Sign in

Mark Dredze

@mdredze.bsky.social
2.8K followers 377 following 67 posts

John C Malone Professor, Johns Hopkins Computer Science Director, Data Science and AI (DSAI) Institute Center for Language and Speech Processing, Malone Center for Engineering in Healthcare. Part-time: Bloomberg LP #nlproc

PostsRepliesMedia
Reposted by Mark Dredze
JHU Computer Science @jhucompsci.bsky.social · 13/05/2026
“AI can revolutionize the way in which Americans engage in our democracy,” says @mdredze.bsky.social, who was part of the team that used artificial intelligence to analyze comments in the public record to determine how people feel about the federal government’s reclassification of cannabis.
cs.jhu.edu
Americans support cannabis rescheduling, study finds
Researchers at Johns Hopkins and UC San Diego used artificial intelligence to analyze comments in the public record.
001
Reposted by Mark Dredze
JHU Computer Science @jhucompsci.bsky.social · 07/05/2026
Most people strongly support the federal government’s reclassification of cannabis, according to a new study by @jhu.edu and @ucsandiego.bsky.social researchers including @mdredze.bsky.social and @vijay-murari.bsky.social that used AI to analyze more than 40,000 comments in the public record.
hub.jhu.edu
Americans support cannabis rescheduling, study finds
Researchers at Johns Hopkins and UC San Diego used artificial intelligence to analyze comments in the public record
022
Reposted by Mark Dredze
David Broniatowski @broniatowski.bsky.social · 09/12/2025
This NYT story about A.I. travel avatars is being treated as a novelty. “Synthetic influencers are here!” But the underlying platform architecture is the same. What’s changing is how efficiently actors can exploit it. A.I. influencers simply remove the human bottlenecks.
nytimes.com
These Travel Influencers Don’t Want Freebies. They’re A.I.
271
Mark Dredze @mdredze.bsky.social · 08/12/2025
Fantastic talk by Hopkins Medicine's Peter Najjar at the @jhumceh.bsky.social symposium on Human + AI to redefine the standard of care in medicine. What can AI do to improve quality and safety in medicine?
031
Reposted by Mark Dredze
Alexis Battle @alexisbattle.bsky.social · 08/12/2025
Loving the “Pitch a problem” session at the Johns Hopkins Symposium on Engineering in Healthcare. Lots of animated conversations! @jhumceh.bsky.social
022
Reposted by Mark Dredze
JHU Malone Center for Engineering in Healthcare @jhumceh.bsky.social · 17/10/2025
Congratulations again to John C. Malone Professor of Computer Science @mdredze.bsky.social on this accomplishment!
malonecenter.jhu.edu
Mark Dredze named director of Johns Hopkins Data Science and AI Institute
Dredze, a member of JHU’s faculty since 2009, has been selected to lead the university institute dedicated to harnessing the power of AI to translate data-driven discovery into real-world impact.
182
Reposted by Mark Dredze
JHU Computer Science @jhucompsci.bsky.social · 09/09/2025
Congratulations to CS faculty @mdredze.bsky.social, Jason Eisner, Peter Kazanzides, and @tom-lippincott.bsky.social on their @jhu.edu Nexus Awards! Learn more about their funded projects here: www.cs.jhu.edu/news/compute...
Headshots of Mark Dredze, Jason Eisner, Peter Kazanzides, and Tom Lippincott.
051
Reposted by Mark Dredze
William Jurayj @williamjurayj.bsky.social · 20/02/2025
🚨 You are only evaluating a slice of your test-time scaling model's performance! 🚨 📈 We consider how models’ confidence in their answers changes as test-time compute increases. Reasoning longer helps models answer more confidently! 📝: arxiv.org/abs/2502.13962
11710
Mark Dredze @mdredze.bsky.social · 19/02/2025
Please read and share this excellent FAQ on University indirect costs by my friend @broniatowski.bsky.social He explains why these funds are essential and a critical investment for research in the United States. www.linkedin.com/posts/david-...
linkedin.com
David Broniatowski on LinkedIn: NIH has just announced that it will save $4 billion by capping university… | 61 comments
NIH has just announced that it will save $4 billion by capping university indirect costs on federal grants. Will this actually save money? Bottom line: No. It… | 61 comments on LinkedIn
043
Mark Dredze @mdredze.bsky.social · 05/02/2025
I know I can improve my ARR reviews, but there really is no need for name calling. 😁
1180
Mark Dredze @mdredze.bsky.social · 30/01/2025
Helpful Insightful Probing Valuable Thoughtful Illuminating Constructive In author feedback, these are synonyms for "we hate your review."
