Classifying Quranic Verse Topics using Word Centrality Measure

Muslims believe that, as the speech of Allah, The Quran is a miracle that has specialties in itself. Some of the specialties that have studied are the regularities in the number of letters, words, vocabularies, etc. In the past, the early Islamic scholars identify these regularities manually, i.e. by counting the occurrence of each vocabulary by hand. This research tackles this problem by utilizing centrality in quranic verse topic classification. The goal of this research is to analyze the effect of The Quran word centrality measure on the topic classification of The Quran verses. To achieve this objective, the method of this research is constructing the Quran word graph, then the score of centralities included as one of the features in the verse topic classification. The effect of centrality is observed along with support vector machine (SVM) and naïve Bayes classifiers by performing two scenarios (with stopword and without stopword removal). The result shows that according to the centrality measure the word “الله” (Allah) is the most central in The Quran. The performance evaluation of the classification models shows that the use of centrality improves the hamming loss score from 0.43 to 0.21 on naïve Bayes classifier with stopword removal. Finally, both of classification method has a better performance in word graph that use stopword removal.

Relating centralities in graphs and the principal eigenvector of its distance matrix

In this work a new centrality measure of graphs is presented, based on the principal eigenvector of the distance matrix: spectral closeness. Using spectral graph theory, we show some of its properties and we compare the results of this new centrality with closeness centrality. In particular, we prove that for threshold graphs these two centralities always coincide. In addition we construct an infinity family of graphs for which these centralities never coincide.

Three Perspectives on Centrality

This chapter presents three different perspectives on centrality. In part, the motivation is definitional: what counts as a centrality measure and what doesn’t? But the primary purpose is to lay out ways that centrality measures are similar and dissimilar and point to appropriate ways of interpreting different measures. The first perspective the chapter considers is the “walk structure participation” perspective. In this perspective, centrality measures indicate the extent and manner in which a node participates in the walk structure of a graph. A typology is presented that distinguishes measures based on dimensions such as (1) what kinds of walks are considered (e.g., geodesics, paths, trails, or unrestricted walks) and (2) whether the number of walks is counted or the length of walks is assessed, or both. The second perspective the chapter presents is the “induced centrality” perspective, which views a node’s centrality as its contribution to a specific graph invariant—typically some measure of the cohesiveness of the network. Induced centralities are computed by calculating the graph invariant, removing the node in question, and recalculating the graph invariant. The difference is the node’s centrality. The third perspective is the “flow outcomes” perspective. Here the chapter views centralities as estimators of node outcomes in some kind of propagation process. Generic node outcomes include how often a bit of something propagating passes through a node and the time until first arrival of something flowing. The latter perspective leads us to consider the merits of developing custom measures for different research settings versus using off-the-shelf measures that were not necessarily designed for the current purpose.

Characterizing Phase Behavior Through Time-Varying Microarchitecture Independent Characteristics Clustering

Programs often exhibit repeating behaviors, which are known as program phases. The automatic discovery of such structured behavior has benefited many applications. However, many existing phase signatures lack the ability to reason about what are the key factors of each phase. Also, programs exhibit phase behavior at many different granularities, and some exhibit hierarchical phase behavior. Many techniques focus on a single granularity, which can cause an out of sync classification with the actual phase behavior. We solve these problems by adopting a recently proposed method of subsequence clustering of multivariate time series. Using this method, the phases started to have a much more interpretable signature (MRF). We graphically showed that the method partitions the execution into a temporally consistent way. We showed the effectiveness of MRF's signature by using a centrality measure to identify the most important characteristics within a program phase. Finally, we present a case study to show the relationship between the MRF signature and source code.