Christopher Menzel
Texas A&M University
cmenzel@tamu.edu

Background to the Questions

• The Current Paradigm for Searching the Web
  • The current paradigm for searching and extracting information from the web is a function of current web architecture — unstructured text and graphics.
  • Google: Semantic content inferred (sort of) from web structure
    • Employs a clever method of inferring content from the hyperlinked structure of the web. Note, though, that this is an inference from structure, NOT from content.
  • Limitations
    • General reliance upon presence of explicit string representations of desired content.
      • Cf. Google's own suggested search techniques:

        USE
          [ Idaho luxury hotel ]
        NOT
          [ a fancy place to stay in Idaho ]

        USE
          [ tutorial ] [ introduction ] OR [ overview ]
        NOT
          [ help ]

    • No explicit inference possible, as content is not represented in a uniform, processable form with a fixed, sharable semantics and a corresponding inference engine.
      • E.g., current paradigm does not accommodate inferences from "fancy place to stay" to "luxury hotel".
• The Semantic Web I: Canonical Content
  • XML-based Languages: http://www.w3.org/XML
    • The next step in the evolution of the web under active development — the so-called "Semantic Web" — involves the development of web pages whose content is explicitly representated in specialized, XML-based formal languages.
    • These languages are designed with computational tractability in mind, and hence, expressively, are sublanguages of full FOL.
      • Resource Description Framework (RDF): http://www.w3.org/RDF
        • Basically existential, conjunctive logic. See P. Hayes: "RDF Semantics" (http://www.w3.org/TR/rdf-mt).
      • RDF Schema (RDFS): http://www.w3.org/TR/rdf-schema
        • Adds (notably) classes, properties, lists, and various related notions to the basic semantics of RDF.
      • Web Ontology Language (OWL): http://www.w3.org/2001/sw/WebOnt
        • Extends RDFS with further primitives.
  • Logic-based languages
    • Designed for exchange and integration of information and for metatheory, and hence include (usually properly) full FOL.
      • Knowledge Interchange Format (KIF): http://logic.stanford.edu/kif/kif.html
      • Common Logic (CL): http://cl.tamu.edu
      • CycL: http://www.cyc.com/cycl
      • OWL Full: http://www.w3.org/2001/sw/WebOnt
      • Conceptual Graphs: http://www.bestweb.net/sowa
• The Semantic Web II: Ontologies and Intelligent Agents
  • Intelligent Agents
    • Critical to the success of the Semantic Web will be the development of software agents that can extract content from the web in response to complex queries by reasoning upon its explicit, canonical content.
  • The Need for Ontologies
    • Typically, however, just as in ordinary discourse, most explicit content presupposes a vast amount of implicit, higher-level, "background" information, e.g.:
      • That a hotel is a place to stay.
      • That hotels are typically made of mostly concrete, steel, and glass.
    • Implicit background information is crucial for making relevant inferences:
      • That one is likely to find beds in hotels.
      • That hotels are large, heavy, objects.
      • That hotels are difficult or impossible to move from place to place.
      • That hotels make poor camping equipment.
    • Background information must therefore be made explicit to support the inferences needed to support complex queries to the explicit, distributed content in next-generation web pages.
    • Ontologies are explicit representations of this sort of higher-level background information, and are thus crucial to the evolution of the Semantic Web.
• Some Notable Loci of Formal Ontology
  • Cycorp
    • Upper Cyc Ontology
      • "[A]pproximately 3,000 terms capturing the most general concepts of human consensus reality."
  • Teknowledge
    • SUMO upper ontology
  • Laboratory for Applied Ontology
    • Trento (Nicola Guarino) and Rome
    • Numerous foundational papers on formal ontology
  • Institute for Formal Ontology and Medical Information Science (IFOMIS)
    • Leipzig (Barry Smith)
    • Basic Formal Ontology
  • National Institute for Standards and Technology (NIST)
    • Process Specification Language (Michael Grüninger)
  • John Sowa
    • One-man ontology machine from an evangelical Peircean perspective.

The Questions

The above suggests a broader, "public service" question than the first of the advertized questions:

Should philosophers participate in the development of the Semantic Web? That is, should philosophers, as a discipline, actively engage in development, refinement, and maintenance of formal and applied ontologies and reasoning systems to enhance intelligent search on the World Wide Web?

