Dipe-R: a knowledge representation language

https://doi.org/10.1016/S0169-023X(02)00190-8Get rights and content

Abstract

The paper reports the design of Dipe-R, a knowledge representation language. It meets two requirements: (1) Dipe-R should enable the appropriate expression of mental states and processes, viz. in line with basic insights on the nature of (human) communication and knowledge, and (2) Dipe-R should offer the possibility of identifying concepts by describing them by their features. The first requirement means that Dipe-R is capable of representing knowledge adequately. The second reflects a common way of exploring knowledge; this seems useful in various applications, e.g., in Information Retrieval.

For meeting these requirements, Dipe-R has as main characteristics: (a) two types of expressions for representing thoughts (using binary relations), (b) for each expression Dipe-R creates at least one sentence in natural language, thus distinguishing language from ‘meaning’, and (c) each expression in Dipe-R is provided with information on its origin. Some relation types are pre-represented, e.g., for creating type hierarchies. Experiments (in Information Retrieval) indicate that Dipe-R indeed meets its requirements.

Introduction

For various reasons, a person may need knowledge on a specific subject area. One common reason is that it is necessary for successfully doing a task or one’s work, another common reason is curiosity. One way of becoming familiar with a specific subject area is to read (textual) documents about that area. By doing so, the reader gets to know the main concepts, and their names, and how they relate to each other and to other concepts. However, such reading may take a lot of effort, since often one has to read many texts before the answers one needs are given. Especially, it is difficult to obtain a quick overview of the area.

An alternative way of familiarising oneself with a subject area is to represent knowledge of that area in a more structured (formalised) manner, and to explore the thus obtained knowledge representation with a software tool. For example, the user may ask everything that is ‘known’ (i.e., represented) about the concept named ‘optical storage device’. The tool may then present all the statements (‘facts’) in the knowledge representation that contain this concept. Applications of this approach can be found in Information Retrieval, organisational or corporate memories, and expert systems.

It is the latter approach (i.e., the formalisation of knowledge) that we try to improve on in our research. We have explored the theory, and on that basis we have designed an improved knowledge representation language: Dipe-R.

Dipe-R was developed to support a tool that interactively formulates queries, i.e., Information Retrieval. The tool (Dipe-D) is reported in [44], and both Dipe-D and Dipe-R were originally reported in [43]. Our reason for developing a new knowledge representation language is that existing representation languages did not meet the requirements set by Dipe-D (and similar tools). Hence, they do not adequately represent knowledge. These shortcomings go beyond the specific purpose of query formulation; they seem generic for representing knowledge.

The paper is structured as follows. Section 2 poses requirements on Dipe-R. Section 3 discusses related research on representation languages. Section 4 introduces the design of Dipe-R, and 5 Content and expression in Dipe-R, 6 Source information, 7 Derivation work out the design. Section 8 provides some experimental support. Finally, Section 9 draws conclusions and makes suggestions for future research.

Section snippets

Assumptions and requirements

On the basis of a number of assumptions (Section 2.1), we specify our requirements for Dipe-R (Section 2.2).

Related research

The section explores related research in the field of knowledge representation as far as relevant to Information Retrieval. In the first place, these are representations common in Information Retrieval, such as (a) thesauri and (b) dictionaries. Additionally, there are also representations from outside Information Retrieval that partially meet our requirements: (c) semantic networks, (d) the KL-ONE family, and the products of two large-scale projects in knowledge representation, viz. (e) the

Overall design of Dipe-R

The main features of Dipe-R are introduced one by one.

Content and expression in Dipe-R

Section 5 treats the content of represented thoughts. Section 5.1 presents what types of thoughts Dipe-R represents and Section 5.2 elaborates Dipe-R’s expression mechanism.

Source information

In line with the general design of Section 4, Dipe-R stores source information for each represented thought. We decide to represent source information along with the content of a represented thought (instead of separately). This allows for a simple implementation (not needing second order expressions or other difficulties), and still allows for the source information to be used.

