spaCy, o se faletusi masani gaosi gagana

Faʻalauiloa AI faʻailoa le tatalaina o o le fou lomiga o le leai se totogi faletusi «SpaCy»Lea e iai lona faʻatinoina o gagana masani gaosi algorithms (NLP). I le faʻataʻitaʻiga, e mafai ona faʻaaogaina le poloketi e fausia ai ni autoresponder, bot, faʻavasega tusitusiga, ma faʻavasega talanoaga e iloa ai le uiga o fasifuaitau.

Faletusi ua fuafuaina e saunia ai se faifai pea API E le fesoʻotaʻi i le algorithms faʻaaogaina ma sauni e faʻaoga i oloa moni. Faletusi faʻaaoga le alualu i luma lata mai i le NLP ma sili ona lelei algorithms avanoa e faʻagasolo ai faʻamatalaga.

Afai e aliali mai se sili atu algorithm lelei, o le faletusi pasi atu ia ia, ae o lenei suiga e le aafia ai le API poʻo talosaga.

O se vaega o spaCy o se fausaga foi fuafuaina e faʻagasolo atoa pepa, e aunoa ma le muaʻi faʻapipiʻiina i le au muamua e vaevaeina le pepa i fuaitau. O faʻataʻitaʻiga o loʻo ofoina atu i ni faʻamatalaga se lua: mo le maualuga o le gaosiga ma le maualuga o le saʻo.

O vaega taua o spaCy:

  • Lagolago mo lata ile 60 gagana.
  • Ua maeʻa aʻoaʻoina faʻataʻitaʻiga avanoa mo gagana eseese ma faʻaoga.
  • Multitask aʻoaʻoga faʻaaogaina muamua transformers pei o BERT (Bidirectional Encoder Renderings of Transformers).
  • Lagolago mo muamua-aʻoaʻo fualaʻau ma upu tuʻufaʻatasi.
  • Maualuga faatinoga.
  • Sauni e-faʻaaoga i-le-galuega faʻataʻitaʻiga faiga faʻavae.
  • Faʻalauiloaina gagana.
  • O loʻo avanoa vaega ua sauni mo fesoʻotaʻiga kamupani igoa, makaina o vaega o le tautala, faʻavasegaina o tusitusiga, auiliiliina faʻavae faʻavae faalagolago, vaeluaina o fuaiʻupu, makaina o vaega o le tautala, faʻataʻitaʻiga faʻapitoa, faʻavaeina, etc.
  • Lagolago mo le faʻalauteleina o gaioiga ma tu ma aga masani ma uiga.
  • Lagolago e fausia au oe lava faʻavae faʻavae luga o le PyTorch, TensorFlow ma isi faʻavaʻa.
  • Mea faigaluega na fausia e fai ma Faʻailogaina o Igoa ole igoa ma le Syntax Visualization (NER, Named Entity Recognition).
  • Faigofie faiga o afifiina ma deploying faʻataʻitaʻiga ma puleaina workflow.
  • Maualuga saʻo.

Faletusi o loʻo tusia i le Python ma elemeni i le Cython, o se Python faʻalautelega e faʻatagaina tuʻusaʻo galuega valaʻau i le C gagana.

Le poloketi poloketi ua tufatufaina i lalo o le laisene MIT. Gagana faʻatusa ua sauni mo 58 gagana.

E uiga i le fou lomiga o spaCy 3.0

O le spaCy 3.0 version tu mai mo le faʻatinoina o aiga faataitai toe aʻoga mo 18 gagana ma 59 laina paipa na aʻoaʻoina i le aofaʻi, aofia ai 5 fou suia-faʻavae paipa

O le faʻataʻitaʻiga o loʻo ofoina atu i ni faʻamatalaga se tolu (16 MB, 41 MB - 20 afe vector ma 491 MB - 500 afe vector) ma ua faʻalelei e galue i lalo ole CPU uta ma aofia ai le tok2vec, morphologizer, parser, sender, ner, atribut_ruler, ma lemmatizer vaega.

Ua silia nei ma le tausaga o matou galulue i le spaCy v3.0, ma toeititi atoa le lua tausaga pe a e faitauina uma galuega na faia i le Thinc. O la matou sini autu ma le faʻalauiloaina o le faʻafaigofieina ona aumai a oe lava faʻataʻitaʻiga i le SPACY, aemaise lava o le state-of-the-art faʻataʻitaʻiga pei o transformers. E mafai ona e tusiina ni faʻataʻitaʻiga e fafaga ai spaCy vaega i ni faʻafanua pei o le PyTorch poʻo le TensorFlow, e faʻaaogaina ai le matou masini fou fou e faʻamatala ai au tulaga uma. Ma talu ai o gaioiga faʻaonapo nei NLP e masani ona aofia ai le tele o laʻasaga, e i ai le fou auala faigaluega e fesoasoani ia te oe tausia lau galuega faʻamaopoopo.

Isi taua taua o loʻo tu matilatila mai le lomiga fou:

  • Taʻavale fou mo faʻataʻitaʻiga faʻataʻitaʻi.
  • Fou faiga fou.
  • Lagolago mo transformer-faʻavae laina paipa, talafeagai mo multitasking aʻoaʻoga.
  • Le agavaʻa e faʻafesoʻotaʻi au oe faʻataʻitaʻiga e faʻaaoga ai masini eseese e faʻataʻitaʻi ai masini, pei o le PyTorch, TensorFlow, ma le MXNet.
  • Polokalama lagolago e faʻatonutonu uma tulaga o alavai, mai i le muaʻi faʻagaioiga i le faʻataʻitaʻiga faʻatino.
  • Lagolago mo le tuʻufaʻatasia ma Data Version Control (DVC), Streamlit, Weight & Biases ma Ray afifi.
  • Fou vaega fausia-i totonu: SentenceRecognizer, Morphologizer, Lemmatizer,
  • AttributRuler ma Transformer.
  • Fou API e fausia a oe lava vaega.

Mulimuli ane, pe afai e te fiafia e iloa atili e uiga i ai o lenei fou lomiga pe e uiga ile spaCy, oe mafai ona siakiina auiliiliga I le fesoʻotaʻiga lenei.


O mataupu o le tusitusiga e tausisi ia tatou mataupu silisili o amio lelei faʻatonu. E lipotia se mea sese kiliki iinei.

Ia avea muamua ma faamatalaga

Tuʻu lau faamatalaga

o le a le lomia lou tuatusi imeli. O nofoaga e manaʻomia e makaina *

*

*

  1. E tali atu mo faʻamatalaga: Miguel Ángel Gatón
  2. Faamoemoega o faʻamatalaga: Pulea le SPAM, faʻamatalaga pulega.
  3. Tulaga faʻatulafonoina: Lau maliega
  4. Fesoʻotaʻiga o faʻamatalaga: O faʻamatalaga o le a le fesoʻotaʻi atu i isi vaega vagana i tulafono faʻatulafonoina.
  5. Teuina o faʻamatalaga: Faʻamaumauga tuʻufaʻatasia e Occentus Networks (EU)
  6. Aia Tatau: I soo se taimi e mafai ai ona e faʻatapulaʻaina, toe maua ma aveʻese au faʻamatalaga.