Insegnamento a.a. 2022-2023

20597 - NATURAL LANGUAGE PROCESSING

Department of Computing Sciences

Course taught in English
Go to class group/s: 31
DSBA (6 credits - II sem. - OP  |  ING-INF/05)
Course Director:
DIRK HOVY

Classes: 31 (II sem.)
Instructors:
Class 31: DIRK HOVY


Suggested background knowledge

To feel comfortable in this course, you should have good knowledge of programming in Python, as well as simple linear algebra (what are vectors and matrices, how are they multiplied) and probability theory (what is a probability distribution, what is conditional probability). Additional knowledge of data structures (Counter, defaultdict) make many of the applications easier to solve.

Mission & Content Summary

MISSION

Natural Language Processing tools are becoming ubiquitous: from everyday tools like Siri or Alexa to decision making processes in industry or politics and to text analysis tools in social science research. Machine-learning based text analysis tools provide a range of possibilities and are a growing field of expertise. Whether it is the exploration of text to find structures and topics, or the construction of a classifier to predict the sentiment or author characteristics of a text, this course provides an overview and hands-on experience in all relevant techniques.

CONTENT SUMMARY

Preparation: how do I work with text:

  • Data formats.
  • Preprocessing.
  • Storage and retrieval.

Exploration: exploring structure in the data:

  • Clustering.
  • Topic models.
  • Static and contextual word embeddings.

Prediction: finding patterns to impute new values:

  • Text classification (sentiment analysis, author attributes).
  • Logistic Regression.
  • Perceptron.
  • Feed-forward Neural Nets.
  • Convolutional Neural Nets.
  • Structured perceptron.
  • Recurent Neural Nets.
  • Transformer architecure
  • BERT

Intended Learning Outcomes (ILO)

KNOWLEDGE AND UNDERSTANDING

At the end of the course student will be able to...
  • Describe different text analysis problems.
  • Talk about the linguistic foundations.
  • Distinguish between exploration and prediction approaches.
  • Know which algorithm to choose for a given problem.
  • Understand the trade-offs between different approaches.

APPLYING KNOWLEDGE AND UNDERSTANDING

At the end of the course student will be able to...
  • Apply their knowledge to practical text analysis problems.
  • Implement a variety of algorithms for text exploration and classification in Python.

Teaching methods

  • Face-to-face lectures
  • Online lectures
  • Guest speaker's talks (in class or in distance)
  • Exercises (exercises, database, software etc.)
  • Individual assignments
  • Group assignments
  • Participation in external competitions

DETAILS

  • Each lecture features hands-on exercises in Jupyter notebooks.
  • Each student completes several individual assignments to get experience in implementation details.
  • Students work together in groups to solve a joint task.
  • If applicable/available, students have the option to participate in external competitions such as Kaggle competitions or shared tasks in natural language processing.
  • If available, guest speakers from data-science companies present their work on text and language processing.

Assessment methods

  Continuous assessment Partial exams General exam
  • Individual assignment (report, exercise, presentation, project work etc.)
 x x  
  • Group assignment (report, exercise, presentation, project work etc.)
     x
  • Active class participation (virtual, attendance)
 x    

ATTENDING STUDENTS

  • Individual Assignments (50%) 

Individual projects (Jupyter Notebook) on material from first half of class. 

Each student completes an individual assignment to get experience in implementation details. 

 

  • Final Group project (50%) 

Group final project (Jupyter Notebook) on second half of class material.  

Students work together in groups to solve a joint task. 

It is graded based on the performance of the system and the quality of the report. 

 

Assessment of projects will include their clarity of presentation and performance of models used. 


NOT ATTENDING STUDENTS

  • Individual Assignment (100%) 

Individual project (Jupyter Notebook) on material from entire class.

Each student completes an individual assignment to get experience in implementation details. 

 

Assessment of project will include their clarity of presentation and performance of models used. 


Teaching materials


ATTENDING AND NOT ATTENDING STUDENTS

  • Lecture slides and notes provided on Bboard.
  • D. HOVY, Text Analysis in Python for Social Scientists: Discovery and Exploration, Cambridge University Press, 2020. (https://www.cambridge.org/core/elements/text-analysis-in-python-for-social-scientists/BFAB0A3604C7E29F6198EA2F7941DFF3)
  • JURAFSKY, DAN, J.H. MARTIN, Speech and language processing, Vol. 3. London, Pearson, 2014.
  • C.D. MANNING, H. SCHUTZE, Foundations of statistical natural language processing, MIT press, 1999.
  • S. MARSLAND, Machine learning: an algorithmic perspective, CRC press, 2015.
  • F. CHOLLET, Deep learning with Python, Manning Publications Co., 2017.
Last change 10/01/2023 14:14