October 2020 - nlplab - mailman.cs.vt.edu

[NLPLab] next week meeting schedule
by Lifu Huang 25 Oct '20

25 Oct '20

Monday: 10-10:30am, Sijia, check implementation progress 10:30-11am, Adheesh, check updated baseline score and implementation of new idea 2-2:30pm, Shuaicheng, check implementation progress 2:30-3pm, Ali, fake news detection dataset survey 5:30-6:30pm, reading group discussion, Adheesh lead. Wednesday: 9-10am: Kaiwen, dialogue summarization Thursday: 1:30-2pm, Sijia, check implementation progress 2-2:30pm, Ali, fake news detection dataset survey 2:30-3pm, Adheesh, check implementation progress 3-3:30pm, Shuaicheng, check implementation progress Lifu

1 0

[NLPLab] next week meeting schedule
by Lifu Huang 18 Oct '20

18 Oct '20

Monday: 10-10:30am, Sijia, check implementation progress 10:30-11am, Adheesh, get coreference baseline errors done 2-2:30pm, Shuaicheng, check implementation progress 2:30-3pm, Ali, discuss plans for fake news detection survey 5:30-6:30pm, reading group discussion, shuaicheng lead. Wednesday: 9-10am: Kaiwen, dialogue summarization Thursday: 1:30-2pm, Sijia, check implementation progress 2-2:30pm, Ali, discuss papers about fake news detection 2:30-3pm, Adheesh, discuss approaches for joint entity and event coreference 3-3:30pm, Shuaicheng, check implementation progress In your progress slides, please summarize (1) what you have done since the last meeting (2) what you will finish by the next meeting. Thanks, Lifu

1 0

gpu server
by Lifu Huang 15 Oct '20

15 Oct '20

Our gpu server is ready to use. I have created an account for Sijia, Shuaicheng an Adheesh. The user name and password is your pid. The external IP address is: 128.173.236.231 The dns entry is: nlplab1.cs.vt.edu (Both should be active now.) nlplab1 has /home mounted from a NFS server, so we will have 86 TB of disk to play with in /home. I have created a /home/data so that you can put all useful datasets, benchmarks, resources here. When you put a dataset/resource in the server other than your home dir, please add a description here so that we can track the usage of it later: https://docs.google.com/spreadsheets/d/19vy_9ReS9f50vnI5D_N4RKYY4lN920zEtuu… Please also sign-up a gpu here: https://docs.google.com/spreadsheets/d/19vy_9ReS9f50vnI5D_N4RKYY4lN920zEtuu… Each card can be signed by 2-3 students. @Ali, I can also create an account for you if you can send me your VT pid. Let me know if you have any questions and suggestions. Lifu

1 0

Next week meeting schedule
by Lifu Huang 11 Oct '20

11 Oct '20

Monday: 5:30-6:30pm: reading group discussion, Sijia lead. Wednesday: 9-10 am: Kaiwen, on dialog summarization 2:30-3:30: Ali, on potential research topics. 3:30-4:30: Sijia, on QAIE progress 4:40-5:30: Shuaicheng, on temporal relation prediction 5:30-6:30: Adheesh, on error analysis and new ideas for coreference resolution Please send me your progress slides one day before the meeting. Thanks, Lifu

1 0

Fwd: DLRL meeting of Oct. 8 - guest speaker: Dr. Lourentzou: Semi-supervised Learning tutorial
by Lifu Huang 08 Oct '20

08 Oct '20

---------- Forwarded message ---------- From: EdFox <fox(a)vt.edu> Date: Oct 1, 2020, 9:33 PM -0400 To: dlrl-g(a)vt.edu <dlrl-g(a)vt.edu> Cc: EdFox <fox(a)vt.edu>, Lourentzou, Ismini <lourent2(a)illinois.edu> Subject: DLRL meeting of Oct. 8 - guest speaker: Dr. Lourentzou: Semi-supervised Learning tutorial > Hi! > > Please spread the word about the guest speaker presentation at next week’s DLRL meeting. As usual, this will be > Thu 3:30-4:30pm in Zoom (see details below) > Thanks, eaf > > Speaker: Dr. Ismini Lourentzou > (New faculty member in CS; recall her seminar from April: > https://cs.vt.edu/News/Seminars/ilourentzou.html) > > Title: Semi-supervised Learning tutorial > > Abstract: > The recent success of deep learning can be accredited to large annotated datasets and better computational infrastructures available. However, in new domains and applications, acquiring large amounts of data is a costly and time-consuming process. Even with active learning annotation pipelines, typically, most of the data available remain unlabeled. Semi-supervised learning is a class of algorithms between supervised and unsupervised learning, that aims to model the data distribution using both labeled and unlabeled data, in order to train a better performing model than what we could obtain by using only the labeled subset. This talk will cover a broad overview of recent advances in semi-supervised learning in a coherence timeline, from self-training and co-training to FixMatch (NeurIPS 2020). Finally, we will conclude with some limitations and open problems for future research. > > Link: > https://virginiatech.zoom.us/j/94931034669?pwd=U1NYcXF3Sk1XUGdLeHJja1V4bEZt… > > PW: JCDL20 > > Meeting ID: 949 3103 4669 >

1 0

[NLPLab] next week meeting schedule
by Lifu Huang 03 Oct '20

03 Oct '20

I’m very happy to introduce a new member, Ali Taleb Zadeh Kasgari, who is a senior PhD from ECE department, will work with us for his MS in Computer science. Welcome Ali! Ali will talk about his past research in machine learning during the group meeting. Monday: 5:30-6:30pm: reading group discussion, Adheesh lead. Everyone please read papers before the meeting, and think about what are the main contributions and limitations of the work and how can they be improved and applied to our research. Wednesday: 9-10am: Kaiwen, with Yixin 2:30-3:30pm: Adheesh 3:30-4:30pm: Sijia 4:30-5:30pm: Shuaicheng Thursday: 4:30-5:30pm: group meeting, Ali will talk about his research for 30 mins. Sijia, Shuaicheng and Adheesh will briefly introduce their current research projects. Please send me your progress slides one day before the meeting. Also a quick reminder, the deadline of NAACL’20 is Nov. 23. If you want to catch it, you need to plan your progress, create an overleaf for the paper, and send the link to me. Zoom meeting link: https://virginiatech.zoom.us/j/5961845863 Thanks, Lifu

1 0