{"id":13801,"date":"2017-10-26T09:59:03","date_gmt":"2017-10-26T07:59:03","guid":{"rendered":"https:\/\/www.dase-analytics.com\/blog\/?p=13801"},"modified":"2020-02-06T18:48:35","modified_gmt":"2020-02-06T16:48:35","slug":"kam-az-moze-zajst-analyza-dat","status":"publish","type":"post","link":"https:\/\/www.dase-analytics.com\/blog\/sk\/kam-az-moze-zajst-analyza-dat\/","title":{"rendered":"Kam a\u017e m\u00f4\u017ee z\u00e1js\u0165 anal\u00fdza d\u00e1t?"},"content":{"rendered":"

Google len predned\u00e1vnom predstavil nov\u00e9 API<\/a>. Ide hlavne o vylep\u0161enie mo\u017enost\u00ed anal\u00fdzy d\u00e1t z obr\u00e1zkov, vide\u00ed alebo textu. Tie otv\u00e1raj\u00fa nov\u00e9 obzory v anal\u00fdze d\u00e1t.<\/strong><\/p>\n

Google Vision API<\/h2>\n

DEMO<\/strong>\u00a0<\/a>
\nT\u00e1to API n\u00e1m poskytuje ve\u013ek\u00fa silu, nako\u013eko pomerne jednoducho dok\u00e1\u017ee\u00a0analyzova\u0165 obr\u00e1zky<\/strong>\u00a0a d\u00f4le\u017eit\u00e9 inform\u00e1cie z\u00edskan\u00e9 z nich.
\nPri jej teste som do nej nahral obr\u00e1zok odfoten\u00fd z vrcholu Eiffelovej ve\u017ee.
\nGoogle Vision API ur\u010dila, presne\u00a0odkia\u013e bol obr\u00e1zok odfoten\u00fd<\/strong>, spolu s GPS s\u00faradnicami miesta. V mojom pr\u00edpade Google Visual API spr\u00e1vne identifikovala miesto ako Par\u00ed\u017e, konkr\u00e9tne Eiffelovu ve\u017eu.<\/p>\n

\"\"<\/p>\n

\u010ealej bol obr\u00e1zok zaraden\u00fd do kateg\u00f3ri\u00ed s pravdepodobnos\u0165ou, s akou si je API ist\u00e1, \u017ee je v\u00fdsledok spr\u00e1vny. Tieto kateg\u00f3rie n\u00e1m m\u00f4\u017eu pom\u00f4c\u0165 pri anal\u00fdze\u00a0ve\u013ek\u00e9ho mno\u017estva obr\u00e1zkov<\/strong>\u00a0a zisti\u0165, ktor\u00e9 z nich s\u00fa napr\u00edklad najzdie\u013eanej\u0161ie. Pr\u00edpadne vedia zisti\u0165 \u010di existuje nejak\u00e1\u00a0kateg\u00f3ria, ktor\u00e1 ich sp\u00e1ja<\/strong>. V mojom pr\u00edpade bol obr\u00e1zok zaraden\u00fd do t\u00fdchto kateg\u00f3ri\u00ed:<\/p>\n

\"\"<\/p>\n

\u010ealej dok\u00e1\u017ee API vyhodnoti\u0165, \u010di sa na internete nach\u00e1dzaj\u00fa\u00a0podobn\u00e9 obr\u00e1zky<\/strong>\u00a0a tie\u017e ur\u010d\u00ed, \u010do sa na obr\u00e1zku nach\u00e1dza. Pre mnou nahran\u00fd obr\u00e1zok to boli:<\/p>\n

\"\"<\/p>\n

Google Visual API tie\u017e dok\u00e1\u017ee vyhodnoti\u0165, \u010di je obsah\u00a0obr\u00e1zka vhodn\u00fd pre deti<\/strong>\u00a0alebo neobsahuje n\u00e1silie.<\/p>\n

Ak napr\u00edklad do tejto API vlo\u017e\u00edte obr\u00e1zok tv\u00e1re, dok\u00e1\u017ee vyhodnoti\u0165 aj moment\u00e1lne\u00a0em\u00f3cie \u010dloveka<\/strong>\u00a0na obr\u00e1zku. Ak obr\u00e1zok obsahuje text, API dok\u00e1\u017ee tento\u00a0text identifikova\u0165<\/strong>\u00a0a pre\u010d\u00edta\u0165.<\/p>\n

