Siamese Neural Networks for Curriculum Alignment

EVENT TIME:

• 5:30pm to 6pm: Networking
• 6pm: Doors close (building entrance/elevator get locked... sorry). Talk starts
• 7-7:30pm: Clean up and more networking
• 7:30pm: Head to Malone's Social Lounge & Taphouse Head for more networking (reservation at 7:45pm)

SPEAKER: Pablo Duboue, PhD

Dr. Duboue works on applied language technology and natural language generation. He is the director and owner of Textualization Software
Ltd., a local Vancouver AI consulting company. With a Ph.D. degree in
Computer Science from Columbia University (NYC), he is passionate
about improving society through language technology and splits his time between teaching, doing research and contributing to free
software projects. His first book "The Art of Feature Engineering" is currently in copy editing and coming Spring 2020 via Cambridge University Press: http://artoffeatureengineering.com/

ABSTRACT:
Last October, I was very fortunate to have been invited to a hackathon
on semi-automatic curriculum alignment in a crisis context (explained
below). This hackathon was held at Google San Francisco and was co-hosted by Google.org, Vodafone foundation, UNHCR, Learning Equality and UNESCO. It featured about 30 participants from a variety of backgrounds, including six teachers from two camps operated by the UN High Council on Refugees.

What is curriculum alignment?

When students move from one school system to another, due to hanging cities, provinces or countries, somebody has to align the topics the students have learned with the topics their new classmates have already learned. With this alignment, remedial lessons and better understanding of the interaction and behaviour of the new students can be accomplished.

What constitutes a crisis context?

This process, curriculum alignment, faces particular challenges for
students that have been displaced and seek refuge on camps operated by international relief organizations: the number of students that require curriculum alignment can be very large and students arrive without advance notice from a variety of jurisdictions. Some of the people at camps walk 3,000 kilometers to get to them. Moreover, the teachers at the camps have to tend to 140-student classrooms and some do not have formal training in education, which makes curriculum alignment particularly hard for them.

A San Diego non-profit contributing in the educational technology
space, Learning Equality, has worked with educators and other
stakeholders to identify curriculum alignment as an important
technology piece needed at the camps.

What did we build?

The hackathon had a number of deliverables. In this talk, I will
present the work from the team I joined, that built a Siamese neural
network for curriculum alignment. This network improved 10x from a
baseline system, using data annotated by a separate team of curricular
experts. Besides sharing my overall experience at the hackathon, I
will also discuss the details behind the developed solution, available
at
https://colab.research.google.com/drive/1S0VJI4kFQxrSQ-HNLg3p8J7l75G-wNyy
. A small demo I put together after the hackathon is also on-line at
http://textualization.com/le_hackathon2019/