Lecture 1: Introduction to the Course (Practical Information)

Machine Learning for Language Technology 2015http://stp.lingfil.uu.se/~santinim/ml/2015/ml4lt_2015.htm

Introduction to the CourseThe Flipped Classroom Model

Marina [email protected]

Department of Linguistics and PhilologyUppsala University, Uppsala, Sweden

Autumn 2015

http://stp.lingfil.uu.se/~santinim/ml/2015/ml4lt_2015.htm

mailto:[email protected]

Course Website & Contact Info:http://stp.lingfil.uu.se/~santinim/ml/2015/ml4lt_2015.htm

• Contact details:

– [email protected]



Lecture 1: Introduction to the Course 2

http://stp.lingfil.uu.se/~santinim/ml/2015/ml4lt_2015.htm




Students’ Email addresses


Make sure you send your email addresses to me.

Send me the email address that you check daily.

The communication via email is important for this course.

Once I got your email, you will receive my reply and the enrolment key to access the Scalable Learning platform.

Outline

• Attendance

• Examination

• Course Organization

• Course Structure

• The Flipped Classroom

• Learning Outcomes

• Course Website and Summary


ATTENDANCE


Attendance is Mandatory

• There is a mandatory 80% attendance requirement for both:– presentations delivered through the online platform– in-class lab sessions

• The whole course is made of 11 lectures.• 80% attendance means that you can miss a couple of

lectures – ie: 11:100=x:80x=80*11/100x=8.880%=8.89 (rounded

up)

• In other words, you should attend 9 lectures out of 11 to be eligible to pass the course


More than 20% absence may result in additional assignments

• Plan carefully, if you already have committments.

• Notify me if you already know that you cannot comply with the 80% attendance requirement


Attendance is mandatory because...

• ...this course is quite tough for linguists.

• It contains math and statistics, and requires practice with a software workbench...

• Regular attendance is helpful to smooth difficulties.


Lectures are paired: online presentation + lab session = 1 lecture

• 80% virtual attendance to online presentations (minimum: 9 out of 11 presentations)

• 80% physical attendance to in-class lab sessions (minimum: 9 out of 11 lab sessions)

• If you do not attend an online presentation, there is no point in attending the matching lab session because lectures are paired.


Attendance: Paired lectures

• Not valid: – online presentation=no;

lab=yes;

• Not valid: – online presentation=yes;

lab=no;

• Valid: – online presentation=yes; lab=yes.


+ paired lab

+ paired lab

+ paired lab

+ paired lab

+ paired lab

+ paired lab

+ paired lab

+ paired lab

[...]

EXAMINATION


Examination: 9 graded INDIVIDUAL lab assignments

• The course is examined by means of 9 in-class lab assignments, from lecture 2 to lecture 11.

• All graded lab assignments have equal weight.

• Each lab assignment will be graded with the following marks: – Underkänd (U) [Fail]– Godkänt (G) [Pass]– Väl Godkänt (VG) [Distinction]

• In order to pass the course (ie to receive the passing grade G on the course), a student must submit all the lab assignments and at least 5 of them need to be a G.

• In order to receive pass with distinction (VG), the majority of all the submitted lab assignments have to meet the criteria for distinction.

• If a student fails the examination, additional assignments will be required in order to receive a passing grade on the course.


Graded assignments: single and double

• Summary: – The course is examined by means of INDIVIDUAL graded lab

assignments.– Students must complete correctly at least 5 lab assignments out of the

total number of submitted lab assignments.

• 8 single assignments: Students will sit for a lab assignment every lab session. Students can submit immediately (i.e. at the end of the lab session) OR the next day, if more time for reflection is needed. The deadline of a single lab assignment is the day after the lab at 1pm.

• 1 double assignment: ”double” means that the last assignment stretches over 2 lectures. Joakim will prepare a programming lab assignment with starter code in python. The basic task will be to implement the basic perceptron. You will work on this programming assignment during the last two lab sessions of the course.


