Authors: S. V. Kogilavani, S.Malliga, C. S. Kanimozhiselvi
Abstract: In any language, sentences with different structures may convey the same meaning. Paraphrase identification is defined as the task of deciding whether two given text fragments have the same meaning or not. The proposed work focuses on the detection of paraphrases in Tamil language using the statistical and semantic analysis of sentences. The statistical analysis calculates the similarity between two Tamil sentences based on Jaccard, Dice, Cosine and Word distance. The semantic analysis determines the similarity of two sentences based on Word order. Classification is a data mining technique that assigns categories to a collection of data in order to aide in more accurate predictions and analysis. The proposed approach utilizes five different machine learning algorithms for classification of given sentence pair using statistical and semantic features. The accuracy and performance of different classification methods are measured on the basis of evaluation parameters like accuracy, precision, recall and f-measures. The combination of statistical and semantic similarity features helps to identify whether the pair of sentences is paraphrase or not.
Full Issue
| View or download the full issue | PDF 106-119 |