This paper investigates the need for tools that will facilitate a higher-order demand for interoperability between dispar ate systems. An ontology importer for Amine is described which enables ontologies written in linear form to be used with Amine version 4. Furthemore, the ontology importer is shown as an intermediary between CharGer and Amine, thus demonstrating interoperation between two conceptual structures tools, as well as discussed in a wider context by means of a Web service or by interoperating with Proeteg´-O´WL.
The Amine platform provides a comprehensive software suite for symbolic
programming, intelligent system programming and intelligent agents programming[
The importer addresses the desire for CG software tools to interoperate. To have these tools in some sense compatible and functioning with each other is a key goal that would be welcomed by many in the CG community. This ontology importer is a small experiment that could be taken further in the future to achieve the end goal of interoperability between other CG tools. 3.1
As an Implementation Currently the ontology importer consists of a GUI with a text-editor. An input in straight ASCII format is complied and accepted, then linking it to Amines’ APIs. The ontology importer is fully functioning with error handling facilities. It is a Java library, and as such can be accessed from any Java application. One application that comes with the ontology importer for example is a text editor to input the text, which also contains output for informative messages. Currently the prototype accepts the input as a Java String. However the ontology importer will soon provide the facility for a user to enter the URL of a file and have the file compiled and if there are no errors in the input create an ontology. It is significant that the input to the ontology importer is a Java String. This is useful for the following reasons; firstly it enables easy creation of ontologies that have been automatically constructed from other sources, thereby enabling interoperability between Amine and other knowledge representation software. Secondly it enables quicker and easire creation of ontologies by using the keyboard rather than the mouse. 4
The syntax used by the ontology importer is modelled upon that of Prolog+CG
2.0 (hereafter referred to as Prolog+CG). Prolog+CG is a GC tool built in
Java. It is similar to Prolog except it has extensions for handling CG.
Professor Adil Kabbaj, the creator of Amine, was also the creator of Prolog+GG[
Instances can also be created easily in Prolog+CG. For example the following instances can easily be created in Prolog+CG: Boy = Sam, Richard, Gary.
Girl = Greta, Rebecca, Sheila.
Like the ontology importer, Prolog+CG enables ontologies to be entered in CG linear form. However it is not interoperable in that it lacks a cohesive, distinct component for creating and interoperating ontologies. It also does not return an Amine API ontology or lexicon object that can then be used by Amines’ ontology builder. Given all these factors, the ontology importer was designed as a distinct component from the outset rather than to retro- fit Prolog+CGs’ legacy version to the problem in hand. 5
Creating an Ontology with the Ontology Importer As in Amine the topmost type and the topmost relation type must be specified at the start of the ontology. This is achieved by simply entering the word T‘OP’ and specifying its identifier after the :‘:=’ assignment of s ymbol. This process is the equivalent to the dialogue box that appears on Amines’ ontology builder GUI that asks for the topmost type in the hierarchy. The topmost relation in the hierarchy must also be specified by entering R‘ELATION TOP’ and passing the identifier name after the :‘:=’ assignment symbol. The on tology importer then expects these to be the first types in the hierarchy. (NB If an identifier other than the TOP’s identifier is the first in the hierarchy an error is flagged. If RELATION TOP’s identifier is a subtype of any type other than TOP’s iden tifier an error is flagged. Similarly if TOP and RELATION TOP are omitted the ontology would not compile). Adding an ontology using the prototype is thereby much simpler than using Amines’ GUI; indeed one can simply pa ste an ontology into the text editor. The following ontology can then be created as shown in Figure 1: TOP ::= Universal.
RELATION_TOP ::= Relation.
An example of how to automate the ontology importer can be illustrated using
CharGer as an example. CharGer is a CG software tool[
The above code can be either automatically transformed or manually edited to produce the type hierarchy in the format accepted by the ontology importer. One current shotcoming of the ontology importer is that it requires the types to be specified top-down. Hence we need to reorder the list befor e processing with the ontology importer.
TOP ::= Universal.
RELATION_TOP ::= Relation.
The above ontology now can be accepted and compiled by the ontology importer. Achieving the goal of automation across applications has been met, though there were some amendments that need to be done to the ontology in order for it compile. This is clearly a shortcoming which we intend to address in the near future. 7
The primary goal for this work was to be able to use Amine ontologies in Web Services, without using a mouse for data entry. Part of this goal has been achieved; since the ontology importer is a Java library, it can be incorporated into any Java Web Service together with Amine. Similarly it would be useful to enable the creation of Amine ontologies as a web service. Once implemented this will achieve the goal of interoperability. Any application can then add a reference specifying the location of the Web Service and connect to the ontology importers’ API. When connected to the importers’ API the user is indirectly a ccessing Amine’s APIs to create an ontology. An Amine ontology would then be returned, thus achieving the goal which was originally formulated.
The ontology importer presented in this paper provides Amine with a convenient plug-in tool for its ontology builder. This extends to Web Se rvices and mainstream non-CG ontology tools such as Proetge´-O´WL. Future research would include the following. First, the restriction on the order of the lines needs to be lifted. Second, a convenient method needs to be added, such that a URL can be given to the ontology importer, which is then fetched and used as the input to the method which accepts a String. Third, a Web Service needs to be created which provides Amine ontology creation services through a Web Service interface. The ontology importer will be released as Open Source software, thereby benefitting the CG and broader knowledge representation communities.