This paper describes the Spyglass tool, which is designed to help analysts explore very large collections of unstructured text documents. Spyglass uses a domain ontology to index documents, and provides retrieval and visualization services based on the ontology and the resulting index. The ontology based approach allows analysts to share information and helps to ensure consistency of results. The approach is also scalable and lends itself very well to parallel computation. The Spyglass system is described in detail and indexing and query results using a large set of sample documents are presented.

Evaluating changes in documentation of technical products is a key issue in knowledge management. A product may be declined in different versions and one way to evaluate changes is to compare the sets of documents which describe each version. The aim of this paper is to propose a framework for exhibiting changes between sets of documents. This framework is based on the representation of the sets of documents in terms of association rules and on the definition of first order predicates for reasoning with these association rules. The aim of the reasoning stage is to exhibit the differences between the sets of documents. These predicates show what rules are specific to a corpus or how differs the usage of concepts appearing in the associations rules. The framework is  experimented with the comparison of two corpuses of documents which describe documentation about two different versions of a spatial component.

A discriminative large margin classifier based approach to anaphora resolution for neuroscience abstracts is presented. The system employs both syntactic and semantic features. A support vector machine based word sense disambiguation method combining evidence from three methods, that use WordNet and Wikipedia, is also introduced and used for semantic features. The support vector machine anaphora resolution classifier with probabilistic outputs achieved almost four-fold improvement in accuracy over the baseline method.

Coinduction has recently been introduced into logic programming by Simon et al. The resulting paradigm, termed coinductive logic programming (co-LP), allows one to model and reason about infinite processes and objects. Co-LP extended with negation has many interesting applications: for instance in developing top-down, goaldirected evaluation strategies for Answer Set Programming. In this paper we show yet another application of co-LP, namely, elegantly realizing Boolean SAT solvers

Clustering and visualization of large text document collections aids in browsing, navigation, and information retrieval. We present a document clustering and visualization method based on Latent Dirichlet Allocation and self-organizing maps (LDA-SOM). LDA-SOM clusters documents based on topical content and renders clusters in an intuitive two-dimensional format. Document topics are inferred using a probabilistic topic model. Then, due to the topology preserving properties of self-organizing maps, document clusters with similar topic distributions are placed near one another in the visualization. This provides the user an intuitive means of browsing from one cluster to another based on topics held in common. The effectiveness of LDA-SOM is evaluated on the 20 Newsgroups and NIPS data sets.

As robots become more common, it becomes increasingly useful for them to communicate and effectively share knowledge that they have learned through their individual experiences.  Learning from experiences, however, is often-times embodiment-specific; that is, the knowledge learned is grounded in the robot’s unique sensors and actuators.  This type of learning raises questions as to how communication and knowledge exchange via social interaction can occur, as properties of the world can be grounded differently in different robots.  This is especially true when the robots are heterogeneous, with different sensors and perceptual features used to define the properties.  In this paper, we present methods and representations that allow heterogeneous robots to learn grounded property representations, such as that of color categories, and then build models of their similarities and differences in order to map their respective representations.  We use a conceptual space representation, where object properties are learned and represented as regions in a metric space, implemented via supervised learning of Gaussian Mixture Models.  We then propose to use confusion matrices that are built using instances from each robot, obtained in a shared context, in order to learn mappings between the properties of each robot.  Results are demonstrated using two perceptually heterogeneous Pioneer robots, one with a web camera and another with a camcorder.

We present a semantically-driven approach to uncertainties within and across ontologies. Ontologies are widely used not only by the Semantic Web but also by artificial systems in general. They represent and structure a domain with respect to its semantics. Uncertainties, however, have been rarely taken into account in ontological representation, even though they are inevitable when applying ontologies in `real world' applications. In this paper, we analyze why uncertainties are necessary for ontologies, how and where uncertainties have to be represented in ontologies, and what their semantics are. In particular, we investigate which ontology constructions need to address uncertainty issues and which ontology constructions should not be affected by uncertainties on the basis of their semantics. As a result, the use of uncertainties is restricted to appropriate cases, which reduces complexity and guides ontology development. We give examples and motivation from the field of spatially-aware systems in indoor environments.

A neural net with multiple output nodes is capable of distinguishing among a set of related input classes even in the absence of training. It can do so with an accuracy that is markedly better than random guessing. This is because each class will tend to activate a different set of output nodes. We refer to this tendency as the net's 'inherent' bias. Ascertaining a net's inherent bias  may be thought of as learning the net. One may learn the net either instead of training it, or prior to training it. Furthermore, one only needs a small number of samples from each input class in order to reliably learn the net. If a net has been previously trained on a different, related set of classes, then ascertaining the inherent bias is a form of knowledge transfer. When such a net is trained to respond in accordance with its inherent bias, one may obtain substantially higher accuracies than is provided by nets trained in the standard fashion. Furthermore, when using a deep net, we were able to obtain such improvements while only allowing the top layer of the net to train. This layer contained only about 5.7% of the net's free parameters.

Knowledge compilation is a common technique for propositional logic knowledge bases. A given knowledge base is transformed into a normal form, for which queries can be answered efficiently. This precompilation step is expensive, but it only has to be performed once.  We apply this technique to knowledge bases defined  in the Description Logic ALC. We discuss an efficient satisfiability test as well as a subsumption test for precompiled concepts and Tboxes. Further we use the precompiled Tboxes for efficient Tbox reasoning. Finally we present first experimental results of our approach.

Students must be able to competently compose essays in order to succeed in school and progress into the workplace. Current intelligent tutoring systems (ITS) attempt to provide individual training that is lacking in the current educational system. To provide efficient individual training through ITS, the systems must be able to effectively assess writing input from students. Necessary components for computer-based writing tutors are algorithms that mimic human judgments of writing. The current study attempts to establish a connection between paragraph position and human ratings of paragraph type through the use of computational measures provided by Coh-Metrix. We find that expert raters do not easily identify paragraph type and ratings of paragraph type do not map onto paragraph position.