Reading Research Genre: The Impact of Thematic Progression

The research genre has specific communicative purposes which require students to understand the tone, generic and disciplinary conventions. The present study explored the potential of thematic progression (TP) to contribute to research argument readability. TP concerns how clauses encode information and how that information is carried forward. Three major types of TP include: (a) constant TP; (b) linear TP; and (c) derived TP. Overuse of constant TP often prevents a text from developing, while linear TP better contributes to cohesion. It was hypothesized that an effective TP pattern, if present, may help graduate students better grasp the gist in research arguments. Two groups of participants, native and non-native speakers of English, were recruited to read six Introductions of varying TP patterns and conceptualize their readability. The results revealed that the TP pattern may not have strong predictive power; rather, fewer information breaks better predict readability. Regardless of TP patterns, heavy themes and rhemes may impede understanding. Static and simple themes also do not facilitate readability. The use of metadiscursive devices may facilitate readability on condition that the themes are informative. Effective lexical chains and marked themes, which signal the ties between clauses, also ease processing.

Improvement in Text Categorization Using Semi-Supervised Approach and Lexical Chains

Text categorization is used for assigning the class labels to the available data set or providing a conceptual view to a data set. The text categorization can be performed in two ways supervised way, and in an unsupervised way. But alone neither can perform well in the categorization of data set. So a semi-supervised model with the combination of lexical chains is used to perform the task of categorization. In the proposed semi-supervised model the lexical chains are used to determine the numbers of clusters has to be formed using k-means clustering. This ‘k-means’ will divide the data set into different categories and then onto these different categories the support vector Machine (SVM) model is applied for the classification task. The purpose is to improve the performance of support vector Machine by having data already in some pattern, otherwise, support vector Machine will take a lot of time in the training of data set.