A Scalable Framework for Stylometric Analysis Query Processing

Abstract

Stylometry is the statistical analyses of variations in the author's literary style. The technique has been used in many linguistic analysis applications, such as, author profling, authorship identifcation, and authorship verifcation. Over the past two decades, authorship identifcation has been extensively studied by researchers in the area of natural language processing. However, these studies are generally limited to (i) a small number of candidate authors, and (ii) documents with similar lengths. In this paper, we propose a novel solution by modeling authorship attribution as a set similarity problem to overcome the two stated limitations. We conducted extensive experimental studies on a real dataset collected from an online book archive, Project Gutenberg. Experimental results show that in comparison to existing stylometry studies, our proposed solution can handle a larger number of documents of different lengths written by a larger pool of candidate authors with a high accuracy.

A scalable framework for cross-lingual authorship identification

“…However, the main problem of the PkNN method is that the classifier is sensitive to outliers. To mitigate this problem, we use a stylometric data representation, which makes use of set similarity search [36] such that the stylistic variations between documents can be measured as a set distance [21].…”

“…On the other hand, in a study which employed machine translation [4], the logistic regression classification has led to superior accuracy using a larger set of features (i.e., word n-grams) in comparison to other classifiers. The nearest neighbor classifier has reported a reasonable accuracy (71%) when used in a corpus with a large number of candidate authors [36].…”

“…As a result, they are not completely language-independent. As for the kNN-based method [36], the feature sets include language-specific feature types: lexical and syntactic. Hence, it cannot be directly applied to cross-lingual settings.…”

“…Consider our CLSS method as an example. We first formulate a 10D feature space called VSP and then apply the set similarity PkNN method [36] to the analysis part. Similarly, for the competitor [28], the authors proposed a set of vocabulary richness (VRFS) features and applied an array of machine learning algorithms to the extracted VRFS feature vectors.…”

“…There exists a long history of stylistic investigations focusing on authorship identification since the 19th century [32]. Most of the existing research in this area has used monolingual corpora and English is the most studied language [9,29,36,43,45].…”