MindShare

Publication has been postponed, waiting on funding to conduct the experiments. The software has been implemented as a platform for testing the thesis and refining my MS thesis using social software in a real-world environment, not as a traditional Media Lab "demo." (Although it did give good demo.)

MindShare is a new way of sharing knowledge within an organization or community of interest. MindShare uses each person's personal information-organizing scheme as a basis for navigating and presenting new information, and it leverages each person's everyday Web information-finding and -organizing activities for the benefit of the whole.

The MindShare approach was the focus of my M.S. thesis at MIT, MindShare: Knowledge Sharing via Personalized Views on a Composite Ontology. Abstract:

Ontologies, or ways of categorizing and describing things, have application to organizing information for human access and allowing software agents to exchange knowledge. The traditional approaches to ontology development involve dedicated effort by a small team of knowledge engineers in a slow and expensive process. We show that a composite ontology of information resources can be constructed from the personal ontologies of individuals in an organization via an ongoing process that combines normal personal organizing activites with low-impedence computer mediation. The MindShare approach is designed to facilitate this process within an organization of people who wish both to organize internal and public information for their personal benefit, and to leverage information previously discovered by others within the organization. MindShare software provides the user with an animated graphical ontology browser for organizing Web documents according to his or her personal ontology, and also for accessing the composite ontology as extensions of the personal ontology with an assisted bimodal browsing interface. The personal ontologies are mapped to the composite ontology by a process that includes MindShare term-based information retrieval methods to suggest existing topics when a user chooses to add a topic, and a variety of user interface affordances that encourage the user to add exposed pieces of the composite ontology to their personal ontology (and later build upon those pieces). We implemented the MindShare software and conducted an initial trial with a small group of users. The MindShare approach behaved as hoped, suggesting that the approach has promise and merits further investigation.

The MindShare approach has its origins in my earlier ideas for distributed ontology evolution for what's now called the Semantic Web. (Tim Berners-Lee was much more capable than I at advocating the "Semantic Web" idea.) Excerpt from the thesis introduction:

If two agents — be they human or software — are to communicate and share knowledge effectively with each other, they must have a common conceptual model of the domain of their discourse, and common terms for concepts in that model. We call the common conceptual and terminological model an ontology.

Here we will use the term "ontology" more in the sense it is used by ontological engineers, as a machine-readable formal specification of a conceptualization, rather than in the metaphysical senses it is used by philosophers [Gruber][Guarino][Smith]. Another way to look at an ontology is as a model for categorizing or describing objects — their names, their properties, the relationships among them, the contraints they obey, etc. Some examples of ontologies within applications include:

• Web directories such as Yahoo employ simple topical category ontologies of Web sites and pages, to help human agents locate desired information.

• If Web pages had semantic encoding in terms of a common ontology, rather than merely presentation-oriented HTML or ad-hoc XML DTDs, the Web could be treated as a massive-scale machine-accessible distributed knowledgebase, over which agents could perform more precise term-based information retrieval [Luke], make inferences to answer questions [Heflin], and present information in forms best suited to the user's current interface modality.

• An ontology of consumer products permits description of product categories and subcategories, features, vendors, pricing structures, etc. If vendors' and consumers' agents commit to a common ontology, this can facilitate ecommerce services such as cross-vendor product selection aids and product offerings tailored to the needs and preferences of the individual consumer [Guttman].

Common ontologies are essential for many desirable applications, but arriving at a common ontology is often difficult, sometimes prohibitively so. One of the greatest difficulties is in the development and maintenance of the ontology. Identifying the pertinent concepts and consistently describing or classifying objects in terms of the concepts is a major undertaking in sheer volume of work, and often requires special ontology engineering skills.

[...proceeds into brief survey of traditional approaches...]

The actual approach employed by MindShare is substantially different than the approaches I'd been thinking of for Semantic Web. The general class of approach is necessarily highly sensitive to the particular application and to the HCI and motivations of its distributed human computational elements.

To receive occasional announcements about MindShare, please email neil@neilvandyke.org.