Sasha Rush

@srush_nlp

NLP/ML researcher. Cornell Tech. arush@cornell.edu

New York, NY
Vrijeme pridruživanja: prosinac 2015.

Tweetovi

Blokirali ste korisnika/cu @srush_nlp

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @srush_nlp

  1. proslijedio/la je Tweet
    30. pro 2019.

    Looking to hire a developer/intern in the NYC area. Here is a quick (approximate) description. Please apply here: DM (or email) me with any questions. Please RT to spread the word! 🙏

    Prikaži ovu nit
    Poništi
  2. 11. pro 2019.

    My group is looking for PhD students this year. We're researching text generation, struct. prediction, generative NLP / vis, and compression. Also excited to hear what interests you. I'm at Cornell Tech in NYC (student's view below). If you are interested, apply to Cornell CS.

    Poništi
  3. 4. pro 2019.

    Dear NLP, \aclfinalcopy % Uncomment this line to make Overleaf work 🤯💪🧙‍♀️

    Poništi
  4. 25. stu 2019.

    One more alignment plot. These require some of the new sparse matrices calculations to scale. (otherwise they push the limits of pytorch internals)

    Prikaži ovu nit
    Poništi
  5. 25. stu 2019.

    GPUs are still wild to me. Let's just casually parse a 1500 word sentence in python/colab in half a second.

    Prikaži ovu nit
    Poništi
  6. 25. stu 2019.

    PyTorch-Struct (v0.4 ). Library `genbmm` for speed/memory improvements. New CUDA kernels for log-space matmul and banded sparse matrices ()

    Prikaži ovu nit
    Poništi
  7. 22. stu 2019.

    Looking for an intern/programmer for the spring to work on some projects in the NYC area. Find me at or send me an email.

    Poništi
  8. 8. stu 2019.

    Lisa interned with us last summer and did incredible work, (and taught us all a lot about NLP). So excited to see this. Not that this needs saying, but you should probably admit her to grad school....

    Poništi
  9. proslijedio/la je Tweet

    Summarization is one of the most important & least solved tasks in Problem with all models: they are not optimized for factual correctness We introduce a new task, dataset and model. Work by Paper

    Poništi
  10. proslijedio/la je Tweet
    29. lis 2019.

    What can we learn from a set of 3000 birth stories? 🐣 , , and I explore an online community's shared understanding of childbirth through a computational analysis of narrative patterns and power dynamics.

    Prikaži ovu nit
    Poništi
  11. proslijedio/la je Tweet
    29. lis 2019.

    New paper by salesforce on learning a model-based fact checker for Abstractive Summarization. Definitely a much needed evaluation approach, let's hope that these kinds of metrics will become a new standard. Link:

    Poništi
  12. 28. lis 2019.
    Prikaži ovu nit
    Poništi
  13. 28. lis 2019.

    PyTorch-Struct (v0.3 ). New features: autoregressive models / beam search, sparse-max dp, alignment/dtw, parallel semi-markov, k-max, pretty docs () Fun example: gradients of time-warping crf under different semirings.

    Prikaži ovu nit
    Poništi
  14. 28. lis 2019.

    Anyone have a list of the accepted NLP workshops? Just curious.

    Poništi
  15. proslijedio/la je Tweet
    26. lis 2019.

    Our ACL tutorial on interpretability got accepted! Please help us crowdsourced potential topics you'd like to see!

    Prikaži ovu nit
    Poništi
  16. 14. lis 2019.

    New analysis tool from folks at MIT/IBM and Harvard NLP.

    Poništi
  17. 10. lis 2019.
    Poništi
  18. 10. lis 2019.

    PyTorch-Struct (v0.2 ). CRF distributions API, documentation, parsing datasets, new structured models, Tree/Span-LSTM, DGL adapters, and perf. Fun example: RL for learning tree network over math (ListOps)

    Poništi
  19. 10. lis 2019.

    Here's the multiheaded tensor tutorial for Named Tensor (Also follow Richard Zou who spec'd / implemented this.)

    Poništi
  20. 10. lis 2019.

    Not sure if there are docs yet, but they put a lot of thought into it that improves on the original version.

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·