Master Students

• Must fulfil the basic requirements of the course (at least 5 Gs out of all submitted lab assignments)

• + a Home Assignment (passing grade=at least a G) [programming assignment]


COURSE ORGANIZATION


People

• Marina Santini: delivering some online presentations and responsible for the lab classes based on Weka and on statistics.

• Joakim Nivre: decided the topics of the course and delivering some online presentations. He is responsible for the programming assignment based on the perceptron.

• Mats Dallhöf: responsible for all administrative issues related to this course.


Students’ Responsibilities

• Attendance (virtual and physical)

• Reading

• Submission of lab assignments (graded)


Students MUST read

• Listen to online presentations AND read the pages associated with the lectures (see the course website).


About the Course

• Introduction to Machine Learning applied to Language Technology.

• The focus of the course is on models that are commonly used in Language Technology and NLP.

• Teaching is based on the Flipped Classroom educational strategy.


Reading: only chapters specified in the course website

• Hal Daumé III (2015). A Course in Machine Learning. Copyright © 2015.

Ian H. Witten, Eibe Frank, Mark A. Hall (2011). Data Mining: Practical Machine Learning Tools and Techniques. 3rd Edition. Morgan Kaufmann Publishers.


Lab assignments

• At each lab session, you will be given the lab assignment of the day.

• You will complete the lab tasks and you will submit either at the end of the class, or you will send the lab assignment to me by 1pm of the next day.


Computers

• Either use your laptop (recommended) or use the computers in the classroom.

• Most of the lab assignments are based on the Weka software package, which means that:– You must install the software and deal with any

issues about memory limits or errors or hardware problems. Choose the computer that you feel confident about. If you have problems on a computer, just change to another. We focus on the software package and basic IT skills.


Math

• There will be a fair amount of math and statistics in the ML4LT course.

• My main effort: to simply as much as I could mathematical and statistical concepts in the online presentations

• Your main effort is to study the theoretical underpinnings berfore engaging in lab assignments.

• Be prepared to carry out some individual study if you feel that you need to refresh basic math knowledge that is not provided in this course

• Basic requirements for this course: – Matematik 2a/2b/2c (områdesbehörighet A7) OR – Matematik B (områdesbehörighet 7)


Interaction during the lab assignments• Tricky part

• Lab assignments are individual (you will get a grade for your work), BUT you are encouraged to talk to each other and discuss in group.

• Challenge: The effort is to learn as much as you can independently by interacting in group.

• I will help as LITTLE as I can during the labs: I will not give you hints to solutions, nor solve your computer-related problems. BUT I will act as a moderator or a facilitator in a discussion (if needed).

• LEARNING OUTCOMES OF LAB ASSIGNMENTS: – Cooperating with others to optimize your understanding of the topic– Fostering independent-thinking – Enhancing problem-solving skills – Finding the best way to show that you master the topic of the day both

practically and theoretically.


Cheating

• Any assignment that is handed in must be your own work.

• However, talking to one another to understand the material better is strongly encouraged: recognizing the distinction between cooperation and cheating is very important!

• COOPERATION with other students IS WARMLY ENCOURAGED!

• Plagiarism—copying from others—is condemned and measures will be taken if it happens.


COURSE STRUCTURE


Hybrid learning approach: virtual and physical attendance

• Online presentations Scalable Learning platform

• Lab sessions Chomsky classroom


Scalable Learning

• The platform: www.scalable-learning.com


http://www.scalable-learning.com/

Scalable Learning: your tasks• Virtual

attendance

• Listen to the video clips

• Answer the quizzes

• Read the associated chapters

• Each presentation has a deadline!


Analytics

• The aim of this e-learning platform is to understand which concepts and topics are more difficult for the students, thus enabling the teacher to provide the appropriate support.


Online Presentations, Video Clips and Quizzes

• A presentation is made of several video clips.

• The length of the presentations and the length of video clips are variable.

• The number of the quizzes per presentation is variable. Quizzes are NOT graded.


Communication and Interaction

• The platform allows both anonymous and non-anonymous communication between students and teachers.

• The aim is to create an interaction that is smooth, unproblematic and effective.

