Pronolninalization has been related to tile idea of a local focusa set of discourse entities in the speaker's centre of attention, for exmnple ill Gundel et al. (1993)'s givenness hierarchy or in centering theory. Both accounts say that the determination of tile tbcus depends on syntactic as well as pragmatic factors, but have not been able to pin those factors down. In this paper, we uncover the major factors which determine the focus set in descriptive texts. This new tbcus definition has been ew, luated with respect to two corporm museum exhibit labels, mid newspaper mtieles. It provides an operationalizable basis for pronoun production, and has been implemented as the reusable module gnome-np. The algorithm l)ehind gnome-np is conlpared with the most recent pronoun generation algorithm of McCoy and Strube (1999).

In this study, adsorption of ammonia on activated carbon from aqueous solutions has been studied in a batch stirred cell. Experiments have been carried out to investigate the effects of temperature, ammonia concentration, and activated carbon dose on ammonia adsorption. The experimental results manifest that the ammonia adsorption rate on activated carbon increases with its concentration in the aqueous solutions. Ammonia adsorption also increases with temperature. The ammonia removal from the solution increases as activated carbon mass increases. The Langmuir and Freundlich equilibrium isotherm models are found to provide a good fitting of the adsorption data, with r 2 ¼ 0.9749 and 0.9846, respectively. The adsorption capacity of ammonia obtained from the Langmuir equilibrium isotherm model is found to be 17.19 mg g À1 . The kinetic study shows that ammonia adsorption on the activated carbon is in good compliance with the pseudo-second-order kinetic model. The thermodynamic parameters (~G8,~H8, S8) obtained indicate the endothermic nature of ammonia adsorption on activated carbon.

But, equally important, the culture in the company gradually evolved to becoming much more open, less defensive, a truly learning organisation. What nextDutch National Quality Award the management team decided to re-embark on a renewed company wide program in order to reach new platforms in Total Quality.In the autumn of 1993, even after winning the first Main topic: New Developments in Oleochemical Research, Technology and Application J . 0 . M e t z g e r and Ursula B i e r m a n n , Oldenburg: Ene additions to unsaturated fatty compounds Thermal reactions of unsaturated fatty compounds e. g. of oleic acid and erucic acid with maleic anhydride are well known'. ' . The reaction is called "maleinization" and takes place by forming a new C,C-bond with migration of the ene double bond and 1,s-hydrogen shift. The diastereoselectivity of the reaction has not yet been examined.We carried out the addition of maleic anhydride to oleic acid (180°C, 24 h) and obtained the ene adduct in an isolated yield of 75%. The addition takes place in positions C-9 and C-10 of the molecule chain. The regioisomers were obtained as a mixture of diastereomers as pure (E)-adducts in a ratio of 1:l. The ratio of threo-and erythro-adduct was 4:l. The assignment could be made definitely using I3C-NMR by comparison with literature data3. 0 3 1 '*ooC' 24b + 0 'J. 0. Metzger u. K . E Leisinger, Fat Sci. Technol. 90, 1 [1988].

No abstract

No abstract

No abstract

In natural language generation, different generation tasks often interact with each other in a complex way. We think that how to resolve the complex interactions inside and between tasks is more important to the generation of a coherent text than how to model each individual factor. This paper focuses on the interaction between aggregation and text planning, and tries to explore what preferences exist among the features considered by the two tasks. The preferences are implemented in two generation systems, namely ILEX-TS and a text planner using a Genetic Algorithm. The evaluation emphasises the second implementation and shows that capturing these preferences properly can lead to coherent text.

Most classification models work by first predicting a posterior probability distribution over all classes and then selecting that class with the largest estimated probability. In many settings however, the quality of posterior probability itself (e.g., 65% chance having diabetes), gives more reliable information than the final predicted class alone. When these methods are shown to be poorly calibrated, most fixes to date have relied on posterior calibration, which rescales the predicted probabilities but often has little impact on final classifications. Here we propose an end-to-end training procedure called posterior calibrated (PosCal) training that directly optimizes the objective while minimizing the difference between the predicted and empirical posterior probabilities. We show that PosCal not only helps reduce the calibration error but also improve task performance by penalizing drops in performance of both objectives. Our PosCal achieves about 2.5% of task performance gain and 16.1% of calibration error reduction on GLUE (Wang et al., 2018) compared to the baseline. We achieved the comparable task performance with 13.2% calibration error reduction on xSLUE (Kang and Hovy, 2019), but not outperforming the two-stage calibration baseline. PosCal training can be easily extendable to any types of classification tasks as a form of regularization term. Also, PosCal has the advantage that it incrementally tracks needed statistics for the calibration objective during the training process, making efficient use of large training sets 1 .