Search

Es wird ein inkrementeller Ansatz zur Koreferenzanalyse deutscher Texte vorgestellt. Wir zeigen anhand einer breiten empirischen Untersuchung, dass ein inkrementelles Verfahren einem nichtinkrementellen überlegen ist und dass jeweils die Verwendung von mehreren Klassifizierern bessere Resultate ergibt als die Verwendung von nur einem. Zudem definieren wir ein einfaches Salienzmass, dass annähernd so gute Ergebnisse ergibt wie ein ausgefeiltes, auf maschinellem Lernen basiertes Verfahren. Die Vorverarbeitung erfolgt ausschliesslich durch reale Komponenten, es wird nicht - wie so oft - auf perfekte Daten (z.B. Baumbank statt Parser) zurückgegriffen. Entsprechend tief sind die empirischen Ergebnisse. Der Ansatz operiert mit harten linguistischen Filtern, wodurch die Menge der Antezedenskandidaten klein gehalten wird. Die Evaluierung erfolgt anhand der Koreferenzannotationen der TüBa-D/Z.

In this paper we focus on anaphora resolution for German, a highly inflected language which also allows for closed form compounds (i.e. compounds without spaces). Especially, we describe a system that only uses real preprocessing components, e.g. a dependency parser, a two-level morphological analyser etc. We trace the performance drop occurring under these conditions back to underspecification and ambiguity at the morphological level. A demanding subtask of anaphora resolution are the so-called bridging anaphora, a special variant of nominal anaphora where the heads of the coreferent noun phrases do not match. We experiment with two different resources in order to find out how to cope best with this problem.

We introduce a system for anaphora resolution for German that uses various resources in order to develop a real system as opposed to systems based on idealized assumptions, e.g. the use of true mentions only or perfect parse trees and perfect morphology. The components that we use to replace such idealizations comprise a full-fledged morphology, a Wikipedia-based named entity recognition, a rule-based dependency parser and a German wordnet. We show that under these conditions coreference resolution is (at least for German) still far from being perfect.

We describe the task of automatically detecting interactions between proteins in biomedical literature. We use a syntactic parser, a corpus annotated for proteins, and manual decisions as training material. After automatically parsing the GENIA corpus, which is manually annotated for proteins, all syntactic paths between proteins are extracted. These syntactic paths are manually disambiguated between meaningful paths and irrelevant paths. Meaningful paths are paths that express an interaction between the syntactically connected proteins, irrelevant paths are paths that do not convey any interaction. The resource created by these manual decisions is used in two ways. First, words that appear frequently inside a meaningful path are learnt using simple machine learning. Second, these resources are applied to the task of automatically detecting interactions between proteins in biomedical literature.

We describe the task of automatically detecting interactions between proteins in biomedical literature. We use a syntactic parser, a corpus annotated for proteins, and manual decisions as training material. After automatically parsing the GENIA corpus, which is manually annotated for proteins, all syntactic paths between proteins are extracted. These syntactic paths are manually disambiguated between meaningful paths and irrelevant paths. Meaningful paths are paths that express an interaction between the syntactically connected proteins, irrelevant paths are paths that do not convey any interaction. The resource created by these manual decisions is used in two ways. First, words that appear frequently inside a meaningful path are learnt using simple machine learning. Second, these resources are applied to the task of automatically detecting interactions between proteins in biomedical literature.

BACKGROUND: Research scientists and companies working in the domains of biomedicine and genomics are increasingly faced with the problem of efficiently locating, within the vast body of published scientific findings, the critical pieces of information that are needed to direct current and future research investment. RESULTS: In this report we describe approaches taken within the scope of the second BioCreative competition in order to solve two aspects of this problem: detection of novel protein interactions reported in scientific articles, and detection of the experimental method that was used to confirm the interaction. Our approach to the former problem is based on a high-recall protein annotation step, followed by two strict disambiguation steps. The remaining proteins are then combined according to a number of lexico-syntactic filters, which deliver high-precision results while maintaining reasonable recall. The detection of the experimental methods is tackled by a pattern matching approach, which has delivered the best results in the official BioCreative evaluation. CONCLUSION: Although the results of BioCreative clearly show that no tool is sufficiently reliable for fully automated annotations, a few of the proposed approaches (including our own) already perform at a competitive level. This makes them interesting either as standalone tools for preliminary document inspection, or as modules within an environment aimed at supporting the process of curation of biomedical literature.

Research scientists and companies working in the domains of biomedicine and genomics are increasingly faced with the problem of efficiently locating, in the vast amount of published scientific results, the critical pieces of information that are needed in order to assess current and future research investment. In this paper we describe approaches taken within the scope of the second Biocreative competition in order to solve two aspects of this problem: the detection of novel protein interactions reported in scientific articles, and the detection of the experimental method that was used to confirm the interaction. Our approach is based on a high-recall protein annotation step, followed by two sharp disambiguation steps. The remaining proteins are then combined according to a number of lexico-syntactic filters, which deliver high-precision results, while maintaining a reasonable recall.

Computer Assisted Language Learning (CALL) has been one of the first incarnations of E-Learning, which has not only become a trend nowadays but also a commercial factor. CALL has its own international conferences and a wide range of academic prototypes and commercial products are available. As with E-Learning software in general, advanced CALL systems are based on the principles of multi media design, facilitating modalities like audio and video, e.g. speech input and output, animated graphics and avatars for guiding the user. Although these design standards do improve the quality of E-Learning systems, it is widely accepted that an optimal learning environment is one where the learner is guided by an intelligent personalized tutor, that adapts to the level of expertise of the learner, for example in tailoring explanations to the user's domain knowledge and selecting appropriate exercises. This is the area of Intelligent Tutorial Systems (ITS) and User Modelling (UM), so far a mainly academic discipline. If at all, only few commercial (CALL) products incorporate such capabilities. The problem is that domain knowledge, reasoning and learning facilities and - in the case of verbal explanations - natural language understanding capabilities are to be integrated; altogether a non-trivial task for many application areas. As a consequence, intelligent tutorial systems seem to be suited well only for restricted domains, e.g. circuit design (Davis 1984).