Skip to content

BSc Thesis - Convolutional neural network for text classification in TensorFlow

License

Notifications You must be signed in to change notification settings

bornabesic/cnn-text-classification

Repository files navigation

PDF | Croatian Digital Theses Repository

Analiza teksta pomoću dubokih neuronskih mreža

Cilj ovog rada je opisati primjenu dubokih neuronskih mreža u klasifikaciji odsječaka teksta različite duljine s obzirom na jednu pripadajuću kategoriju. U prvom dijelu objašnjavaju se motivacija za korištenje modela u vektorskom prostoru te osnovni koncepti u pozadini Word2vec modela. Predstavljena je i detaljno analizirana korištena arhitektura konvolucijske neuronske mreže s naglaskom na inicijalizaciju težina, aktivacijsku funkciju, metodu optimizacije, funkciju pogreške te regularizaciju. U drugom dijelu opisani su korišteni skupovi podataka za treniranje modela neuronske mreže te su navedeni eksperimentalni rezultati dobiveni koristeći TensorFlow biblioteku u okviru programskog jezika Python.

Text Analysis Using Deep Neural Networks

The aim of this paper is to describe the use of deep neural networks in the single-label classification of various length texts. In the first part, the motivation for vector space models and the basic concepts behind the Word2vec model are explained. Then the used convolutional neural network architecture is analyzed in detail, with emphasis on weights initialization, activation function, optimization method, error function and regularization. In the second part, the datasets used to train the neural network model are described and experimental results are reported. The results were obtained using the TensorFlow library and the Python programming language.

About

BSc Thesis - Convolutional neural network for text classification in TensorFlow

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published