An Approach to Knowledge Aware Systems

Case Study: Designs On Knowledge

Minimalist data model

DOK employs a minimalist approach, motivated by the following goals

  • To allow maximal flexibility

  • To allow emergent design and structure

  • To give programming and design access to the most elemental levels of organization

  • To be structured to mimic or approach metaphors found semantic networks, complex systems, cognitive and neuroscience, connectionist, AI and intelligent agent designs

  • To be easily combined and recombined into tree and network structures.

  • Separation of theory, implementation and modeling

  • Allow for shape and dimensionality (each data element is able to participate in and contain embedded networks, cycles of internalizations (structures within structures)

Two Key Components and How They are Combined

The DOK system is composed of two elemental structures: Relations and Contents (in the vocabulary of networks, Relations may be thought of as Paths or Edges, and Contents as Nodes.) A A Content is composed of a Label and Text. Relations are composed of a Predicate and any number of Contents. A Content may have any number of Relations attached to it. A Content may contain any number of Relations within it. Thus the Content and Relations are composed of each other and are linked to each other. Relational databases are used to store Relations and Contents, as well as the mapping of their connections to each other. The Content Text field may contain raw text, and XML (Personal Knowledge Markup Language). As such the Content may be composed of any combination of natural language text, semi-structured data trees or structured data records.

Content may be a special type, such as Event (things that happen in time) or Topic (Person, Place or Theme). Relations are comprised an of any number and of Events or Topics. In practice most relations take the form Predicate(TopicA, TopicB): example IsA(Cat, Mammal) or Influenced (Mozart, Beethoven), where Cat, Mammal, Mozart and Beethoven are Content of the type Topic.

Implementation: Data layer

Below is the working description of a functional data layer implementational for Designs On Knowledge.

Hybrid Relational Database with optional Heterogenous XML Markup

  • Characteristics: one main table with shared content for each entry. One text field for character content of arbitrary length, including optional markup (xml or other). Relations are stored in a separate table. Each Content can be linked to an arbitrarily large number of relations (0-many) via a Content-to-Relations-Map table. Each relation represented uniquely in the system (i.e. each combination of arguments occurs only once). The Content Type is extended relationally with any number Extension Tables, which are are used to add fields for specific types of Content (such as people, places, things, concepts, events, etc). Note: all content records are represented in the Content Table and thus, all main searches and functions are enacted on one field.

  • Tables

    • Content table (with entries for every element in the system)

      • Unique ID

      • Title (short)

      • Text (arbitrary length)

      • Author/Contributor/Source

    • Relations table

      • 3-tuple (expandable to N-Tuple)

    • Content-to-Relations-Map table

    • N-Extension Tables. These extend the content types to include fields not shared with other content. For example: Events included temporal information (for example: Begin DateTime, End DateTime).