UNL 2005 Specifications

7 June 2005
Copyright © UNL Center of UNDL Foundation

Chapter 6  Knowledge Representation in UNL

In this chapter, how the various knowledge represented in natural languages should be defined and how the knowledge in UNL should be organized are described. In addition, what kinds of knowledge are necessary and how they should be represented in UNL for reasoning by computers are also described. This chapter includes two sections as follows:

1. Three Levels of Knowledge
2. Representation of the Three Levels of Knowledge

6.1 Three Levels of Knowledge

We consider that knowledge can be divided into three levels: linguistic knowledge, definitions of concepts, and others.

(1) Linguistic knowledge is the knowledge providing semantic information of words for understanding sentences of natural languages. For example, an “author” is a “person”, which can take various actions that a person can take. This level of knowledge is necessary to provide semantic background of natural language sentences.

(2) Definitions of concepts provide the knowledge of the concepts in connection with other concepts that can specify the concepts. For example “a person who write a book” is the sentence to define what is an “author”. Through this definition, it is possible to know that a person who wrote a book will be the author of the book. This level of knowledge is indispensable for reasoning in information retrieval. etc.. 

(3) The third level of knowledge includes every kind of knowledge provided by natural language sentences such as encyclopedia. This level of knowledge is for providing human with necessary information of concepts, and for providing computers with various information and knowledge on concepts necessary for reasoning and providing information on the concepts.

In the UNL system, linguistic knowledge is provided in UNLKB with UWs representing concepts. Definitions of concepts and knowledge on concepts described in natural language sentences are represented by UNL expressions. Figure 6.1 shows the relations of the three levels of knowledge and how they are represented in the UNL Knowledge System.
   

Figure 6.1 Three Levels of Knowledge in the UNL Knowledge System

6.2  Representation of the Three Levels of Knowledge

Figure 6.2 shows how the three levels of knowledge are represented in UNL and how they are linked with each other in the UNL Knowledge System.

·         UWs representing concepts of words are defined in UNLKB. The circles of nodes indicate UWs and arrows indicate relations in figure 2. Concepts of UWs are defined in the UW System of the UNLKB. The UW System is a lattice type of hierarchy made of UWs related to each other. The details of the UW System and UNLKB are explained in section 2.1.

·         A definition of a concept (UW) is described in a file. Each file consists of sentences to define the concept of the UW. Sentences are expressed in UNL. Each definition file is linked with corresponding UW using relation ‘equ’. All UWs used in the UNL expressions of definition must be defined in UNLKB.

·         UNL encyclopedia is composed of files of sentences (UNL documents) describing information and knowledge about UWs. A file contains information and knowledge about a UW. Information and knowledge about a UW can be provided in different files. Each file of encyclopedia is differentiated using the developer name of the content as the file name. These files are linked with corresponding UWs using relation ‘cnt’. All UWs included in UNL expressions of information and knowledge must be defined in UNLKB.

UWs are linked with definition files or encyclopedia files in the following format:

Where,

A UNL document means a text file or a HTML file in which UNL expressions for the original natural sentences are included. A UNL document is attached to with extension “.unl” as in “http://www.undl.org/UNLencyclopedia/GreatWall.html.unl” or “x:/unlencyclopedia/local/GreatWall.html.unl”.

Figure 6.2 Representation of the three levels of Knowledge in the UNL Knowledge System