Catálogo de publicaciones - libros

The longer the input sentences, the worse the syntactic parsing results. Therefore, a long sentence is first divided into several clauses, and syntactic analysis for each clause is performed. Finally, all the analysis results are merged into one. In the merging process, it is difficult to determine the dependency among clauses. To handle such syntactic ambiguity among clauses, this paper proposes two-step clause-dependency determination method based on machine learning techniques. We extract various clause-specific features, and analyze the effect of each feature on the performance. For the Korean texts, we experiment using four kinds of machine-learning methods. Logitboosting method performed best and it also outperformed the previous rule-based methods.

Currently the best algorithms for transcription factor binding site predictions are severely limited in accuracy. However, a non-linear combination of these algorithms could improve the quality of predictions. A support-vector machine was applied to combine the predictions of 12 key real valued algorithms. The data was divided into a training set and a test set, of which two were constructed: filtered and unfiltered. In addition, a different “window” of consecutive results was used in the input vector in order to contextualize the neighbouring results. Finally, classification results were improved with the aid of under and over sampling techniques. Our major finding is that we can reduce the False-Positive rate significantly. We also found that the bigger the window, the higher the F-score, but the more likely it is to make a false positive prediction, with the best trade-off being a window size of about 7.

This study tests the sensitivity of input feature space selection on credit rating using four classifiers as backpropagation(BP), Kohonen self-organizing feature map, discriminant analysis(DA), and logistic regression. The results of the study are that at individual methods applied, BP network outperforms two statistical counterparts while Kohonen network shows the least accuracy among the models. The results also show that the selection of the feature spaces to the accuracy outcome may not be very sensitive when we test the four methodologies altogether at aggregate level.

This paper describes an adaptive electronic video store application that monitors customers’ actions and provides dynamic movie recommendation. The adaptive recommendation is formed based on double stereotypes that have been constructed for user modeling. The construction of stereotypes has been based on a novel approach that uses an Immune Network System (INS). In particular, the INS has been applied on data collected from 150 users of an earlier version of the e-commerce application. Specifically, the INS clustered users’ interests as well as movies and represented each resulting cluster with corresponding antibodies. The double classification (users’ interests – movies) was performed in a hierarchical way that resulted in several levels of user stereotypes: These stereotypes are then used dynamically by the e-commerce application to infer users’ interests in movies based on a small set of observed users’ actions.