MARC details
| 000 -LEADER |
|---|
| fixed length control field |
10948nam a22002537a 4500 |
| 008 - FIXED-LENGTH DATA ELEMENTS--GENERAL INFORMATION |
|---|
| fixed length control field |
210830s2018 |||||||f mb|| 00| 0 eng d |
| 040 ## - CATALOGING SOURCE |
|---|
| Original cataloging agency |
EG-CaNU |
| Transcribing agency |
EG-CaNU |
| 041 0# - Language Code |
|---|
| Language code of text |
eng |
| Language code of abstract |
eng |
| 082 ## - DEWEY DECIMAL CLASSIFICATION NUMBER |
|---|
| Classification number |
005 |
| 100 0# - MAIN ENTRY--PERSONAL NAME |
|---|
| Personal name |
Amr Al-Khatib |
| 245 1# - TITLE STATEMENT |
|---|
| Title |
Emotional Tone Detection in Arabic text using Deep Convolutional Neural Networks / |
| Statement of responsibility, etc. |
Amr Al-Khatib |
| 260 ## - PUBLICATION, DISTRIBUTION, ETC. |
|---|
| Date of publication, distribution, etc. |
2018 |
| 300 ## - PHYSICAL DESCRIPTION |
|---|
| Extent |
78 p. |
| Other physical details |
ill. |
| Dimensions |
21 cm. |
| 500 ## - GENERAL NOTE |
|---|
| General note |
Supervisor: Samhaa El-Beltagy |
| 502 ## - Dissertation Note |
|---|
| Dissertation type |
Thesis (M.A.)—Nile University, Egypt, 2018 . |
| 504 ## - Bibliography |
|---|
| Bibliography |
"Includes bibliographical references" |
| 505 0# - Contents |
|---|
| Formatted contents note |
Contents:<br/>Abstract ........................................................................................................................................ II<br/>Acknowledgements ................................................................................................................... IV<br/>Table of Contents ....................................................................................................................... VI<br/>List of Figures .............................................................................................................................. IX<br/>List of Tables ............................................................................................................................... X<br/>1 Introduction .............................................................................................................................. 1<br/>1.1 Emotion Models .................................................................................................................... 2<br/>1.2 Artificial Neural Networks ................................................................................................. 2<br/>1.2.1 Historical Background ..................................................................................................... 2<br/>1.2.2 ANNs Design ....................................................................................................................... 3<br/>1.2.3 Activation Functions ......................................................................................................... 5<br/>1.2.4 Backpropagation ............................................................................................................... 8<br/>1.2.5 Regularization .................................................................................................................. 11<br/>1.3 Emotion Detection Approaches ...................................................................................... 13<br/>1.4 Objectives and Contributions .......................................................................................... 14<br/>1.5 Thesis Outline ...................................................................................................................... 15<br/>VII<br/>2 Datasets for Emotion Detection in Arabic Text ............................................................... 17<br/>2.1 Collecting Data .................................................................................................................... 17<br/>2.2 Lexical Approach in Data Annotation ............................................................................ 19<br/>2.3 Automatic Data Annotation .............................................................................................. 19<br/>2.4 Summary .............................................................................................................................. 20<br/>3 Representation of Short Text .............................................................................................. 22<br/>3.1 Different Model Architectures for Distributed Word Representation ................... 23<br/>3.2 Continuous Bag of Words and Skip Gram ..................................................................... 24<br/>3.2.1 Continuous Bag of Words Model (CBOW) ................................................................. 24<br/>3.2.2 Skip Gram Model ............................................................................................................. 24<br/>4 State of The Art ....................................................................................................................... 27<br/>4.1 Application of Deep Learning to Sentiment Analysis ................................................. 27<br/>4.2 Application of Convolutional Neural Networks to Sentiment Analysis ................. 28<br/>4.2.1 Multiple CNN Model Variations for Sentence Classification .................................. 30<br/>4.2.2 Effect of Words’ Representation on System’s Performance .................................. 32<br/>4.2.3 Character to Sentence Level Representation for Text Classification ................... 34<br/>4.2.4 Arabic Sentiment Classification with CNN ................................................................ 37<br/>4.2.5 Short Text Classification with CNNs ........................................................................... 39<br/>4.2.6 Sensitivity Analysis of CNN Components ................................................................... 40<br/>4.3 Summary .............................................................................................................................. 43<br/>5 Collecting and Experimenting with the Arabic Emotions Twitter Dataset ............... 