whether the meetings is real or simulated, a list of task types, as defined in the ontology, that it covers followed by an example of an actual instance of the meeting.

1. Plan a party real

   
   Task types:
   

   planning

   
   Example:
   

   Participants will plan an office party. Participants will decide the date and plan different tasks that each member will carry out, complete with timelines for doing so and different coordination points. A moderator will run the meeting and 5-7 people will attend the meeting. The meeting is expected to take 45 minutes. Additional meetings will be scheduled and recorded if needed. No domain expertise is required.

2. Executive summary generation simulated

   
   Task types:
   

   dissemination of information, decision making

   
   Example:
   

   Participants are asked to take a look at the recent events in the news in order to produce an executive summary. Participants are shown a document describing the task to be performed and the constraints the participants must work within (imparted time, number of news items, summary format, etc.).

3. Interactive game playing simulated

   
   Task types:
   

   competitive performance, dissemination of information

   
   Example:
   

   Players of Once upon a Time create a story together using cards that show typical elements from fairy tales. The Storyteller creates astory using the ingredients on her cards, guiding the plot toward her own ending. The other players use their cards to interrupt her and to become the new Storyteller. The winner is the first player to play out all her cards, ending with the "Happy Ever After" card.

4. Staff meeting real

   
   Task types:
   

   planning, negotiation, brainstorming, decision-making and dissemination of information

   
   Example:
   

   Participants give a briefing about their progress, discuss a technical topic, brainstorm ideas and make decisions and/or plan their activities. The number of participants may vary from five to eight. The meeting is expected to take 60 minutes. Several sessions of this meeting type will be recorded. The same and/or different people will participate in different sessions.

5. Technical presentation real

   
   Task types:
   

   dissemination of information

   
   Example:
   

   A lecturer gives a presentation on a technical topic. The presentation might be followed by a question answering session. Several sessions of this meeting type will be recorded. The same and/or different people will participate in different sessions.

6. Interaction with an expert simulated

   
   Task types:
   

   decision-making and dissemination of information

   
   Example:
   

   Participants will decide on how to furnish their office. An office design expert will be present to assist the participants. Participants shall decide on the type, color and arrangement of the desk, chair, file cabinets, bookcases and carpets for their office. There will be 5 to 8 participants in these meetings. The meeting is expected to take 45 minutes. Several sessions of this scenario will be recorded. Different people will enact different sessions. No domain expertise is required.

7. Focus group discussion simulated

   
   Task types:
   

   dissemination of information

   
   Example:
   

   Round table discussions. A group of 5-8 participants will discuss one of the current events topics listed below. A moderator will run the meeting. Expert subjects may be invited to start the discussion with a brief presentation. Books, magazines, pictures and video clips would be available to assist in the discussion. Participants are encouraged to read a handout and/or watch a videotape before attending the meeting. The meeting is expected to take 60 minutes. Several sessions of this scenario with different topics will be recorded. The same people may participate in discussions of different topics. No domain expertise is required. However, participants are assumed to have a general knowledge of the topics being discussed. The following are some possible topics for focus group discussions:

   • Bioterrorism: Experts say the U.S. is ill prepared for bioterrorism, the deliberate release of disease-casing germs to kill large numbers of people. Are you worried? What can be done to fight it? Join the debate.
   • Public schools: How good are our schools? How safe are they?
   • Vacations: Where do you plan on going on vacations? What do you intend to do?
   • Email: How has email changed your life? Has it helped you to be more productive or less productive? What email software do you prefer? Are you happy with current email technology? What would you change? Do you still correspond in postal mail?
   • Holidays: How and where will you celebrate the holidays this year? Do you enjoy the holidays? What does the coming New Year mean to you? How will you celebrate it?
   • Movies: What movie genres do you prefer? What is your favorite movie of all time? What was the last movie you saw? Did you enjoy it? Who is your favorite actor? What is the next movie you plan to see?
   • Cellphone ethics: Do you have a cell phone? If so, how long have you had one? How have cellphones affected your life? Do you think drivers should be permitted to use cellphones while they drive? Do you think hands-free devices make driving with a cellphone safer?