Schedule
Monday, June 24
8:00 - 9:00 am
Registration/check-in
Space: BALL-214
9:00 - 9:30 am
Welcome and introductions
Space: BALL-214
9:30 - 11:00 am
Lectures
Space: BALL-214
11:00 - 11:30 am
Coffee break
Space: BALL-214
11:30 - 1:00 pm
Tutorials
Space: DAN-416 , DAN-402
1:00 - 2:30 pm
Lunch

2:30 - 3:30 pm
Invited talk
Alexander Sasha Rush, Assistant Professor of Computer Science, Harvard University
Space: ETIC
Title: Conditional Text Generation and Pretraining

Rush will discuss recent efforts to build robust text generation systems for problems like summarization and data-to-text generation and the challenges compared to highly constrained tasks such as machine translation. The focus of this talk will be on pretraining and whether it has the potential to change the landscape of this area. Rush will discuss early experiments showing its usefulness, but argue that high-precision conditional text generation still remains a difficult task.
3:30 - 4:30 pm
Networking opportunity with the invited speaker
Space: ETIC
4:30 - 6:00 pm
Supervised project work I (Pick teams & projects)
Space: DAN-417
6:00 - 7:00 pm
Dinner

7:00 - 8:30 pm
Supervised project work II
Space: DAN-417
Tuesday, June 25
8:00 - 9:00 am
Breakfast
Space: Fox Hall
9:00 - 11:00 am
Lectures
Space: BALL-214
11:00 - 11:30 am
Coffee break
Space: BALL-214
11:30 - 1:00 pm
Tutorials
Space: DAN-416 , DAN-402
1:00 - 2:30 pm
Lunch

2:30 - 3:30 pm
Invited talk
João Sedoc, Assistant Research Professor, Johns Hopkins University
Space: Ball-214
Title: Evaluating Conversational Agents

There has been a renewed focus on dialog systems, including non-task driven conversational agents. Dialog is a challenging problem since it spans multiple conversational turns. To further complicate the problem, there are many possible valid utterances which may be semantically different. This makes automatic evaluation difficult, which is why the current best practice for analyzing and comparing dialog systems is the use of human judgments. This talk focuses on evaluation, presenting a theoretical framework for the systematic evaluation of open-domain conversational agents, including the usage of Item Response Theory (Lord, 1968) for efficient chatbot evaluation and evaluation set creation. We introduce ChatEval (https://chateval.org) a unified framework for human evaluation of chatbots that augments existing tools and provides a web-based hub for researchers to share and compare their dialog systems.

3:30 - 4:30 pm
Networking opportunity with the invited speaker
Space: Ball-214
4:30 - 6:00 pm
Supervised project work I
Space: DAN-417
6:00 - 7:00 pm
Dinner

7:00 - 8:30 pm
Supervised project work II
Space: DAN-417
Wednesday, June 26
8:00 - 9:00 am
Breakfast
Space: Fox Hall
9:00 - 11:00 am
Lectures
Space: BALL-214
11:00 - 11:30 am
Coffee break
Space: BALL-214
11:30 - 1:00 pm
Tutorials
Space: DAN-416 , DAN-402
1:00 - 2:30 pm
Lunch

2:30 - 3:30 pm
Invited talk
Kate Saenko, Associate Professor of Computer Science, Boston University
Space: Ball-214
Title: Grounding Language in Pixels: Recent Work on Neural Vision and Language Understanding

Saenko will talk about recent work on joint vision and language understanding using neural networks. The talk will cover a range of problems including referring expression grounding, visual question answering, text-to-image and text-to-video retrieval, image and video captioning, and language-based navigation. Saenko will discuss issues of feature representation, relational reasoning and explainability of the resulting models.

3:30 - 4:30 pm
Networking opportunity with the invited speaker
Space: Ball-214
4:30 - 6:00 pm
Supervised project work I
Space: DAN-417
6:00 - 7:00 pm
Dinner

7:00 - 8:30 pm
Supervised project work II
Space: DAN-417
Thursday , June 27
8:00 - 9:00 am
Breakfast
Space: Fox Hall
9:00 - 10:30 am
Lectures
Space: BALL-214
10:30 - 4:00 pm
Social event
Lowell National Historical Park – Canal Tour & Boott Cotton Mills Museum
4:30 - 5:30 pm
Tutorials
Space: DAN-416 , DAN-402
5:30 - 6:30 pm
Supervised project work I
Space: DAN-417
6:30 - 7:00 pm
Dinner

7:00 - 8:30 pm
Supervised project work II
Space: DAN-417
Friday, June 28
8:00 - 9:00 am
Breakfast
Space: Fox Hall
9:00 - 11:00 am
Lectures
Space: BALL-214
11:00 - 11:30 am
Coffee break
Space: BALL-214
11:30 - 12:15 am
Lectures
Space: BALL-214
12:15 - 1:45 pm
Lunch

1:45 - 2:30 pm
Tutorials
Space: DAN-416 , DAN-402
2:30 - 3:30 pm
Invited talk
Jason Weston, Research Scientist, Facebook
Space: Ball-214
Title: Putting together the threads of conversational AI?

Maybe we don't have enough threads yet to knit together the whole, but let's try anyway! We present our view of what is necessary for conversational AI, and the pieces we have worked on so far to get there. In particular: software (ParlAI, a unified platform for dialogue research), various neural architectures for memory, reasoning, retrieval and generation, and interactive learning, tasks for employing personality (PersonaChat), knowledge (Wizard of Wikipedia) and perception (Image-Chat), evaluation studies & techniques (dialogue NLI), and a recent competition (ConvAI2) we ran that shows unfortunately how far we still have to go.

3:30 - 4:30 pm
Networking opportunity with the invited speaker
Space: Ball-214
4:30 - 6:00 pm
Supervised project work I
Space: DAN-417
6:00 - 7:00 pm
Dinner

7:00 - 8:30 pm
Supervised project work II
Space: DAN-417
Saturday, June 29
9:00 - 11:00 am
Group project presentations
Space: ETIC
11:00 - 11:30 am
Project presentations
Space: ETIC
11:30 - 12:30 am
Closing session
Space: ETIC
Computational resources
A teaching lab with 30 GPU workstations will be available at UMass Lowell for use by the participants
Additional remote virtual GPU machines will be provided by the MIPT lab
Participants will also be able to use colaboratory
Sample Topics for Participants Projects
Participants will select from a number of projects for which initial code base will be made available. Sample topics include:
Engaging chit-chat model to keep interlocutor engaged in the conversation using Reddit
Infotainment chat model using Wizard of Wikipedia dataset
Multi-task goal-oriented chat model using MultiWOZ dataset (A Large-Scale Multi-Domain Wizard-of-Oz Dataset for Task-Oriented Dialogue Modelling)
Made on
Tilda