AutoBrief is an experimental system that automatically creates interactive presentations in coordinated text and information graphics. The current prototype is implemented in the domain of transportation scheduling to assist human transportation analysts using an incremental scheduling system (DITOPS). AutoBrief acts as an intelligent assistant providing high-level briefings about the DITOPS schedules. The briefings summmarize the schedules, analyze possible problems in them, and suggest ways to address the problems.

Navigational links in the presentation enable the analyst to request more detailed information. Also, implemented in a VISAGE environment, AutoBrief supports an information-centric approach. For example, the analyst can drag highlighted text or elements of a graphic from AutoBrief to other parts of the environment, e.g., to control DITOPS, to populate a user-created graphic for data exploration, or to create a personalized briefing.

Internally, AutoBrief has two main components, the Schedule Analyzer and the Presentation Generator. The Schedule Analyzer, which uses knowledge about the transportation scheduling domain, stores analyses of properties of the schedules in a database for use by the Presentation Generator. The primary focus of our research, the domain-independent Presentation Generator is intended to be portable to other domains. It integrates natural language explanation generation with task-based graphic design.

The Presentation Generator employs a modified pipeline architecture. In the first stage, the content and organization of a response to the user's information request is planned by the Planner. The Planner employs a library of presentation plan operators, which are instantiated with data and concepts from the database, to create a plan for the presentation. The lower level of this hierarchical plan consists of communicative acts.

Next, the Media Allocation component decides which communicative acts are to be expressed in text and/or graphics. Then two media generators, the Text Generator and Graphics Generator, are invoked to realize the acts that have been assigned to them. In addition to this straight pipeline flow of control, the Graphics Generator may send requests to the Text Generator for realization of text to be used within a graphic.

The Text Generator is responsible for expressing communicative acts as coherent text. During the first phase of text generation, called text microplanning, the communicative acts of the plan are transformed into a sequence of abstract sentence specifications. The text microplanner must select content words and syntactic forms to perform the communicative acts and decide which acts to perform in which sentences. In the second phase, called text realization, the abstract sentence specifications are converted to grammatical word sequences by a general-purpose English sentence generator (FUF/Surge).

Graphics generation is also divided into microplanning and realization phases. The graphics microplanner interprets and aggregates communicative actions and re-represents them as conceptual tasks that the user will need to perform with the graphic in order to achieve the effect of the action. This task representation is then fed into the Sage designer, which employs task-based design knowledge to produce a graphic that supports the tasks.