Tree Matching Using Data Shaping

Real time big data analytics has become important to meet the business as well as other decision making needs in many complex applications. A significant portion of such data is available and stored in semi-structured form. A tree-based organization is commonly used in such cases. Tree matching is a core component for many applications such as fraud detection, spam filtering, information visualization and extraction, user authentication, natural language processing, XML databases, bioinformatics, etc. Comparing ordered (unordered) trees is compute-intensive, in particular for Big Data. To facilitate comparison of ordered trees, in this paper we address the problem of shaping the semi-structured data to enable time efficient processing on contemporary hardware such as a GPGPU (General Purpose Graphics Processing Unit) and INTEL MIC (a multi-core processors). Specifically, our data shaping approach enables pre-computation of partial edit distance values in parallel. We evaluate our work using real world data sets. Our experimental results show that our SIMT-based PTED-GPU (Parallel Tree Edit Distance using GPU) implementation shows speedup of up to 12X when compared to the state-of-the-art in tree edit distance (TED) computation.