We perform Noun Phrase Bracketing by using a local, maximum entropy-based tagging model, which produces bracketing hypotheses. These hypotheses are subsequently fed into a reranking framework based on support vector machines. We solve the problem of hierarchical structure in our tagging model by modeling underspecified tags, which are fully determined only at decoding time. The tagging model pe...

Full discourse parsing in the PDTB framework is a task that has only recently been attempted. We present the Two Taggers approach, which reformulates the discourse parsing task as two simpler tagging tasks: identifying the relation within each sentence, and identifying the relation between each pair of adjacent sentences. We then describe a system that uses two CRFs to achieve an F1 score of 39...

Precision and Recall, as well as their combination in terms of FMeasure, are widely used measures in computer science and generally applied to evaluate the overall performance of ontology matchers in fully automatic, unsupervised scenarios. In this paper, we investigate the case of supervised matching, where automatically created ontology alignments are verified by an expert. We motivate and de...

Statistical tests that compare classification algorithms are univariate and use a single performance measure, e.g., misclassification error, F measure, AUC, and so on. In multivariate tests, comparison is done using multiple measures simultaneously. For example, error is the sum of false positives and false negatives and a univariate test on error cannot make a distinction between these two sou...

We present an on-line system which learns a SPARQL query from a set of wanted and a set of unwanted results of the query. The sets are extended during a dialog with the user guided by recall and F1 measure. The system leverages SPARQL 1.1 and does not depend on any particular RDF graph.

In this paper we present the Arib system for Arabic spelling error detection and correction as part of the second Shared Task on Automatic Arabic Error Correction. Our system contains many components that address various types of spelling error and applies a combination of approaches including rule based, statistical based, and lexicon based in a cascade fashion. We also employed two core model...

This paper describes our system in the shared task of CoNLL-2013. We illustrate that grammatical error detection and correction can be transformed into a multiclass classification task and implemented as a single-model system regardless of various error types with the aid of maximum entropy modeling. Our system achieves the F1 score of 17.13% on the standard test set.

In order to identify variations between two or several versions of Clinical Practice Guidelines, we propose a method based on the detection of noun phrases. Currently, we are developing a comparison approach to extract similar and different elements between medical documents in French in order to identify any significant changes such as new medical terms or concepts, new treatments etc. In this...

This paper proposes state-of-the-art models for time-event relation extraction (TERE). The models are specifically designed to work effectively with relations that span multiple sentences and paragraphs, i.e., inter-sentence TERE. Our main idea is: (i) to build a computational representation of the context of the two target relation arguments, and (ii) to encode it as structural features in Sup...