Named Entity Recognition Methods Research

Now, many companies have a problem of extracting characteristics values from stock items names. There is still no common solution for this problem. The problem is alike the problem of named entity recognition in text and connected with the problem of texts and parts of text embedding problem. There are some papers on these topics, however, in those papers they analyze natural language text, which could be simply splitted into words, most of which are used in texts more than once, and from which a dictionary could be built. This makes impossible to simply apply results of those papers in described problem. However, stock item names also consist of symbols and have a common structure. As well as in text, there are some hidden connections between symbols. Therefore, methods and algorithms described in those papers could be adapted to the problem of characteristics extraction. In this work, this approach is researched. The present report comprises 41 pages, 3 equations, 17 images, one table and 2 appendix. 41 links to sources are present. Keywords: characteristics extraction, machine learning methods, texts analysis, classification, information retrieval.