Contributions in Merlin

null

Despite the many initiatives in recent years aimed at creating Language Engineering standards, it is often the case that different projects use different approaches and often define their own standards. Even within the same project it often happens that different tools will require different ways to represent their linguistic data. In a recently started EU project focusing on the integration of Information Extraction and Data Mining techniques, we aim at avoiding the problem of incompatibility among different tools by defining a Common Annotation Scheme internal to the project. However, when the project was started (Sep 2002) we were unaware of the standardization effort of ISO TC37/SC4, and so we commenced once again trying to define our own schema. Fortunately, as this work is still at an early stage (the project will last till 2005) it is still possible to redirect it in a way that it will be compatible with the standardization work of ISO. In this paper we describe the status of the work in the project and explore possible synergies with the work in ISO TC37 SC4.

null

Recent studies on global warming have introduced the inherent uncertainties associated with the costs and benefits of climate policies and have often shown that abatement policies are likely to be less aggressive or postponed in comparison to those resulting from traditional cost–benefit analyses (CBA). Yet, those studies have failed to include the possibility of sudden climate catastrophes. The aim of this paper is to account simultaneously for possible continuous and discrete damages resulting from global warming, and to analyse their implications on the optimal path of abatement policies. Our approach is related to the new literature on investment under uncertainty, and relies on some recent developments of the real option in which we incorporated negative jumps (climate catastrophes) in the stochastic process corresponding to the net benefits associated with the abatement policies. The impacts of continuous and discrete climatic risks can therefore be considered separately. Our numerical applications lead to two main conclusions: (i) gradual, continuous uncertainty in the global warming process is likely to delay the adoption of abatement policies as found in previous studies, with respect to the standard CBA; however (ii) the possibility of climate catastrophes accelerates the implementation of these policies as their net discounted benefits increase significantly.

The true power of anaphora resolution algorithms can only be gauged when embedded into specific Natural Language Processing (NLP) applications. In this paper we report on the anaphora resolution module from ExtrAns, an answer extraction system. The anaphora resolution module is based on Lappin and Leass' original algorithm, which used McCord's Slot Grammar as the inherent parser. We report how to port Lappin and Leass' algorithm to Link Grammar, a freely available dependency-based parsing system that is used in a range of NLP applications. Finally, we report on how the equivalence classes that result from the anaphora resolution algorithm are incorporated into the logical forms used by ExtrAns.

Intertemporal models of the current account generally assume that global shocks do not affect the current account. We use this assumption to identify global and country-specific shocks in a bivariate VAR of output and the current account. Cross-country evidence from the G7 economies suggests that this identification works surprisingly well. We then employ our method to collect stylized facts on international macroeconomic fluctuations. We find that long-term output growth is driven mainly by global factors in most G7 countries and that country-specific shocks are less persistent and generally less volatile than global shocks.

Members of boards of directors of large corporations who also serve together on an outside board, form the so-called interlock graph of the board and are assumed to have a strong influence on each others’ opinion. We here study how the size and the topology of the interlock graph affect the probability that the board approves a strategy proposed by the Chief Executive Officer. We propose a measure of the impact of the interlock on the decision making, which is found to be a good predictor of the decision dynamics outcome. We present two models of decision making dynamics, and we apply them to the data of the boards of the largest US corporations in 1999.

Large-scale parsing is still a complex and time-consuming process, often so much that it is in-feasible in real-world applications. The parsing system described here addresses this problem by combining finite-state approaches, statistical parsing techniques and engineering knowl- edge, thus keeping parsing complexity as low as possible at the cost of a slight decrease in performance. The parser is robust and fast and at the same time based on strong linguistic foundations.

There is no doubt that learning in groups is in general more fun, often more stimulating, and sometimes more successful than learning individually. This simple insight is reflected in the computer supported collaborative work (CSCW) tools omnipresent in class-management systems (e.g., WEBCT, cf. www.webct.com) and web-based learning (WBL) platforms (e.g. MILCA, milca.sfs.uni-tuebingen.de). It is also corroborated by the results of the new problem-based learning (PBL) paradigm which is essentially a collaborative work learning strategy used together with a new concept of teaching (Rhem 1998).