At the Indiana Philosophy Ontology (InPhO) Project, we are working to create a dynamic computational ontology for the discipline of philosophy. This knowledge representation is currently being developed primarily to serve the metadata needs of the Stanford Encyclopedia of Philosophy (SEP), although it may have a wide array of other uses. The SEP is a peer-managed, open-access, dynamic reference work, with new entries constantly being published and old entries asynchronously edited. This creates a serious challenge: how can the SEP's metadata (cross-references, tables of contents, search keywords) be efficiently updated to reflect the changing content? The scale, complexity, and dynamic nature of many digital reference works like the SEP calls for sophisticated information management tools to capture and harness their metadata. Structured representations of their subject domains are required for scholarly functions such as semantic search, automated generation of cross-references and tables of contents, and ontology-driven conceptual navigation. For scholarly projects with limited resources, human-intensive methods for coding, populating, and maintaining ontologies are not viable. "Social tagging" approaches leading to so-called "folksonomies" are useful for some purposes, but may not meet standards of academic review.

The Indiana Philosophy Ontology contains categories for persons, documents, organizations, and, most importantly, philosophical ideas. The "person" ontology currently contains a great deal of biographical information about philosophers, and we are working on populating our "document" ontology with a full array of citation information. However, our taxonomy of philosophical ideas is the most useful area of the ontology for addressing the SEP's metadata needs. This taxonomy is generated by first using statistical text processing to produce ranked candidate lists of hypernym and hyponym candidates for each idea term, which are then passed to SEP authors and other users of the system for evaluation. Interested individuals can sign up for an InPhO account and contribute to the extension and population of the ontology. With the users' feedback in hand we can use answer set programming techniques to construct an taxonomy classifying these ideas according to their location in the "intellectual space" of the discipline [1].

Figure 1: The metacontent generating engine.
(For a larger view of figure 1, click here.)

By harnessing a dynamic reference work's most valuable informational resource, the domain experts that serve as their editors and contributors, InPhO tries to bridge the gap between the Web 2.0's wisdom of the crowd approach and intricate, expert designed ontologies that have been proposed by the semantic web community. Through statistical text processing and information extraction algorithms, taxonomic and non-taxonomic information is retrieved from both the documents of the SEP as well as external sources such as Wikipedia and academic genealogy datasets. The human expert feedback is used to evaluate the automated methods' recommendations without presuming knowledge of ontology design or placing undue demands on the contributors' time. Web interfaces are provided to the authors of the SEP to evaluate the suggestions generated by the automatic methods. For example, users can score the relatedness of two ideas in philosophy or the influence one philosopher has on another on a five point scale (see Figure 2). Each of these evaluations is stored and a logic program is run to construct a global populated ontology from the local feedback facts. The feedback from the users provides overlapping pieces of expert knowledge which are ultimately put together by the nonmonotonic answer set program we have developed for this task [1]. This program takes into account the user's levels of expertise in the relevant sub-area of philosophy when resolving feedback inconsistencies [2]. A detailed description of the methodology can be found in [1] and [3]. Figure 1 depicts a diagram of the system. The populated taxonomy can be browsed online.

Figure 2: The Idea Tree interface allows users of the system to evaluate automatically
populated relationships between ideas in philosophy. 
(For a larger view of figure 2, click here )

References

[1] Mathias Niepert, Cameron Buckner, and Colin Allen. Answer set programming on expert feedback to populate and extend dynamic ontologies. In Proceedings of FLAIRS, pages 500-505. AAAI Press, 2008.

[2] Colin Allen, Cameron Buckner, and Mathias Niepert. The World is Not Flat: Expertise and InPhO. Selected papers from the Ninth Annual WebWise Conference. First Monday, Volume 13, Number 8, 2008. [article]

[3] Mathias Niepert, Cameron Buckner, and Colin Allen. A dynamic ontology for a dynamic reference work. In Proceedings of JCDL, pages 288-297. ACM Press, 2007.