• Ask questions through the platform or by email: either you will receive an individual answer or we will discuss your questions in class.


Lab Sessions: two parts

• First, general comments on your virtual attendance to the online presentations (questions, quizzes, etc. )

• Then, you will be give a lab assignment to complete.


Lab Sessions: Lab AssignmentsExample of a Lab Assignment...


Lab Sessions: Interaction

Lab assignments are individual (you will get a grade) but cooperation among students is strongly encouraged:

• We warmly encourage students to help one another understand the assignments, the tasks, the theoretical concepts and general issues relevant to the course.

• Research shows that cooperation among students combined with individual thinking is an effective way to acquire and activate new knowledge.


FLIPPED CLASSROOM


What is a ”flipped classroom”?

• Short answer: The flipped classroom inverts traditional teaching methods, delivering theoretical knowledge online outside ofclassroom and moving exercises into the classroom.


Flipping learning is upside down

• The basic idea is to reverse the structure of traditional teaching.

• Traditional teaching usually is based on: – lectures that are delivered in a classroom by a lecturer – homework carried out by students by themselves, not in

the classroom

• With the flipped approach, we will do the opposite: – you will listen to the online presentations at home– you will be in the classroom to do your homework (that

we will call lab sessions)


The Flipped Classroom Model

• Students watch lectures at home at their own pace, communicating with peers and teachers via email or via the platform.


Learning Process

• Passive phase: that we can call the receptive phase, where the student/learner opens the mind by listening, reading and receiving new information. In this phase the student lets new knowledge come in.

• Active phase: that we can call the production phase, where the student/learner processes the new knowledge, constructs a personal concept map, creates cross-references with previous knowledge. In this phase, the student will become able to apply the new knowledge and to solve practical tasks.


Research says that …

… often with traditional teaching, where the passive phase is carried out in the classroom, learning outcomes are poor. For ex:


Thanks to Technology and eLearning…

eLearning: thanks to the availability and successof online videos used for pedagogical purposes, and the increased access to technology, it is nowpossible to stop this negative trend.


The benefits

• It allows students to personalize the learning at their own pace.

• You can replay the videos as many time as you like, you stop them and resume them if you need to look up a word in a dictionary, or if you need to brush up a concept, or if you are tired or hungry, etc.

• Therefore there is both a cognitive and physical advantage in doing the passive phase at home.


The Scalable Learning Platform

• We will use platform that has been developed in Sweden (by Swedish Institute of Computer Science and Uppsala University) and it is called Scalable Learning.

• Create your own account and sign up for the course using the enrolment key that will be sent to you.


Scalable Learning at Uppsala Uni

• The platform is already successfully used at Uppsala University.

• David Black-Schaffer (Department ofInformation Technology, UU) is regularly usingit for his own courses.

Watch David’s video presentation for motivation, aims, and outcomes.


http://www.it.uu.se/katalog/davbl791

https://www.youtube.com/watch?v=57MvwhSbv3k

Online material

• Video clips

• Quizzes

• Lecture transcripts (or Lecture notes)

• Slides (pdf)

• Slides (slideshare)


LEARNING OUTCOMES


What is a learning outcome?

• Learning outcomes describe what students areable to demonstrate in terms of knowledge, skills, and values upon completion of a course.


Expected Learning Outcomes

1. apply basic principles of machine learning to natural language data;

2. show theoretical and practical knowledge of the following machine learning methods:– decision trees– naïve bayes classifiers– logistic regression– the perceptron

3. use of a standard machine learning package for practical classification and evaluation (the Weka workbench)


COURSE WEBSITE & SUMMARY


Schedule, News and more


Check the website regularly

Summary: basic stuff

• 80% attendance requirement

• Paired lectures: 1 lecture = online presentation + matching lab session

• Lab assignments are graded; quizzes are not graded

• Examination: In order to pass the course successfully, 5 Gs is the minimum requirement (for bachelor students). Master students must submit a home assignment, in addition to the minimum requirement.

• Cooperation is encouraged, cheating is condemned.


Questions?


Lecture 1: Introduction to the Course (Practical Information)

Education