010
Mark Dredze @mdredze.bsky.social · 27/01/2025
Do reviewers purposely write confusing reviews with typos to demonstrate that the review wasn't written by a LLM?
020
Mark Dredze @mdredze.bsky.social · 24/01/2025
Golden idea for an NLP paper: a group of llamas is called a "cria herd". That would make a great name for a LLM method, model, or paper. Just remember to acknowledge me in your paper. You're welcome.
090
Mark Dredze @mdredze.bsky.social · 21/01/2025
Idea for GenAI app: rewrite click bait headlines to normal headlines in the browser. Input: you’ll never guess this one company organizing the best deals of the year Output: Amazon has a modest sale on phone chargers
2232
Mark Dredze @mdredze.bsky.social · 20/01/2025
Good idea!
000
Mark Dredze @mdredze.bsky.social · 20/01/2025
The ARR submission checklist is already pretty extensive, but I suggest we add an additional question: "I certify that I know the difference between \citet and \citep."
1221
Mark Dredze @mdredze.bsky.social · 20/01/2025
ARR: Reviews are due today. Me:
010
Mark Dredze @mdredze.bsky.social · 17/01/2025
I feel seen. This is why I always access my API keys from my laptop.
0101
Mark Dredze @mdredze.bsky.social · 14/01/2025
Do you have any of those fortune cookies that mock academics? Sure!
160
Mark Dredze @mdredze.bsky.social · 02/01/2025
Starting a new year and reflecting on how lucky I am to work at @hopkinsengineer.bsky.social with amazing people @jhucompsci.bsky.social @jhuclsp.bsky.social. I was promoted to full professor in 2023, and my students presented me with this amazing poster of current and former PhD students.
0122
Mark Dredze @mdredze.bsky.social · 02/01/2025
Listen to @karaswisher.bsky.social's new podcast where she interviews @ruchowdh.bsky.social, @ghadfield.bsky.social and me about AI Ethics and Safety. The podcast was recorded before a live audience at @jhu.edu Bloomberg Center. podcasts.apple.com/us/podcast/a...
podcasts.apple.com
AI Ethics and Safety — A Contradiction in Terms?
Podcast Episode · On with Kara Swisher · 01/02/2025 · 53m
020
Mark Dredze @mdredze.bsky.social · 22/12/2024
Examining the generated QA pairs, you can really see the difference. Our generations (bottom) look harder and more interesting. Try our strategy for your synthetic generation task? Check out our paper, being presented at #ML4H2024 . arxiv.org/abs/2412.04573
020
Mark Dredze @mdredze.bsky.social · 22/12/2024
Training a Clinical QA system on our data gives big improvements, whether we generate data from Llama or GPT-4o. These improvements are both in F1 and any overlap between the extracted and true answers.
110
Mark Dredze @mdredze.bsky.social · 22/12/2024
The generated pair has a lot of advantages: it doesn't use the same language as the report, it includes harder questions, and the answers are sometimes not in the report (unanswerable questions.) The result? Harder, more diverse and more realistic QA pairs.
110
Mark Dredze @mdredze.bsky.social · 22/12/2024
Second, we use a summarize-then-generate strategy. The LLM first summarizes a given clinical record in a structured format. The summary keeps the key points but loses the details, such as specific terminology and content. We then use the summary to generate a new QA pair.
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
We explore two strategies. First, we craft instructions to encourage QA diversity. We formulate these as constraints on the answers to the questions. It helps, but we need more.
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
We can ask an LLM to write QA pairs, but they turn out to be too easy and repetitive. They don't come close to what you can get with real data. We need more diverse data! Typical methods (e.g. annealing) don't work. What can we do?
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
Paper at #ML42024! Clinical QA can help doctors find critical information in patient records. But where do we get training data for these systems? Generating this data from an LLM is hard. 🧵
130
Mark Dredze @mdredze.bsky.social · 22/12/2024
Takeaways: If you can fine-tune a model on a specific clinical domain, that's great. If you can't, you should probably use models that are better overall, even if they aren't trained on clinical data. Many more details in the paper! arxiv.org/abs/2412.05845
arxiv.org
Are Clinical T5 Models Better for Clinical Text?
