An Investigation of the Usage of Dynamic Time Warping in String Similarity Estimation

Abstract

String similarity estimation is important in many fields, including natural language processing, information retrieval, and data mining. Dynamic Time Warping (DTW) has emerged as a widely used technique for measuring sequence similarity, effectively accommodating variations in length and temporal distortions. This paper presents an examination of the use of DTW in string similarity estimation. We delve into the adoption of DTW in string similarity estimation in various contexts, such as approximate string matching and spelling correction.  We investigate DTW’s strengths and limitations through empirical analysis, particularly in capturing complex patterns and variations within strings, while taking into account factors such as adaptability and robustness. Furthermore, we discuss the impact of various traditional similarity metrics in comparison with DTW based on their evaluation on two different experimental settings. The findings of this study provide important insights into the effectiveness and challenges of using DTW in string similarity estimation. This work may open up alternative ways for the development of more adaptive string similarity estimation techniques through the use of DTW.