PORSCE II is comprised by a planning component, a concept relevance module for discovering similarities among OWL ontology concepts, a visual component, and a service replacement component, which cooperate to achieve the goal of automated semantic web service composition.
The planning module is responsible for expressing the web service composition problem as a planning problem and exporting it to the standard language for planning domain and problem description, PDDL. In addition, it cooperates with independent external PDDL-compliant planning systems (LPG-td, JPlan), which search for composition plans by matching OWL-S profile input and output parameters, and evaluates the produced plans, in order to use this evaluation when assessing the quality of each complex service.
The concept relevance module exploits knowledge obtained from reasoning over ontologies to semantically enhance the input and output requirements of the web services, by including semantically related ontology concepts based on different notions of concept similarity.
The visual component, apart from the obvious benefits of visualization of the complex services, their statistics and their quality assessment values, also accommodates intervention to the complex services by replacing certain atomic services with alternatives, which might either be equivalent or semantically similar to the original simple services, as determined by the concept relevance module.
A screenshot of the interface of the system showing the representation of complex services obtained from planners is shown below.
The full implementation code of the PORSCE II system complemented with example files can be found here (last updated on Tuesday, November 3rd, 2009).
Alternatively, a runnable demo can be downloaded from here (last updated on Tuesday, September 22nd, 2009).
Some indicative PDDL domains and corresponding problems produced by the PORSCE II system can be found here (last updated on Tuesday, November 3rd, 2009).
More information about PORSCE II can be found in relative papers in Publications.
