Antti Ajanki

@anttiajanki

Data science, functional programming, complex systems

Vrijeme pridruživanja: lipanj 2014.

Tweetovi

Blokirali ste korisnika/cu @anttiajanki

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

  1. Prikvačeni tweet
    11. sij

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

    Poništi
  2. 26. sij

    Updated my Finnish POS and lemmatization comparison : * includes new spaCy Finnish model * batched evaluation makes most models run much faster * auxiliary verbs as a distinct tag * tokenization fixes

    Prikaži ovu nit
    Poništi
  3. 6. sij

    "You look like a thing and I love you" by is a delightful introduction to AI. It studies cases where algorithms fail, often with silly and/or terrifying consequences.

    Poništi
  4. 5. sij
    Poništi
  5. proslijedio/la je Tweet
    2. sij

    The 2010s were an eventful decade for NLP! Here are ten shocking developments since 2010, and 13 papers* illustrating them, that have changed the field almost beyond recognition. (* in the spirit of and , exclusively from other groups :)).

    Prikaži ovu nit
    Poništi
  6. 2. sij
    Poništi
  7. 28. pro 2019.

    A great explanation on what transformers are and how they are applied on sequence modeling:

    Poništi
  8. 22. pro 2019.

    In exploratory search, relevant results are those that reveal available options and possible query refinements. Relevance can't really be assigned to individual results but to the whole result set.

    Poništi
  9. 21. pro 2019.

    Suomen kielen sanojen taivutus: 😋 oppilaiksi → oppilaki perusteluissa → perusteluu aikajanoja → aikajano ääri-ilmiöissä → ääri-ilmiyö asiantuntijoiksi → asiantuntijoki perustuvan → perustupa aurinkoamme → aurinkoamme

    Poništi
  10. proslijedio/la je Tweet
    30. stu 2019.

    Overhyped claims about AI have contributed to past AI winters.  fears that we could be headed down that same path again.  Here's what we can do to stop it.

    Poništi
  11. proslijedio/la je Tweet

    A history of procedural text generation:

    Poništi
  12. proslijedio/la je Tweet
    12. stu 2019.

    On the linguistic patterns of "her/hers" and "him/his" which results in data with gender imbalance: “linguistic difference that is not inherently biased could still result in a biased machine learning model.” — Robert Munro

    Poništi
  13. 3. stu 2019.

    Granted, I'm on CPU and processing one sentence at a time. GPU and longer documents would hopefully reduce the required time.

    Prikaži ovu nit
    Poništi
  14. 3. stu 2019.

    Lemmatization is expensive. Halving the error rate requires increasing the computational effort more than 1000-fold!

    Prikaži ovu nit
    Poništi
  15. 3. stu 2019.

    Comparing Finnish part-of-speech tagging and lemmatization algorithms. Turku neural parser, FinnPos and StanfordNLP are the most accurate in my tests.

    Prikaži ovu nit
    Poništi
  16. 26. lis 2019.

    A comprehensive review of document embedding methods:

    Poništi
  17. proslijedio/la je Tweet
    19. lis 2019.

    Evaluation metrics are crucial to progress in machine learning. In light of recent interest in transfer learning in NLP, gives a comprehensive overview of some of the most important evaluation metrics in language modeling.

    Poništi
  18. 12. lis 2019.

    My colleague has been training hate speech detectors on multiple languages. She gave a presentation of what she has learned so far at Check out the slides!

    Poništi
  19. 5. lis 2019.
    Poništi
  20. 23. ruj 2019.

    Reasons why English is not representative of all natural languages:

    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:

    ·