String Kernel-Based Techniques for Native Language Identification

Abstract In recent years, Native Language Identification (NLI) has shown significant interest in computational linguistics. NLI uses an author’s speech or writing a second language to figure out their native language. This may find applications forensic linguistics, teaching, acquisition, authorship attribution, identification of spam emails phishing websites, etc. Conventional pairwise string comparison techniques are computationally expensive and time-consuming. paper presents fast based on kernels such as spectrum, presence bits, intersection incorporating different learners Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting-XGBoost (XGB). Feature sets for the proposed generated using combinations features n-word grams noun phrases. Experimental analyses carried 8235 English articles from 10 linguistic backgrounds typical NLP benchmark dataset. The experimental results show that technique spectrum kernel with RF classifier outperformed existing character n-gram SVM, RF, XGB classifiers. Also, comparable were observed among kernels. Interestingly, random forest SVM classifiers feature sets. All demonstrated promising improvement training time, best result attaining more than 95 percent decrease time. reduced time makes it well suited scale production.