🌺Education chatbots should be *empathetic* and encouraging!
Announcing EDEN, a spoken conversation partner for English learning. Through a user study, we find that certain design choices that make EDEN more empathetic correlate with higher resilience for language learners!🧵
📣✨ I'm starting my PhD
@columbianlp
in the fall! I will be working with
@Zhou_Yu_AI
and Dr. Julia Hirschberg on dialogue system applications in mental health, education, and healthcare!! I'm very excited and thankful for everyone who has supported me along the way :)
In a conversation, you should ideally say things. But when should you say those things? And how can you teach a model to know when to say those things?
I am super hyped to present our paper at
@sigdial
2022! This is joint work with
@AshwinParanjape
. 🧵
Coming to a Findings of EMNLP near you: GPT-3 fails to demonstrate a structural understanding of noun compounds once we remove some lexical clues in our prompts.
Joint work with Riley Carlson and
@ChrisGPotts
. Paper link: 🧵
Hi, I made a thing: . This is for people who signed up for arxiv daily emails and just want the relevant papers without going through the full email!! It is not hosted publicly just yet, you would need to set up Google text to speech and run locally :)
I will NOT be at
#ACL2023NLP
next week and will experience severe FOMO, and the
#1
reason why I will be wailing into the void at some point next week is that I HAVE a Canada visitor visa I just don't have a valid visa for U.S. reentry 🙃.
I have so much anxiety over the grad school decision process that at this point I'm tempted to just open a Twitter poll and get Twitter to decide where I'll be spending the next 4-6 years of my life for me.
I'm on the faculty market!
My goal is to build language systems that we understand deeply through discovery and by design, so we can precisely control them and treat their failures.
Let's tackle this grand challenge of science and engineering together.
Should I change my publication name from Siyan Li to Li Siyan so that it says Siyan et al instead of Li et al? I do already have some publications under Siyan Li though so there may be some issues with consistency.
Hi friends! I’ve been busy working on a health tech startup, and we’re looking for our founding ML engineer. If you're passionate about building at the forefront of AI and are looking to make an outsized impact in healthcare, DM me!
Reasons why I want to do a PhD, inspired by my networking event today:
1. I want to be a professor so I kinda have to get a PhD
2. I like going to school, like, a lot
3. Research is a jolly good time
4. MOST IMPORTANTLY, I still get to be a "student" and won't need to mature
🌞Hi all! We are recruiting Native Mandarin speaker participants for an English-teaching Chatbot user study! Please retweet so that this post gains more visibility. Here's the informed consent form for the study:
I reviewed a paper that I gave a low score to. The other reviewers gave higher scores so we discussed. I was willing to increase my scores (to 2.5), but didn't make changes / respond since I wanted to see if we can reach a better agreement. Today the authors withdrew :(
As a Chinese living in the United States, both candidates not directly answering the China relationship question is honestly tiring, irritating, and terrifying. Wow. That was not reassuring at all.
Unrelated note, PLEASE MUTE PENCE.
I forgot to RSVP to the CMU Open House lol. Too late to arrange faculty meetings now but I will still get to network with the students which is a plus. Still, really can't believe that I forgot?
Hi all!! Please retweet to help reach more native Mandarin speakers learning English! We are developing a chatbot for English education that can provide grammatical and empathetic feedback for the users :)
In a conversation, you should ideally say things. But when should you say those things? And how can you teach a model to know when to say those things?
I am super hyped to present our paper at
@sigdial
2022! This is joint work with
@AshwinParanjape
. 🧵
And finally, a perfect SNL clip that describes and motivates this work. I have gotten back to re-watch this on multiple occasions during the past year. (N/N)
Excited to share that I'll be joining the
@blei_lab
at Columbia for my PhD and thankful to receive the Provost's Diversity Fellowship! Very grateful for my mentors & collaborators at UNC (
@Mohitban47
,
@uncnlp
,
@CSS_Science
) and Meta, and for the support of my friends & family :)
@boknilev
@naaclmeeting
I managed to avoid it by getting my advisor to register for the workshop paper I'm presenting at (but advisors can only be tied to up to 2 papers), but yeah this is awful :(((
Happy New Year from the hotel I'm quarantining in in China because of customs safety regulations. There are a lot to look forward to in 2021 and I am so very excited!!!!
Best way to spend quarantine time: drink booze, do code, while playing all of
@mulaney
's Netflix specials while reciting them from memory simultaneously.
#MyNameIs
Siyan and it means "aspiring to be a great and knowledgeable person" in Mandarin. My English name is Sylvia and it means "goddess of the forest". The former I can try my best to be, but the latter might be a bit hard.
I've been waiting for this for a while.
#MyNameIs
Prithviraj and it means "Earth King" in Sanskrit. A hard name to live up to but at least, hopefully, one day our names will be normalized
We instead formulate this as a continuous problem, Lead Time to Initiation prediction. Given an event horizon (2 seconds in this work), an ideal model would produce the following predictions. (2/N)
@AstroKatie
@GalaxyKate
I always address someone as "Dr." if they have a PhD (medical settings are different, I am only talking about academia here) because the person who earned the PhD really deserves to have all their hard work acknowledged.
@lucy3_li
I low-key want to host a Standup comedy for CS PhDs workshop and hire actual standup comedians to teach us how. When I'm in New York I'll look this up lmao.
@MirowskiPiotr
Hi! I'm the improvisor who asked questions at AIIDE. I would like to follow up with some questions: (1) what context exactly is fed into the chatbot during the show? (2) if the GPT2 is only trained on some movie script corpora, where would the comedy come from?
@colinraffel
@_AlecJacobson
Percy Liang's class on Developing and Understanding Large LMs (CS 324) at Stanford uses this format! I took this class as well, it is very interesting to see different perspectives on the same paper!
With Valentine's Day coming up and me seeing a lot of my friends finding their other halves I've been pondering if I should find a partner or just marry academia.