March 18, 2017 Dataset Open Access

Russian Distributional Thesaurus (RDT): Word Embeddings

Alexander Panchenko; Nikolay Arefyev; Dmitry Ustalov; Natalia Loukachevitch; Denis Paperno; Chris Biemann; Natalia Konstantinova

This resource is a part of the Russian Distributional Thesaurus (RDT): see http://russe.nlpub.ru/downloads and http://nlpub.ru/RDT. 

This dataset contains a large scale word embeddings model for Russian trained using the SGNS model (Mikolov et al., 2013) on a 12.9 billion word collection of books in Russian. According to the results of our participation in the shared task on Russian semantic similarity (Panchenko et al., 2015), this approach scored in the top 5 among 105 submissions (Arefyev et al., 2015). Following our prior experiments (Arefyev et al., 2015) we have selected the following parameters for the model: minimal word frequency – 5, number of dimensions in a word vector – 500, three or five iterations of the learning algorithm over the input corpus, context window size of 1, 2, 3, 5, 7 and 10 words. Parameters of the model are listed below:

• Model: skip-gram
• Corpus: a 150Gb sample of the lib.rus.ec book collection.
• Context window size: 10 words
• Number of dimensions: 500
• Number of iterations: 3
• Minimal word frequency: 5

References:

• Panchenko A., Ustalov D., Arefyev N., Paperno D., Konstantinova N., Loukachevitch N. and Biemann C. (2016): Human and Machine Judgements about Russian Semantic Relatedness. In Proceedings of the 5th Conference on Analysis of Images, Social Networks, and Texts (AIST'2016). Communications in Computer and Information Science (CCIS). Springer-Verlag Berlin Heidelberg

• Panchenko A., Loukachevitch N. V., Ustalov D., Paperno D., Meyer C. M., Konstantinova N. (2015): RUSSE: The First International Workshop on Russian Semantic Similarity. In Proceedings of the 21st International Conference on Computational Linguistics and Intellectual Technologies (Dialogue'2015). Moscow, Russia. RGGU

• Arefyev N., Panchenko A., Lukanin A., Lesota O., Romanov P. (2015): Evaluating Three Corpus-Based Semantic Similarity Systems for Russian. In Proceedings of the 21st International Conference on Computational Linguistics and Intellectual Technologies (Dialogue'2015). Moscow, Russia. RGGU