Using String Vector based KNN for Keyword Extraction

In this research, we propose the string vector based KNN as the approach to the keyword extraction. The keyword extraction may be viewed as an instance of word classification, encoding words into numerical vectors may cause the main problems, such as the huge dimensionality, the sparse distribution and the poor transparency, and the problems were solved by encoding texts into string vectors in previous works on text mining tasks. In this research by these motivations, we encode words into string vectors, define the semantic operation on string vectors, and modify the K Nearest neighbor into its string vector based version which is used for the keyword extraction. As the benefits from this research, we expect the better performance and more compact representations than encoding words or texts into numerical vectors. Hence, the goal of this research is to implement the keyword extraction system with the benefits.