Institutional Repository of Vilnius University
Buhalterinių dokumentų atpažinimas ir interpretavimas naudojant neuroninius tinklus /
Abstract
dc:descriptionRecognition and Interpretation of Financial Documents using Neural Networks In this research the recurrent neural network (RNN) structure was investigated, followed by one of its popular models – long short-term memory (LSTM). Additionally optical character recognition (OCR) technology was explored and its relation to neural networks. After research part, an experiment was conducted in which Tesseract OCR engine was used and its lithuanian language neural network model was improved. The improvement was targeted toward financial documents – the symbols were chosen which were missing in lithuanian language model, but were needed to read financial documents. When the lithuanian language model was tuned to recognise new chosen symbols. During this research several conclusions were reached: it is difficult to train RNNs and so, it is recommended to use LSTM model instead. LSTM models are being used in OCR engines, such as – Tesseract OCR. Tesseract OCR models can successfully be tuned to recognise additional symbols. Also, Tesseract OCR can read text in lithuanian financial documents, but some extra strategy is required to store the read text in some reusable format.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Mikoliūnas, Vilius,
- Contributors dc:contributor
-
- Valaitis, Vytautas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD81705385&prefLang=en_US