Laboratories


Description

The lab component of the Speech Processing course aims to supplement the theoretical material discussed in class. It provides students with hands-on experience in applying the topics covered in the lectures to real-world challenges.

The labs involve programming exercises in jupyter notebooks for students to explore and experiment with various concepts and tools for speech processing.

The labs are designed to develop problem-solving skills and an understanding of the practical aspects of speech processing. Furthermore, the labs provide a platform for students to get creative and develop their own solutions to problems.

Lab attendance is mandatory!

Materials and Resources

In the Spoken Language Processing lab students will use the Python programming language together with several Python packages to illustrate some of the theoretical concepts discussed in class and to apply them in a specific challenge. The lab deliverables include Jupyter notebooks to train the students in generating more reproducible results through a combination of live code, text, images, and mathematical equations.

Pre-requisites

Basic knowledge of the Python programming language is advised as well as of Markdown basic syntax and in writing mathematical equations in LaTeX.

Ethics

Students are expected to behave ethically both in and out of the lab. Unethical behavior includes, but is not limited to, the following:
  • Reference to, or use of another person's laboratory solutions from the current or previous years.
  • Submission of work that is not done by your laboratory group, except when using open-source code, acknowledging its authors and origin (URL).
  • Allowing another person to copy your laboratory solutions or work.

Lab Enrollment


The lab enrollment will be done in the Fenix system in groups of 2 students in atomic mode. This means that the enrollment is performed by one of the members of the group in the representation of both students.
If a student enrolls another student without his or her consent, the student enrollment will be canceled and will be manually enrolled by the lab instructor after all the other students are enrolled. To prevent any misunderstandings, the student that enrolls in the group should ask for the agreement of the other group member in the form of an email message, for example. 

The lab enrollment will be open from 18:00 May 2nd (Tuesday) till 18:00 May 4th (Thursday). Students without a group by that time may ask the lab instructor to be assigned to a group with another unrolled student.

Lab Assignments


Lab assignments will be mostly based on Jupyter Notebooks. It may include a theoretical introduction and sample code and students will fill in their answers and proposed solutions in text and code cells, as well as the results produced by the code.

Lab deliverables are the Jupyter notebooks with the student's answers and code, as well as a PDF file with the printed version of the document including the code output after the code cells.
The lab assignment may also ask students to provide additional materials such as audio files, a summary of the achieved results, a short document explaining the results, etc.

There will be 3 lab assignments
  • L1 - Speech signal processing (LCO) - Due date: Sunday, May 21st, 21:00
  • L2 - Speech recognition/classification (AA) - Due date: Sunday, June 4th, 21:00
  • L3 - Dialogue systems (BM) - Due date: Sunday, June 18th, 21:00

Lab Grading


Each lab assignment will be graded on an integer scale of 0 to 20.

The group grade will be the average of the 3 assignments' grades submitted by the group rounded to 2 decimal places.

The penalty for missing a lab session will be:
  • 1 missed session: no penalty
  • 2 missed sessions: 10% penalty
  • 3 missed sessions: 20% penalty
  • 4 missed sessions: 30% penalty
  • 5 missed sessions: 50% penalty
  • 6 missed sessions or more: 100% penalty
In case of evidence that the workload was not evenly distributed among group members, students may be called for an oral assessment. This may result in an additional penalty that reflects the lack of student involvement.

The student's lab grade will be the group's grade multiplied by 1-penalty.

The student fails the lab component if the student's lab grade is less than 7.50.

Lab Calendar

Week1
  • L02-Lab00 (May 2nd) - Lab enrollment
  • L02-Lab00 (May 3rd) - Lab enrollment
  • L03-Lab00 (May 3rd) - Lab enrollment
  • L03-Lab00 (May 4th) - Lab enrollment
Week2
  • L1 publication (May 8th)
  • L02-Lab01 (May 9th) - Speech signal processing 1 (L1)
  • L02-Lab02 (May 10th) - Speech signal processing 2 (L1)
  • L03-Lab01 (May 10th) - Speech signal processing 1 (L1)
  • L03-Lab02 (May 11th) - Speech signal processing 2 (L1)
Week3
    • L02-Lab03 (May 16th) - Speech signal processing 3 (L1)
    • L02-Lab04 (May 17th) - Speech signal processing 4 (L1)
    • L03-Lab03 (May 17th) - Speech signal processing 3 (L1)
    • L03-Lab04 (May 18th) - Speech signal processing 4 (L1)
    • L1 Deadline (May 21st)
    Week4
    • L2 publication (May 22nd)
    • Tecnico Day (May 23rd) - no classes
    • L02-Lab05 (May 24th) - Speech recognition/classification 1&2 (L2)
    • L03-Lab05 (May 24th) - Speech recognition/classification 1 (L2)
    • L03-Lab06 (May 25th) - Speech recognition/classification 2 (L2)
    Week5
    • L02-Lab06 (May 24th) - Speech recognition/classification 3 (L2)
    • L02-Lab07 (May 31st) - Speech recognition/classification 4 (L2)
    • L03-Lab07 (May 31st) - Speech recognition/classification 3 (L2)
    • L03-Lab08 (June 1st) - Speech recognition/classification 4 (L2)
    • L2 Deadline (June 4th)
    Week6
    • L3 publication (June 5th)
    • L02-Lab08 (June 6th) - Dialogue systems 1 (L3)
    • L02-Lab09 (June 7th) - Dialogue systems 2 (L3)
    • L03-Lab09 (June 7th) - Dialogue systems 1&2 (L3)
    • Holiday (June 8th) - no classes
    Week7
    • Holiday (June 13th) - no classes
    • L02-Lab10 (June 14th) - Dialogue systems 3&4 (L3)
    • L03-Lab10 (June 14th) - Dialogue systems 3 (L3)
    • L03-Lab11 (June 15th) - Dialogue systems 4 (L3)
    • L3 Deadline (June 18th)