Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    it's that time of the week again :D
    dexterp37
    @dexterp37:matrix.org
    [m]
    I have a very, very , VERY ... naive question. How does one pronounce Coqui?
    1 reply
    Coky? CoQ-y (like acquainted)?
    BΓΌlent Γ–zden
    @bozden:mozilla.org
    [m]
    Very interesting indeed... I think it depends if you mix it with Spanish accent (the species is native to Puerto Rico)...
    https://www.google.com/search?q=how+to+pronounce+coqui
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    In English the best spelling I came up with to show the pronunciation is: co-key
    Last syllable stressed
    BΓΌlent Γ–zden
    @bozden:mozilla.org
    [m]
    The best pronunciation is in their native language thou 🀩
    1 reply
    dexterp37
    @dexterp37:matrix.org
    [m]
    Thank you folks!
    BΓΌlent Γ–zden
    @bozden:mozilla.org
    [m]
    Exactly :)
    skeilnet
    @skeilnet:mozilla.org
    [m]

    I've released my most impressive NLP domain thus far! It's really small and contains only two intents. Don't let it fool you though, this bad boy can answer a huge amount of factual question it never even saw before.

    ->  search the web
    What would you like to search?
    ->  When did Albert Einstein died?
    Searching Google for when did albert einstein died?
    18. april 1955
    ->  Who was Albert Einstein's mother?
    Looking up who was albert einstein's mother using Google.
    Pauline einstein
    ->  How far is the moon?
    Google: how far is the moon?
    238,855 miles
    ->  How tall is Joe Biden?
    Ok Google; how tall is joe biden?
    5 feet, 11.65 inches
    ->  Is Britney Spears in jail?
    Google, is britney spears in jail?
    Britney spears is not in jail
    ->  Where is Switzerland?
    Using Google to find where is switzerland?
    West-central europe
    ->  How tall is mount everest?
    Google, how tall is mount everest?
    8,848 metres
    ->

    It uses google/pegasus-xsum and deepset/roberta-base-squad2 to answer a query from the context of a web page.

    It's not perfect (yet) as it doesn't get the answer right every time but it allows to use other sources such as ddgr, bing or ask. Overall the results are jaw dropping considering how little the actual lm knows. Obviously it's not made to stand alone but rather used in conjunction with a domain like SmallTalk i.e.

    Try it for yourself! It's in my repo for NLU data under Web Search.

    1 reply
    sanjaesc
    @sanjaesc:matrix.org
    [m]
    Why would you rate it as complex? ^^
    skeilnet
    @skeilnet:mozilla.org
    [m]
    Because the only way to get this kind of answer is to to compute it based on some relevant context. You can’t cheat your way out of it πŸ˜…
    sanjaesc
    @sanjaesc:matrix.org
    [m]
    You gave it quite some context, didn't you? Distance and sun
    2 replies
    Guess the model choosing earth and not mars is just the answer with the highest probability ^^
    1 reply
    sanjaesc
    @sanjaesc:matrix.org
    [m]
    And then extracted the required passage right?
    1 reply
    skeilnet
    @skeilnet:mozilla.org
    [m]
    But the point is to be able to do this at scale and fast for almost all queries 😍
    sanjaesc
    @sanjaesc:matrix.org
    [m]
    Guess you know haystack
    2 replies
    How do you crawl the pages?
    2 replies
    skeilnet
    @skeilnet:mozilla.org
    [m]
    BΓΌlent Γ–zden: You might be interested ^^
    BΓΌlent Γ–zden
    @bozden:mozilla.org
    [m]
    I'll check it out, thanks. After a lot of search I had to invest in Scrapy which is rather performing, although you spend too much time per site/page...
    2 replies
    BΓΌlent Γ–zden
    @bozden:mozilla.org
    [m]
    Yep, they are different purpose-wise... Scrapy (with addition of some JS/Browser addon) is very flexible.
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    πŸŽ‰ πŸŽ‰ πŸŽ‰ IT'S FRIDAY πŸŽ‰ πŸŽ‰ πŸŽ‰
    Mi
    @micriot:g3la.de
    [m]
    thorsten.mueller
    @thorsten.mueller:matrix.org
    [m]
    Thanks, but not really πŸ˜†.
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    that's in Golem! I even know Golem and I don't read German news! nice work Thorsten 🀩
    thorsten.mueller
    @thorsten.mueller:matrix.org
    [m]
    Thanks. Yes, it's one of germans most known IT newspages. And the process is described with 🐸 TTS technology ☺️.
    Prime Minister Narendra Modi launched β€˜Digital India Bhashini’ which will enable easy access to the internet and digital services in Indian languages, including voice-based access, and help the creation of content in Indian languages. The key intervention in building AI based language technology solutions for Indian languages will be the creation of multilingual datasets. Digital India Bhashini will enable massive citizen engagement to build these datasets through a crowdsourcing initiative called 'Bhasha Daan.'
    jmukiibi
    @jmukiibi:mozilla.org
    [m]
    The SEPIA offline #speechrecognition server v1.0.0 has been released πŸ₯³ and fully supports
    coqui_ai ASR models now πŸΈπŸ™‚. https://github.com/SEPIA-Framework/sepia-stt-server
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    Talking about Coqui / Voice tech on the Practical AI podcast :D
    josh 🐸
    @josh-coqui:matrix.org
    [m]

    πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰
    πŸŽ‰πŸŽ‰πŸŽ‰ FRIDAY IS HERE πŸŽ‰πŸŽ‰πŸŽ‰
    πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰πŸŽ‰

    https://www.youtube.com/watch?v=1TewCPi92ro

    reuben
    @reuben_m:matrix.org
    [m]
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    1 reply
    AHMAD-AFFANDI
    @chaalinx_twitter
    Hello im new here..
    How to access if limit to create sample? Can i subscribtion?
    1 reply
    thorsten.mueller
    @thorsten.mueller:matrix.org
    [m]
    I am not sure what you mean.
    ramfish
    @ramfish
    Hi ! I am building a toy based on RPI Zero 2@ and would like to incorprate basic STT. Keyword recognition + one extra word (out of 20-30). And it should run offline. Any thoughts or suggestions on how Coqui STT can do that ?
    1 reply
    @josh-coqui:matrix.org Any suggestions?
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    Hi ramfish β€” you might post the question in the STT room πŸ‘
    wannaphong
    @wannaphong:matrix.org
    [m]
    Hello everyone!
    I have some question.
    Sorry, I just found TTS room. I will move room.
    josh 🐸
    @josh-coqui:matrix.org
    [m]
    It's that time of the week again πŸŽ‰
    i hereby challenge anyone to watch that without smiling!! 😁