Named Entity Extraction from Semi-structured Data Using Machine Learning Algorithms

The modern society have been witnessed that intensive development of Internet technologies had followed to information explosion during last decades. This explosion had been expressing by an exponential growth of data volume among the low-quality information. This paper is designed to provide detailed information about some intellectual tools which are support decision taking by automatic knowledge extraction. In the first part of paper, we considered a preprocessing contains morphological analysis of texts. Then we had considered the model of text documents in the form of a hypergraph and implementation of the random walk method to extract semantically close word’s pairs, in other words, pairs that often appears together. Result of calculations is matrix with word affinity coefficients corresponding to each other component of vocabulary vector. In the second part we describe training of neural network for linguistic constructions extraction. These ones include possible values of text named entities descriptors. The neural network enables to retrieve information on one preselected descriptor, for example, location, in the form of the final result of the name of geographical objects. In a general case, the neural network can retrieve information on several descriptors simultaneously.