Papers by Frederik Schadd
Game on, 2007
Real-time strategy games present an environment in which game AI is expected to behave realistica... more Real-time strategy games present an environment in which game AI is expected to behave realistically. One feature of realistic behaviour in game AI is the ability to recognise the strategy of the opponent player. This is known as opponent modeling. In this paper, we propose an approach of opponent modeling based on hierarchically structured models. The top-level of the hierarchy can classify the general play style of the opponent. The bottom-level of the hierarchy can classify specific strategies that further define the opponent's behaviour. Experiments that test the approach are performed in the RTS game Spring. From our results we may conclude that the approach can be successfully used to classify the strategy of an opponent in the Spring game.
A Feature Selection Approach for Anchor Evaluation in Ontology Mapping
Communications in Computer and Information Science, 2014
Ontology mapping is a crucial step for the facilitation of information exchange between knowledge... more Ontology mapping is a crucial step for the facilitation of information exchange between knowledge sources. In the industry this process is often performed semi-automatically, with a domain expert supervising the process. Such an expert can supply a partial alignment, known as anchors, which can be exploited with more elaborate mapping techniques in order to identify the remaining correspondences. To do this we propose a novel approach, referred to as anchor-profiles. For each concept its degree of similarity to each anchor is gathered into a profile for comparison. We evaluated our approach on the Ontology Alignment Evaluation Initiative (OAEI) benchmark dataset using partial alignments that are randomly generated from the reference alignments. The evaluation reveals an overall high performance when compared with mapping systems that participated in the OAEI2012 campaign, where larger partial alignments lead to a higher f-measure.
Word-Sense Disambiguation for Ontology Mapping: Concept Disambiguation using Virtual Documents and Information Retrieval Techniques
Journal on Data Semantics, 2014
Matching ontologies is a crucial process when facilitating system interoperability and informatio... more Matching ontologies is a crucial process when facilitating system interoperability and information exchange. A reoccurring problem in this process is that names can be ambiguous, yielding uncertainty to whether entities of two heterogeneous ontologies are actually related. Linguistic ontologies provide a clear structure of meanings, rather than names, allowing the quantification of the relatedness of any two given meanings. We propose an approach for the automatic allocation of correct meanings within a linguistic ontology through the use of virtual documents and information retrieval techniques. The benefits of this approach are tested and established using a data set from the Ontology Alignment Evaluation Initiative (OAEI) competition, while further improvements are revealed using a benchmark data set from the same competition.
With the growth of the Linked Data Web, time-efficient Link Discovery frameworks have become indi... more With the growth of the Linked Data Web, time-efficient Link Discovery frameworks have become indispensable for implementing the fourth Linked Data principle, i.e., the provision of links between data sources. Due to the sheer size of the Data Web, detecting links even when using trivial specifications based on a single property can be very timedemanding. Moreover, non-trivial Link Discovery tasks require complex link specifications and are consequently even more challenging to optimize with respect to runtime. In this paper, we present a novel hybrid approach to link discovery that combines two very fast algorithms. Both algorithms are combined by using original insights on the translation of complex link specifications to combinations of atomic specifications via a series of operations on sets and filters. We show in three experiments that our approach outperforms SILK by more than six orders of magnitude while abiding to the restriction of not losing any link.
Real-time strategy games present an environment in which game AI is expected to behave realistica... more Real-time strategy games present an environment in which game AI is expected to behave realistically. One feature of realistic behaviour in game AI is the ability to recognise the strategy of the opponent player. This is known as opponent modeling. In this paper, we propose an approach of opponent modeling based on hierarchically structured models. The top-level of the hierarchy can classify the general play style of the opponent. The bottom-level of the hierarchy can classify specific strategies that further define the opponent's behaviour. Experiments that test the approach are performed in the RTS game Spring. From our results we may conclude that the approach can be successfully used to classify the strategy of an opponent in the Spring game.
Facilitating information exchange is a crucial service for ontology-based knowledge systems. This... more Facilitating information exchange is a crucial service for ontology-based knowledge systems. This can be achieved by the mapping of two heterogenous ontologies. Many mapping frameworks utilize language-based knowledge resources such as WordNet. By coupling all ontology concepts to a corresponding entry in WordNet, one can quantify the lexical relatedness of any two ontology concepts. However, coupling the correct entry is a difficult task due to the ambiguous nature of names. Coupling the wrong entries hence yields similarity values that do not correctly express the relatedness of two given concepts, resulting in a poor performance of the overall mapping framework. This paper proposes an approach for the more accurate coupling of ontology concepts with their corresponding WordNet entries. The basis of the proposed approach is the creation of separate virtual documents representing the different ontology concepts and WordNet entries and coupling these according to their document similarities. The extent of improvements using this approach are evaluated using a data set originating from the Ontology Alignment Evaluation Initiative (OAEI). Furthermore, the performance of a framework using our approach is demonstrated using the results of the OAEI 2011 competition.
Uploads
Papers by Frederik Schadd