Subject

Deep Learning for Natural Language Processing

Course Title

Deep Learning for Natural Language Processing
Deep learning for natural language processing

Code

m23_w_028

Study Programme

—

Organizer of the study programme (unit, institute, department or division)

Faculty of Computer Science and Engineering

Degree level (first, second, third cycle)

Second cycle

Academic year / semester

9 / Winter

Number of ECTS credits

Teacher

Sonja Gievska

Prerequisites for enrolling in the course

—

10.

Objectives of the course programme (competences)

The goal of the course is for the student to become familiar with modern deep learning techniques for natural language understanding and text generation. Upon completion of the course, the student will be able to select and apply an appropriate deep neural architecture to problems in the field.

11.

Course content

Some of the topics are dedicated to familiarizing the student with the challenges and achievements in the field of natural language processing: Modeling natural languages. Affective analysis. Detection of antisocial phenomena on the web (e.g., offensive speech, fake news, hate speech, and prejudiced language). Machine translation. Text generation with applications in document summarization, conversational dialogue agents, question answering, and text style transfer. Analysis and interpretation of systems for understanding and generating text. Re-examination of ethical and moral aspects in natural language processing systems.
To address the problems in the field, modern deep learning techniques will be used: deep neural networks with attention, generative adversarial networks, graph neural networks, transformer architectures, transfer learning, and reinforcement learning.

12.

Learning methods

Lectures supported by slide presentations, interactive lectures, exercises (using equipment and software packages), teamwork, case studies, guest lecturers, independent preparation and defense of a project assignment and a seminar paper, and learning in an electronic environment (forums, consultations).

13.

Total available time

6 ECTS x 30 hours = 180 hours

14.

Distribution of available time

60 + 0 + 45 + 45 + 30 = 180 hours

15.

Forms of teaching activities

15.1.	Lectures - theoretical instruction	60 hours
15.2.	Exercises (laboratory, auditory), seminars, teamwork	0 hours

16.

Other forms of activities

16.1.	Project assignments	45 hours
16.2.	Independent assignments	45 hours
16.3.	Home study	30 hours

17.

Assessment method

17.1.	Tests	15 points
17.2.	Seminar paper / project (presentation: written and oral)	45 points
17.3.	Activities and learning	15 points
17.4.	Final exam	0 points

18.

Grading criteria (points / grade)

up to 50 points	5 (five) (F)
from 51 to 60 points	6 (six) (E)
from 61 to 70 points	7 (seven) (D)
from 71 to 80 points	8 (eight) (C)
from 81 to 90 points	9 (nine) (B)
from 91 to 100 points	10 (ten) (A)

19.

Requirement for obtaining a signature and taking the final exam

Activities 15.1 to 15.2 and 16.1 to 16.3 have been completed.

20.

Language of instruction

Macedonian and English

21.

Method for monitoring the quality of teaching

Internal evaluation and survey mechanisms

22.

Literature

22.1.	Required literature 1. Ian Goodfellow, Joshua Bengio, Aaron Courville \| Deep Learning \| MIT \| 2016 2. Jurafsky & Martin \| Speech and Language Processing \| Prentice Hall \| 2021
22.2.	Additional literature No. Author Title Publisher Year