The source information of Dipe-R comprises:

  • 1.

    the name of the person holding the thought, or the URL of the document from

Derivation

As mentioned in Section 4, Dipe-R contains a derivation procedure. We mentioned in Section 4.1 that this allows for the representation of potential thoughts. By deriving obvious potential thoughts, that are not explicitly represented, less labour for representation is required. However, the advantage is at the cost of some certainty: the potential thoughts derived may be not accepted by the origins. Dipe-R only provides derivations that are very likely to be accepted by the users and that seem

Dipe-D

Dipe-R was developed simultaneously with Dipe-D, a tool for formulating (Boolean) queries. Dipe-D formulates queries in a two-step process:

  • 1.

    The identification of concepts (representing a user’s information need), by describing them by their types and features, and then solving this specification by consulting a knowledge base (in Dipe-R).

  • 2.

    The transformation of the collection of concepts into a Boolean query, essentially by looking up the designations for each concept.

In our experiments with

Conclusion(s)

On the basis of the preceding sections, we draw as conclusions:

  • 1.

    Existing knowledge representation languages do not meet the two requirements that we regard reasonable for knowledge representation.

  • 2.

    Dipe-R meets the requirements mentioned, and as such is in theory better suited for representing knowledge than other knowledge representation languages.

  • 3.

    On the basis of experiments, Dipe-R seems well suited for supporting query formulation processes comprising two steps, viz. concept identification and

Ruud van der Pol (1962) holds a MSc in engineering. After doing his military service, he worked as a thermal engineer for the aerospace company Fokker Space and Systems and as a patent attorney for the patent law firm EP&C. In 2000, he received a Ph.D. in Computer Science, titled “Knowledge-Based Query Formulation in Information Retrieval.” Currently he is developing a flexible activity and data management system for consultants and lawyers, thereby using his insights on knowledge

References (53)

  • R.J Brachman

    I lied about the trees

    AI Magazine

    (1985)
  • R Brachman et al.

    KRYPTON: a functional approach to knowledge representation

    IEEE Computer

    (1983)
  • P.J. Braspenning, A framework for modelling complex objects, in: Proceedings of Hierarchical Object-Oriented Design,...
  • I Bratko

    Prolog Programming for Artificial Intelligence

    (1989)
  • B.G Buchanan et al.

    Rule-based expert systems: the MYCIN experiments of the stanford heuristic programming project

    (1984)
  • J.L.G. Dietz, A universal, ontology-free conceptual modelling technique, Research Memorandum, Delft University of...
  • H.C. Doets, A.S. Troelstra, Inleiding in de Wiskunde en Verzamelingenleer. Reader, Universiteit van Amsterdam,...
  • T.R. Gruber, The role of common ontology in achieving sharable, reusable knowledge bases, in: J.A. Allen, R. Fikes, E....
  • N Guarino et al.

    OntoSeek: using large linguistic ontologies for accessing on-line yellow pages and product catalogs

    IEEE Intelligent Systems

    (1999)
  • R.V Guha et al.

    CYC: a midterm report

    AI Magazine

    (1991)
  • R.V Guha et al.

    Enabling agents to work together

    Communications of the ACM

    (1994)
  • ISO 2788, Documentation––Guidelines for the establishment and development of monolingual thesauri, International...
  • S Jones

    A thesaurus data model for an intelligent retrieval system

    Journal of Information Science

    (1993)
  • D.B Lenat

    CYC: a large-scale investment in knowledge infrastructure

    Communications of the ACM

    (1995)
  • D.B. Lenat, The Dimensions of Context-Space, Report Cycorp, Austin, TX,...
  • Cited by (0)

    Ruud van der Pol (1962) holds a MSc in engineering. After doing his military service, he worked as a thermal engineer for the aerospace company Fokker Space and Systems and as a patent attorney for the patent law firm EP&C. In 2000, he received a Ph.D. in Computer Science, titled “Knowledge-Based Query Formulation in Information Retrieval.” Currently he is developing a flexible activity and data management system for consultants and lawyers, thereby using his insights on knowledge representation obtained while working on his thesis. His interests include Information Retrieval, Knowledge Representation and Reasoning, Organisational Memories, WorkFlow Management, as well as cognitive neuroscience.

    View full text