Abstract. Open Mind Common Sense is a knowledge acquisition system designed to acquire commonsense knowledge from the general public over the web. We describe and evaluate our first fielded system, which enabled the construction of a 450,000 assertion commonsense knowledge base. We then discuss how our second-generation system addresses weaknesses discovered in the first. The new system acquires facts, descriptions, and stories by allowing participants to construct and fill in natural language templates. It employs word-sense disambiguation and methods of clarifying entered knowledge, analogical inference to provide feedback, and allows participants to validate knowledge and in turn each other. 1

In a hypermedia authoring task, an author often wants to set up meaningful connections between different media, such as text and photographs. To facilitate this task, it is helpful to have a software agent dynamically adapt the presentation of a media database to the user's authoring activities, and look for opportunities for annotation and retrieval. However, potential connections are often missed because of differences in vocabulary or semantic connections that are "obvious" to people but that might not be explicit. In a hypermedia authoring task, an author often wants to set up meaningful connections between different media, such as text and photographs. To facilitate this task, it is helpful to have a software agent dynamically adapt the presentation of a media database to the user's authoring activities, and look for opportunities for annotation and retrieval. However, potential connections are often missed because of differences in vocabulary or semantic connections that are "obvious" to people but that might not be explicit. ARIA (Annotation and Retrieval Integration Agent) is a software agent that acts an assistant to a user writing e-mail or Web pages. As the user types a story, it does continuous retrieval and ranking on a photo database. It can use descriptions in the story to semi-automatically annotate pictures. To improve the associations beyond simple keyword matching, we use natural language parsing techniques to extract important roles played by text, such as "who, what, where, when". Since many of the photos depict common everyday situations such as weddings or recitals, we use a common sense knowledge base, Open Mind, to fill in semantic gaps that might otherwise prevent successful associations.

■ A long-standing dream of artificial intelligence has been to put commonsense knowledge into computers—enabling machines to reason about everyday life. Some projects, such as Cyc, have begun to amass large collections of such knowledge. However, it is widely assumed that the use of common sense in interactive applications will remain impractical for years, until these collections can be considered sufficiently complete and commonsense reasoning sufficiently robust. Recently, at the Massachusetts Institute of Technology’s Media Laboratory, we have had some success in applying commonsense knowledge in a number of intelligent interface agents, despite the admittedly spotty coverage and unreliable inference of today’s

This paper introduces a model for producing common sense metadata during video capture and describes how this technique can have a positive impact on content capture, representation, and presentation. Metatada entered into the system at the moment of capture is used to generate suggestions designed to help the videographer decide what to shoot, how to compose a shot and how to index their video material to best support their communication requirements. An approach and first experiments using a common sense database and reasoning techniques to support a partnership between the camera and videographer during video capture are presented. 1.

www.cs.umd.edu/hcil We describe the design and use of a system facilitating the sharing of calendar information between remotely located, multi-generational family members. Most previous work in this area involves software enabling younger family members to monitor their parents. We have found, however, that older adults are equally if not more interested in the activities of younger family members. The major obstacle preventing them from participating in information sharing is the technology itself. Therefore, we developed a layered interface approach that offers simple interaction to older users. In our system, users can choose to enter information into a computerized calendar or write it by hand on digital paper calendars. All of the information is automatically shared among everyone in the distributed family. By making the interface more accessible to older uses, we promote symmetrical sharing of information among both older and younger family members. We present our participatory design process, describe the user interface and report on a field study in 3 households.

Everyday family life involves a myriad of mundane activities that need to be planned and coordinated. We describe findings from studies of 44 different families ’ coordination routines to understand how to best design technology to support them. We outline how a typology of calendars containing family activities is used by three different types of families—Monocentric, Pericentric, and Polycentric—which vary in the level of family involvement in the calendaring process. We describe these family types, the content of family calendars, the ways in which they are extended through annotations and augmentations, and the implications from these findings for design.

AI researchers have long dreamed of building systems with human-like 'common sense', the mental skills that most ordinary people share. But all attempts to do so have run into the problem that any given item of knowledge, technique for representing knowledge, method for doing reasoning, or architecture for arranging agents only applies in some situations and not others. We need a new way to think about how to organize the agents in a common sense system; we need an architecture of diversity for commonsense computing (McCarthy, et al., 2002)a framework that supports the organization of arrays of diverse and imperfect methods in order to build societies of agents that together are highly resourceful and robust.

In this thesis, I present a system for reasoning with common sense knowledge in multiple natural languages, as part of the Open Mind Common Sense project. The knowledge that Open Mind collects from volunteer contributors is represented as a semantic network called ConceptNet. Using principal component analysis on the graph structure of ConceptNet yields AnalogySpace, a vector space representation of common sense knowledge. This representation reveals large-scale patterns in the data, while smoothing over noise, and predicts new knowledge that the database should contain. The inferred knowledge, which a user survey shows is often correct, is used as part of a feedback loop that shows contributors what the system is learning and guides them to contribute useful new knowledge.

Video cameras are becoming cheap, small and ubiquitous. With advances in memory, cameras will increasingly be designed to be always ready, always recording. When cameras are always ready, how will videographers -- professional and/or amateur -- decide what to shoot, when to shoot and how to index their video material to best support their communication requirements? In this paper, we describe an approach and early experiments that use a commonsense database and reasoning techniques to support a partnership between the camera and videographer during video capture. We describe a new paradigm for producing commonsense video metadata and describe how it can have a positive impact on video content capture, representation, and presentation.

We describe the design and use of a system facilitating the sharing of calendar information between remotely located family members. Users can choose to enter information into a computerized calendar or to write by hand on digital paper calendars. All of the information is automatically shared among everyone in the distributed family.