Застосування навчання з підкріпленням для побудови рекомендаційної системи

Юрій Щербина, Богдан Романюк, Ольга Пелюшкевич


This paper considers the problems that arise while recognizing symbols for low-quality images with a high level of digital noise, blurring, distortion of digital processing. In this work a new dataset proposed, which consists of synthetic images with overlaying text on a white background. A lot of distortions were obtained via applying resize functions, noise functions, blurring and rotation operators. Applied transformations have random
intensity uniformly distributed. It simulates the image distortions that occur in real life. Creating data in this way allows getting labeled low-quality images of text in any quantity. We used Keras library to build a CRNN and instantiate a custom endpoint layer for implementing CTC loss. A novel pipeline of data preprocessing is suggested for data
preprocessing in order to increase the accuracy of OCR results: an algorithm for horizontal alignment of the text image and an algorithm for cutting a multi-line text image into several
single-line text images. It allows us to reuse model that is suitable for single-line text and
achieve similar accuracy score on the multi-line text images.
We dened an error metric to compare character sequences of labels and model predictions as the ratio of the Levenstein distance between the label and the model prediction
to the label length. That score expose how often model mismatch single character. The value 0.02 of this metric was obtained for our model while recognizing the text of the test dataset. Testing was also performed for state-of-art OCR models Tesseract OCR v 5.0.0.
(alpha) and Google Cloud Vision APIs. The results prove that the built neural network architecture and the described image preprocessing algorithms are eective for recognizing
text of low-quality images.


Повний текст:



1,Chi Wai Lau News Recommendation System Using Logistic Regression and Naive Bayes
Classiers / Chi Wai Lau // Citeseer. 2011.
2. Hejazinia M. Accelerated learning from recommender systems using multi-armed bandit
/ M. Hejazinia, K. Eastman, S.Ye, A. Amirabadi, R. Divvela / arXiv. 2019.
3. Karani D. Electronic Sources: Introduction to Word Embedding and Word2Vec / D. Karani
2018, Available from: https://towardsdatascience.com/introduction-to-wordembedding-
4. Kawale J. Electronic Sources: A Multi-Armed Bandit Framework for Recommendations at
Netix / J. Kawale, E. Chow. 2016. Available from: https://info.dataengconf .com/hubfs/
SF 18. Slides/DS/A Multi-Armed Bandit Framework for Recommendations at Net-
5. Koehrsen W. Electronic Sources: Neural Network Embeddings Expained / W.Koehrsen
2018. Available from: https://towardsdatascience.com/neural-network-embeddingsexplained-
6. Koren Y. Matrix factorization techniques for recommender systems / Y.Koren, R. M. Bell,
C.Volinsky / IEEE Computer. 2009. Vol. 42, 8. P. 3037.
7. Li L. A contextual-bandit approach to personalized news article recommendation / L. Li,
W. Chu, J. Langford, R. E. Schapire //WWW 2010, Raleigh, North Carolina, USA, April
2630. 2010. P. 661670.
8. Lillicrap T.P. Continuous control with deep reinforcement learning / T.P. Lillicrap,
J.J. Hunt, A. Pritzel, N. Heess, T. Erez, Y.Tassa, D. Silver, D. Wierstra // CoRR. 2015.
Vol. abs/1509.02971.
9. Linden G. Amazon.com recommendations: Item-to-item collaborative ltering / G. Linden,
B. Smith, J.York // IEEE Internet Computing. 2003. Vol. 7, 1. P. 7680.
10. Liu F. Deep Reinforcement Learning based Recommendation with Explicit User-Item Interactions
Modeling / F. Liu, R.Tang, X. Li, W. Zhang, Y.Ye, H. Chen, H. Guo, Y. Zhang
// arXiv. 2019.

DOI: http://dx.doi.org/10.30970/vam.2021.29.11016


  • Поки немає зовнішніх посилань.