• Easy answer: Yes!
• Wealth of possibilities for useful philosophical involvement, both critical and constructive.
  • Formal ontology development
    • Formal ontologies the most trenchent point of contact betweeen the philosophical community and the Semantic Web.
      • Though, not surprisingly, the point where the work of philosophers has been drawn upon the most.
      • Still, most of the actual work still largely done by carried out by philosophically literate computer scientists and engineers not themselves trained as philosophers.
        • Gruber, Genesereth, Guarino, Grüninger, Hobbes, Hayes, Sowa, Lenat, Guha
  • Critiques of applied ontologies
    • Many, probably most, applied ontologies not developed by philosophers.
    • Consequently, many issues familiar to philosophers are being rehashed by ontology developers, often with philosophically questionable results.
    • Consequently, there is a great need for the critical eyes of trained philosophers to be cast upon existing ontologies.
  • Contributions to applied ontologies
    • Many applied domains (e.g., automobile manufacturing) present extremely interesting logical and conceptual challenges.
      • Proper representation of time and change.
      • Choice of ontological primitives.
      • Object individuation.
      • Situated reasoning.
    • Career and consulting opportunities.
      • All of the organizations noted above and many private businesses are hiring philosophers to develop applied ontologies.
  • Solutions to the Integration Problem
    • Both formal and applied ontology development efforts tend to focus around some group or individual.
    • Hence, they tend to be somewhat idiosyncratic.
    • This raises the specter of a philosophical Tower of Babel.
      • The problem is not simply one of translating between ontologies based upon different but semantically equivalent lexicons.
      • More typically lexicons and their semantics differ.
      • Might also find different choices of upper-level primitives, e.g., endurantism vs perdurantism; events vs temporally-indexed relations; etc.
      • Might also find different ontologies are based upon different upper-level ontologies altogether, e.g., Upper Cyc vs. SUMO.
    • Critical that there be some overarching framework that enables sharing and integration of information in different ontologies, using different lexicons, possibly based upon different upper-level ontologies.
  • Contributions to Automated Reasoning Systems
    • Many types of reasoning systems of interest to philosophers and philosophical logicians are relevant to the intelligent search.
      • Intensional logics (notably logics of belief)
      • The logic of common knowledge
      • Defeasible reasoning
      • Non-monotonic logics
  • Contributions to formal language development
    • Very large number of difficult issues of form and expressive power surrounding development of both XML-based and logic-based languages for the Semantic Web.
      • Choice of primitives
      • Semantics
      • Tractability/expressiveness tradeoffs
      • Integration of languages
      • See RDF and KIF/CL mailing list archives

Should (can?) philosophers, as a discipline, build a formal ontology of philosophical concepts to enhance searching, linking, and cross-referencing of philosophical content on the World Wide Web?

• Two elements here:
  1. The construction of an "ontology" of philosophical concepts.
  2. The use of that ontology to facilitate intelligent search on the web for peculiarly philosophical content.
• A priori: Surely such an ontology would prove useful in the search for philosophical content if the Semantic Web architecture is itself feasible.
• Two sources of uncertainty:
  • Feasibility of a general philosophical ontology.
  • Quality of content returned questionable when the search space is the entire web in all its anarchic glory.

This leads to the second question:

Should the Stanford Encyclopedia of Philosophy project develop such a formal ontology for managing the metacontent in its reference work (i.e., for managing cross-references as entries come online or are changed, for building other thematic navigation tools, etc), or will statistical techniques in computational linguistics suffice?

• The SEP provides a unique opportunity in that it provides a somewhat constrained environment
  • Uniformity of format
  • Relatively small document base
  • Indeed, an ideal test best for more robust approaches to intelligent search.
    • If it can't work here, it can't work anywhere!
  • But is it scaleable?
    • Not clear that if it can work here, it can work anywhere!
• SEP ontology as a "standard" reference ontology for philosophical documents on the web.
  • "SEP compliant" documents?

The third question, briefly considered…

What do philosophers need to know about markup languages such as XML and XHTML?

• Hopefully not much.
• Question is: How much do they need to know about representing content in a form usable by the Semantic Web or an "ontologized" SEP?
  • Requires representation in some XML-based language like RDFS or OWL.
  • Not clear how much *philosophers* will need to know about this, but if philosophical content is going to get on the web, this is the form it's gonna have to be in. Nonetheless, it would be good at least for philosophers to familiarize themselves with a variety of more specialized languages that are being proposed as the basis for representing content on the Web.

Final question, not considered at all:

What should philosophers be doing about developing common practices, standards and tools for electronic publishing?