Year of Graduation
Named Entity Recognition from Russian Texts
Applied Mathematics and Information Science
The paper considers the problem of Named Entity recognition from Russian texts. The proposed methodology performs machine learning approach for building an efficient name spotting system. Strength of this method is avoiding usage of hand-crafted rules encoding linguistic knowledge. The system uses linear-chain Conditional Random Field for extract entities and compares 3 methods for their classification: Naïve Bayes classifier, Support Vector Machine and Maximum Entropy Classifier. The procedure has been tested on real-life data from the Dialogue conference, which is one of the most popular Russian symposiums on computation linguistics. In this paper we have introduced the system which has 0.66, 0.88, 0.67 F scores for Organization, Person and Location classes.