44<br/>5.1 Data Collection .................................................................................................................... 44<br/>5.2 Data Preprocessing ............................................................................................................ 46<br/>VIII<br/>5.3 Baseline Experiments and Results ................................................................................. 48<br/>5.3.1 Results using the Naïve Bayes Classifier .................................................................... 48<br/>5.3.2 Results using the Complement Naïve Bayes Classifier ........................................... 49<br/>5.3.3 Sequential Minimal Optimization ................................................................................ 50<br/>5.4 Summary .............................................................................................................................. 50<br/>6 Experiments and Building a Convolutional Neural Network for Emotion Detection ...................................................................................................................................................... 51<br/>6.1 The First Deep Convolutional Neural Network Model ............................................... 52<br/>6.2 Eight One Versus All Deep Convolutional Neural Networks .................................... 56<br/>6.3 Using Pretrained Word Vectors ...................................................................................... 59<br/>6.4 Training Word Vectors with Distributed Representation of Sentences ................ 60<br/>6.4.1 Training Word Vectors with Doc2Vec ........................................................................ 62<br/>6.4.2 Training Vectors of Stemmed Words with Doc2Vec ............................................... 64<br/>6.5 Experiment Doc2vec Training Method of Word Vectors with English Datasets .. 65<br/>6.5.1 Experiment Doc2vec Training Method of Word Vectors with SST (Fine Grained) ..................................................................................................................................... 66<br/>6.5.2 Experiment Doc2vec Training Method of Word Vectors with SST (Binary) ..... 67<br/>6.5.3 Experiment Doc2vec Training Method of Word Vectors with TREC ................... 68<br/>6.5.4 Comparison with Other Methods ................................................................................ 69<br/>6.6 Summary .............................................................................................................................. 70<br/>7 Conclusion and Future Work .............................................................................................. 71<br/>8 References ............................................................................................................................... 73 |
| 520 3# - Abstract |
|---|
| Abstract |
Abstract:<br/>Emotion detection in Arabic text is an emerging research area, but the efforts in this new field have been hindered by the very limited availability of Arabic datasets annotated with emotions. In this thesis, we review work that has been carried out in the area of emotion analysis in Arabic text. We also review the work that has been done in sentiment analysis using convolutional neural networks, as it is closely related to the task of emotion detection and has yielded very interesting results.<br/>The efforts and methodologies followed to collect, clean, and annotate an Arabic tweet dataset for experimentation and evaluation is presented. Preliminary experiments carried out on this dataset are described. The results of these experiments are provided as a benchmark for future studies and comparisons with other emotion detection models.<br/>This work is aimed at exploring deep learning as a better model for detecting emotions in Arabic text. The proposed work experimented with two deep learning models. The first model is composed of a convolutional layer with max-pooling function followed by three fully connected layers and soft-max output layer, while the second model is a combination of eight convolutional neural networks all of which are similar to the first model but which employ one versus all strategy for classification (each neural network predicts one emotion versus all the rest emotion classes). The results of the two models did not meet our expectations (about 52% and 47% overall accuracies for the first and the second models respectively), and due to the computational cost of the second model it was omitted from further experiments.<br/>Observing that the results of these models fell below expectations, we created a new approach to train word embeddings for the proposed models. We call this approach, class directed embeddings. In this approach, all words from the same emotion class share a common class vector that acts as the context for that class. As a result, words that are used<br/>III<br/>distinctively within a particular class, will bear vectors that are closer to each other in the embedding space and will be more discriminative towards that class. We then tested our convolutional neural network with the new set of word embeddings, and the results were surprisingly high.<br/>To validate this novel approach, it was applied to English datasets that have been widely tested for other text classification tasks such as sentiment analysis and questions classification. In most cases, the results obtained using the approach proposed in this thesis, outperformed the stated of the art. |
| 546 ## - Language Note |
|---|
| Language Note |
Text in English, abstracts in English. |
| 650 #4 - Subject |
|---|
| Subject |
Wireless Technologies |
| 655 #7 - Index Term-Genre/Form |
|---|
| Source of term |
NULIB |
| focus term |
Dissertation, Academic |
| 690 ## - Subject |
|---|
| School |
Wireless Technologies |
| 942 ## - ADDED ENTRY ELEMENTS (KOHA) |
|---|
| Source of classification or shelving scheme |
Dewey Decimal Classification |
| Koha item type |
Thesis |
| 650 #4 - Subject |
|---|
| -- |
327 |
| 655 #7 - Index Term-Genre/Form |
|---|
| -- |
187 |
| 690 ## - Subject |
|---|
| -- |
327 |