Natural language processing and text classification with Bert & BiLSTM model
Επεξεργασία φυσικής γλώσσας και ταξινόμηση κειμένων με μοντέλο Bert & BiLSTM
View/
Keywords
BiLSTM ; Bert ; Transformers ; Tweets ; COVID-19 ; Optimizers ; Training ; Machine learning modelAbstract
This thesis refers to the architecture of deep learning models . This is
accomplished by using the deep learning models to classify tweets
related to Covid-19 to three different categories. The first category as
tweets which are in favor of vaccination, agree and implement methods
of protection against the virus. The second category as tweets which
are against the vaccination and promote conspiracy theories. The third
category as tweets which they have a neutral stance.
Specifically, to solve the problem above they have been collected a
number of tweets so we can train the two deep learning models we are
using, a Bidirectional LSTM and a Bert model. These models are going
to be analyzed and explained through every step and at the same time
we will emphasize some differences between those two models.
To train these models we have firstly to collect those tweets and
categorize them, also during this step we had to find another collection
method due to twitter policy changes. In addition, we will present the
preparation of those tweets so that the models can accept them and
use them to train as best as possible.
The whole preparation and the process is described in detail later on
and the results and any observations of the models at the end.
Department
Σχολή Τεχνολογιών Πληροφορικής και Επικοινωνιών. Τμήμα ΠληροφορικήςNumber of pages
72Language
EnglishCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού - Παρόμοια Διανομή 3.0 Ελλάδα
Αναφορά Δημιουργού - Παρόμοια Διανομή 3.0 Ελλάδα