On Efficiency of Semantic Relation Extraction through Low-dimensional Distributed Representations for Substrings

By virtue of recent developments in machine learning techniques, higher-level information can now to be extracted from big data. To analyze big data, efficient and smart representations of data achieved by using sufficiently fast algorithms, as well as highly accurate results, are important. In this paper, we focus on extracting multiple semantic relations using light-weight processing through the efficient low-dimensional expression of substrings in text data. We propose an approach to build features for relation classification consisting of only low-dimensional vectors representing substrings between two words, called substring vectors. The experimental results show that, using efficient low-dimensional representations of data and at a small computational cost, our approach achieves a sufficiently high accuracy that is better than most existing approaches. In addition, through experiments, we ensured that mapping substrings to a sufficiently low dimensional space yields better results in terms of both accuracy and efficiency.