EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2849. Classification for natural language processing
Invited abstract in session TA-6: Advancements of OR-analytics in statistics, machine learning and data science 12, stream Advancements of OR-analytics in statistics, machine learning and data science.
Tuesday, 8:30-10:00Room: 1013 (building: 202)
Authors (first author is the speaker)
| 1. | Ioannis Tsiligkaridis
|
| Heritage University |
Abstract
Natural Language Processing (NLP) introduces computational representations of language with the word vectors as a potent approach. This work focuses on Sentiment Analysis for classifying the emotional intent of text. A variety of different types and layers of Neural Networks (NN) are used.
A group of Deep Learning (DL) algorithms are used for classification starting from a Simple Dense Neural Network with Embedding layer (DNE) and continuing with the Convolutional Neural Network (CNN), the Long-Short Term Memory (LSTM), the LSTM with Dropdown (LSTMD), and the Ensemble model with LSTM and CNN(LSTM_CNN). Hyperparameters tuning for the NNs (i.e. learning rate, etc.) are also used. The use of hyper parameters will be able to boost the accuracy levels even higher. The grid search method is used for discovering the LSTM hyperparameters for better accuracy.
All approaches are tested for accuracy with Large Movie Review Dataset (IMDB).
In DNE, misclassification of false positive and false negative is noticed.
CNN, with one-dimensional sequence of words, improves DNE performance at classifying film reviews. The filter of the CNN layers obtained better results at learning short sequences. The LSTM surpasses LDTMD and LSTM_CNN.
Keywords
- Machine Learning
- Analytics and Data Science
- Knowledge Engineering and Management
Status: accepted
Back to the list of papers