Neural Net Model to Identify Author Age Imitation with Easy Interpret Results
https://doi.org/10.1134/S2304487X20020121
Abstract
A method for interpretation of results of identification age style imitation in the text based on a gated recurrent unit (GRU) with an additional network layer to map the activity of the hidden layer for each word in the text has been proposed. At the same time, the special corpus has been collected for the task of age imitation identification. The corpus contains three types of texts: texts written in the author’s natural style, texts with imitation of a younger person style, and texts with imitation of an older person style. The texts have been presented in a segmented form, as words and sentences, and their morphological analysis and lemmatization have been performed using the UDPipe program. The network topology includes: an internal bidirectional GRU layer of 32 neurons providing activity for each word of the document, which is an input of a fully-connected layer with the ReLU activation function and size of 32, which connected to another fully-connected layer with the hyperbolic tangent activation function and 3 neurons (just as the number of age imitation classes). An additional interpretive layer returns the coefficients determining the class to which the text belongs. The analysis of the experiments has revealed that the characteristic features for determining the age imitation type in the text are the beginning and greeting used by a person in the text.
About the Authors
A. G. Sboev
National Research Center Kurchatov Institute; National Research Nuclear University MEPhI (Moscow Engineering Physics Institute)
Russian Federation
Russian Federation
123182
115409
Moscow
I. A. Moloshnikov
National Research Center Kurchatov Institute
Russian Federation
Russian Federation
123182
Moscow
R. B. Rybka
National Research Center Kurchatov Institute
Russian Federation
Russian Federation
123182
Moscow
A. V. Naumov
National Research Center Kurchatov Institute
Russian Federation
Russian Federation
123182
Moscow
References
1. Chung J., Gulcehre C., Cho K. H., Bengio Y. Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555, 2014.
2. Diederik P., Kingma D. P., Ba J. Adam: A method for stochastic optimization. ArXiv preprint arXiv: 1412.6980, 2014.
3. Guestrin C., Ribeiro M. T., Singh S. “why should i trust you?”: Explaining the predictions of any classifier. arXiv:1602.04938, 2016.
4. Ramakrishna A. D., Parikh V. D., Ramprasaath D. B., Selvaraju R., Cogswell M. Grad-cam: Visual explanations from deep networks via gradient-based localization. arXiv: 1610.02391, 2017.
5. Straka M., Strakova J. Tokenizing, pos tagging, lemmatizing and parsing ud 2.0 with udpipe. In Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies. Association for Computational Linguistics. pages 88–99, Vancouver, Canada, August 2017.
Review
For citations:
Sboev A.G., Moloshnikov I.A., Rybka R.B., Naumov A.V. Neural Net Model to Identify Author Age Imitation with Easy Interpret Results. Vestnik natsional'nogo issledovatel'skogo yadernogo universiteta "MIFI". 2020;9(2):189-196. (In Russ.) https://doi.org/10.1134/S2304487X20020121
Views:
106
Email this article 