Title : Decision processes in the verification of monotone quantifiers
| Presenter | Fabian Schlotterbeck |
| Abstract | Psycholinguistic studies have repeatedly demonstrated that downward entailing quantifiers are more difficult to process than upward entailing ones. Although the empirical phenomenon itself is well-documented, it is a matter of current debate what cognitive processes cause the monotonicity effect. Our main aim is to contribute to this debate by testing predictions about the underlying processes that are derived from competing theoretical proposals: two-step and pragmatic processing models. To this end, we model data from two verification experiments, in particular, reaction times and accuracy, using a well-established model of decision making from mathematical psychology, namely the diffusion decision model (DDM). In both experiments, verification of upward entailing ‘more than half’ was compared to downward entailing ‘fewer than half’. One experiment employed a sentence-picture verification task and the other one used a purely linguistic version of the task. Our initial analyses revealed the same pattern of results across tasks: Both non-decision times and drift rates, two of the free model parameters of the DDM, were affected by the monotonicity manipulation. Thus, our initial modeling results support both two-step (prediction: non-decision time is affected) and pragmatic processing models (prediction: drift rate is affected). I will discuss theoretical implications of these results. |
Title : Could we reduce domain classification for datasets to ontology classification for datasets?
| Presenter | Xu Wang |
| Abstract | Research domain classification for datasets is a process of recognizing right research domain which one dataset belongs to. We are looking forwards to reduce the question "Which research domain this dataset belongs to?" to "Which ontology that can explain this dataset very well?", which means that we can know the right domain for dataset if we know suitable ontology for this dataset. We use ontology classification approach to find ontology which can explain dataset very well. Ontology classification for datasets is an approach to rank ontology candidates by calculation similarity between dataset and ontology. We consider similarity between keywords from datasets and keywords from ontology as similarity between dataset and ontology. We also provide a terminology called "ontology specific view", which is considered as keywords of ontology. We run experiments on Mendeley datasets which already have labeled research domain as platinum standard. The result shows that ontology classification approach don't perform well. |
Title : TNO Knowledge Engine: harness the power of domain knowledge
| Presenter | Barry Nouwt |
| Abstract | At TNO (Netherlands Organization for Applied Scientific Research) we work in close collaboration with businesses, governments and universities on innovations that improve the competitive strength of industry and the well-being of society in a sustainable way. We are active in many domains, from Traffic and Transport to Healthy Living and together with our partners we possess a lot of domain knowledge. Apart from applying this domain knowledge in our projects, we capture it in formal domain models (such as ontologies) and leverage it using tools, such as the TNO Knowledge Engine. The TNO Knowledge Engine utilizes ontologies to make working with knowledge more intelligent, human friendly, connected and secure. This talk will address the question of how domain knowledge, in the form of an OWL ontology, can help to integrate knowledge that resides in heterogeneous knowledge bases. The examples from the Defense and Smart Homes domain illustrate how a Prolog-like reasoner uses the ontology to orchestrate knowledge exchange between heterogeneous knowledge bases. Apart from that, we discuss the versatile role of graph patterns within the TNO Knowledge Engine and explain the InterConnect project in which VU and TNO collaborate on making the devices in our homes more interoperable. |