This article describes excerpts from the reorganization of some theoretical, mathematical principles deduced word2vec in; and examines some of the common word2vec achieve, measure their accuracy and other properties, the final analysis word2vec original C code; the status quo is not easy for Java implementation, the original C program ported to Java. Limited time and level, this article does not talk about its development history unfold, only records the necessary knowledge and focus on engineering practice. Although my Java program rate 2 times higher than the original C program, under the same algorithm code with the original C program is still slightly lower than the accuracy of the original case C program (of course, it is still the highest accuracy Java implementation), so not perfect, could be improved. The theoretical part of this large number of reference "...
Continue reading : yard farm >> word2vec principle of derivation and code analysis
Original link : http://www.hankcs.com/nlp/word2vec.html
Reproduced in: https: //my.oschina.net/hankcs/blog/715509