Large language models with a transformer-based encoder/decoder architecture, such as T5, have become standard platforms for supervised tasks. To bring these technologies to the clinical domain, recent...
031
Mark Dredze @mdredze.bsky.social · 22/12/2024
It turns out that when you have just a little supervised data, the models trained on more data and tasks, even when out of domain, do BETTER on the new clinical domain.
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
Maybe the real advantage for domain-tuned models lies in the low resource setting. With lots of supervised data, an out of domain model can do well. What about with just a few training examples?
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
We try a new clinical task and dataset/domain. In this case, the clinical T5 benefits disappear.
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
Comparing 2 clinical with 3 general models on 6 clinical datasets, we find that some clinical models improve. However, these clinical test sets come from the same domain as the clinical training data. Maybe the clinical models are better on THIS clinical data, but not in general?
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
T5 models are the workhorse of many clinical text applications (e.g. information extraction.) Several clinical T5 models have been trained using clinical data to improve performance on these tasks. Do these models work better than general T5 models?
100
Mark Dredze @mdredze.bsky.social · 22/12/2024
Are Clinical T5 Models Better for Clinical Text? That's the question we asked in our #ML4H2024 paper. Turns out clinical models may not be worth it. 🧵 arxiv.org/abs/2412.05845
130
Mark Dredze @mdredze.bsky.social · 22/12/2024
Current feeling.
030
Reposted by Mark Dredze
Conference on Language Modeling @colmweb.org · 17/12/2024
Announcement #1: our call for papers is up! 🎉 colmweb.org/cfp.html And excited to announce the COLM 2025 program chairs @yoavartzi.com @eunsol.bsky.social @ranjaykrishna.bsky.social and @adtraghunathan.bsky.social
06624
Mark Dredze @mdredze.bsky.social · 18/12/2024
Accurate.
050
Reposted by Mark Dredze
masc-conference.bsky.social @masc-conference.bsky.social · 16/12/2024
📢 Want to host MASC 2025? The 12th Mid-Atlantic Student Colloquium is a one day event bringing together students, faculty and researchers from universities and industry in the Mid-Atlantic. Please submit this very short form if you are interested in hosting! Deadline January 6th. #MASC2025
1105
Mark Dredze @mdredze.bsky.social · 12/12/2024
Live shot of me answering email.
030
Reposted by Mark Dredze
Marine Carpuat @marinecarpuat.bsky.social · 11/12/2024
ICYMI: the UMD LSC is looking for a postdoctoral fellow with an interdisciplinary research agenda in language sciences. languagescience.umd.edu/news/job-opp... If your interests connect to #NLP research that helps people communicate across languages, please reach out!
languagescience.umd.edu
Job Opportunity: Language Science Postdoc
The Maryland Language Science Center is recruiting an outstanding postdoctoral researcher for our inaugural Fellowship for Postdoctoral Scholars. The goal of the fellowship is to support early-career ...
131
Mark Dredze @mdredze.bsky.social · 11/12/2024
Congratulations to my JHU CS @jhuclsp colleague Philipp Koehn on being elected an ACL fellow! www.aclweb.org/portal/conte...
aclweb.org
ACL Fellows 2024 | ACL Member Portal
050
Mark Dredze @mdredze.bsky.social · 10/12/2024
Me sitting in the back listening to conference talks and muttering to myself.
0100
Mark Dredze @mdredze.bsky.social · 03/12/2024
Spring 2024 😄
000
Mark Dredze @mdredze.bsky.social · 03/12/2024
It's back online.
010
Mark Dredze @mdredze.bsky.social · 03/12/2024
Still working on my letters this year.
140
Mark Dredze @mdredze.bsky.social · 03/12/2024
Prompt: Where are the parents? ChatGPT: The parents might be the ones smiling warmly in the middle of the table, subtly keeping the Thanksgiving magic alive. Or perhaps they’re taking a moment in the kitchen, checking on the pie in the oven, letting the family enjoy the cozy scene they’ve created!
020
Mark Dredze @mdredze.bsky.social · 03/12/2024
Controlled panic.
0140
Mark Dredze @mdredze.bsky.social · 03/12/2024
Maybe they are the ones taking the picture?
110
Mark Dredze @mdredze.bsky.social · 03/12/2024
The most American.
360