Matthew Honnibal

@honnibal

Computational linguist from Sydney and Berlin. 💫 Author of the NLP tools. 💥 Founder

Vrijeme pridruživanja: svibanj 2008.

Tweetovi

Blokirali ste korisnika/cu @honnibal

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

  1. Prikvačeni tweet
    12. stu 2017.

    Lots of people have been asking for more details about how does parsing and NER in v2.0. In this video, I describe the NER model, which features incremental parsing, a novel Bloom embeddings strategy, and deep residual CNNs.

    Poništi
  2. proslijedio/la je Tweet
    27. sij

    Come join my lab as a PhD. Less than a week left to apply! 🏃‍♀️⌚️

    Poništi
  3. proslijedio/la je Tweet
    28. sij

    Something tells me this won’t be yet another library. I love this massive upgrade to Thinc from its older incarnation.

    Poništi
  4. proslijedio/la je Tweet
    28. sij

    Prodigy built on Spacy is a godsend for anyone working in data poor domains. It allows you to annotate your data and generate a more expressive dataset or just a less sparse one. Currently works for NLP and a few separate use cases but still Check it out. Thanks

    Poništi
  5. proslijedio/la je Tweet
    28. sij

    Check out a fresh take on frontends from the creators of

    Poništi
  6. 28. sij

    Finally, Thinc also offers some extremely practical features. Have a look at the config system, but also the fact that you can easily wrap other frameworks. Thinc is designed to complement other frameworks, not replace them. 5/5

    Prikaži ovu nit
    Poništi
  7. 28. sij

    Speaking of functional programming...The conceptual model behind Thinc is also something that you might find interesting in itself. It's a different way to program backprop, using callbacks instead of a tape. 4/5

    Prikaži ovu nit
    Poništi
  8. 28. sij

    We also have type annotations on the model --- so you can use the types to plan out your model, and again, get type errors if your wiring is wrong. This works especially well with Thinc's functional programming approach. 3/5

    Prikaži ovu nit
    Poništi
  9. 28. sij

    First, the type system: Thinc was designed for type-checking from day 1, and AFAIK it's the first library to offer detailed support for numpy's API. This helps code readability, gives editor support, and catches some bugs. 2/5

    Prikaži ovu nit
    Poništi
  10. 28. sij

    Incredibly proud to release this: a new deep learning library, incorporating lessons we've learned from , and 's FastAPI. Even if you're not looking for new tools, there's lots you might find interesting conceptually. Quick thread: 1/5

    Prikaži ovu nit
    Poništi
  11. proslijedio/la je Tweet
    17. sij

    Ines Montani will be one of the keynote speakers in , Buy your ticket and meet Ines. Get your ticket at: Social Networks: Github: Linkedin:

    Poništi
  12. proslijedio/la je Tweet
    11. sij

    I published a Finnish language model for : POS tagging and dependency parsing for Finnish on !

    Poništi
  13. proslijedio/la je Tweet
    9. sij

    I've left my position at NVIDIA (it was great!) I'm starting a Dask company Wish me luck :)

    Poništi
  14. 9. sij

    Okay yeah I'm convinced: it's not actually weird to have MT as an example problem, even if it's not most people's end-goal. Someone else pointed out that the datasets are actually available too, which is a big consideration I didn't think of. Thanks all!

    Poništi
  15. 8. sij

    Am I off-base in thinking it's weird that there are so many NLP tutorials for translation? Like, approximately nobody needs to make MT models, and the people working on MT know who they are and don't need tutorials. I guess people use them for MT-analogous seq2seq tasks?

    Poništi
  16. 7. sij

    I feel like this is my fault but I'm not sure what I was supposed to do? Persistent disks are really slow, so the experiments did need a scratch disk. I guess I should have added a persistent disk as well, so the state can be backed up periodically?

    Prikaži ovu nit
    Poništi
  17. 7. sij

    Sigh, cloud computing still difficult😰. Put up a GPU instance for a remote intern on GCP. Needs to do lots of i/o, so add NVME disk. Server gets terminated after a week or so -- expected. But now the disk is wiped and a bunch of state is lost on a short internship.

    Prikaži ovu nit
    Poništi
  18. proslijedio/la je Tweet
    6. sij

    Is starting a PhD in on your list of New Year's resolutions? You can now apply for an open-topic Marie Curie TALENT scholarship again, deadline 13 Feb, start date 1 Aug. Feel free to contact me if you're interested in a PhD with me as advisor.

    Prikaži ovu nit
    Poništi
  19. 3. sij

    Woot, exited to see this released! Justin's been working on this on and off since we started working together. It's an interesting modeling problem, and applications could be super cool -- like a virtual math tutor, showing solution steps.

    Poništi
  20. proslijedio/la je Tweet
    31. pro 2019.
    Poništi
  21. proslijedio/la je Tweet
    29. pro 2019.

    2019 has been such an exciting year. It is like a dream come true to work on this awesome open-source NLP library , with the most wonderful, passionate and brilliant , , Adriane and the rest of the team!🤩

    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:

    ·