Ako je to vyu\u017eite\u013en\u00e9?<\/h3>\n

Predstavte si \u017ee m\u00e1me blog, kam m\u00f4\u017eu prispieva\u0165 v\u0161etci registrovan\u00ed pou\u017e\u00edvatelia. M\u00f4\u017eu p\u00edsa\u0165 \u010dl\u00e1nky a m\u00f4\u017eu k nim prida\u0165 aj svoje vlastn\u00e9 fotografie. Pomocou Google Visual API m\u00f4\u017eeme\u00a0analyzova\u0165 obr\u00e1zky z poh\u013eadu licencie<\/strong>\u00a0CC \u2013 \u010di ide o origin\u00e1lny obr\u00e1zok autora, alebo bol stiahnut\u00fd z Googlu s mo\u017en\u00fdm poru\u0161en\u00edm autorsk\u00fdch pr\u00e1v.
\nRovnako mo\u017eno\u00a0analyzova\u0165 nevhodn\u00fd obsah<\/strong>. Takto sa m\u00f4\u017eeme vyhn\u00fa\u0165 pokut\u00e1m, pr\u00edpadne zn\u00ed\u017eeniu n\u00e1\u0161ho renom\u00e9.<\/p>\n

M\u00f4\u017eeme si taktie\u017e v\u0161\u00edma\u0165 aj to, \u010do sa na dan\u00fdch obr\u00e1zkoch nach\u00e1dza a zaradi\u0165 tieto objekty do kateg\u00f3ri\u00ed. M\u00f4\u017eeme zisti\u0165, \u017ee ur\u010dit\u00e9 kateg\u00f3rie obr\u00e1zkov maj\u00fa\u00a0v\u00e4\u010d\u0161\u00ed vplyv na n\u00e1v\u0161tevn\u00edkov<\/strong>\u00a0web str\u00e1nky ako in\u00e9 kateg\u00f3rie.<\/p>\n

Google Natural Language API<\/h2>\n

DEMO<\/strong><\/a>
\nGoogle Natural Language API sa zaober\u00e1 textami. Analyzuje\u00a0vz\u0165ahy medzi slovami<\/strong>, a ka\u017ed\u00e9mu slovu v texte prirad\u00ed v\u00e1hu.
\nNajlep\u0161ie je to op\u00e4\u0165 vysvetli\u0165 na pr\u00edklade.<\/p>\n

Skop\u00edroval som si text a vlo\u017eil ho do API. Po chv\u00edli som mal pred sebou v\u00fdsledok:<\/p>\n

\"\"<\/p>\n

Natural Language API zanalyzovala text a mohol som vidie\u0165\u00a0slov\u00e1, ich d\u00f4le\u017eitos\u0165 a kateg\u00f3rie<\/strong>, pr\u00edpadne entity, kam slov\u00e1 patria. Ak ste niekedy po\u010duli o textovej vate, tak na hornom obr\u00e1zku je t\u00e1to \u201evata\u201d zobrazen\u00e1 sivou farbou. D\u00e1 sa poveda\u0165, \u017ee anal\u00fdza ur\u010dila d\u00f4le\u017eitos\u0165 a\u00a0relevanciu jednotliv\u00fdch slov<\/strong>.<\/p>\n

\"\"<\/p>\n

Zauj\u00edmav\u00fdm je aj\u00a0hodnotenie sentimentality textu<\/strong>, teda toho, ak\u00fa n\u00e1ladu m\u00f4\u017ee text vyvola\u0165 u \u010ditate\u013ea. To je vyjadren\u00e9 indexom na \u0161k\u00e1le od -1 (ve\u013emi negat\u00edvne) po 1 (ve\u013emi pozit\u00edvne).<\/p>\n

\"\"
\nAko je to vyu\u017eite\u013en\u00e9?<\/h3>\n

Mo\u017eno\u00a0analyzova\u0165 cel\u00e9 blogov\u00e9 \u010dl\u00e1nky<\/strong>, popisy pred\u00e1van\u00fdch produktov a ako silne vpl\u00fdvaj\u00fa na pou\u017e\u00edvate\u013ea. To sa m\u00f4\u017ee osved\u010di\u0165 napr\u00edklad pri identifik\u00e1cii hoaxov.
\nRovnako s pomocou Natural Language API mo\u017eno vn\u00edma\u0165 s\u00favislosti a lep\u0161ie analyzova\u0165 obsah dlh\u00fdch textov. T\u00e1to API dok\u00e1\u017ee\u00a0generova\u0165 k\u013e\u00fa\u010dov\u00e9 slov\u00e1<\/strong>\u00a0pre \u010dl\u00e1nky a rovnako dok\u00e1\u017ee ur\u010di\u0165, ktor\u00fd text je len \u201eve\u013ea slov a m\u00e1lo kontextu\u201d, a ktor\u00fd text je, naopak, hutn\u00fd a kvalitn\u00fd.<\/p>\n

Google Video Intelligence API<\/h2>\n

DEMO<\/strong><\/a>
\nVe\u013emi podobn\u00e1 ako Google Vision API je i Video Intelligence API.
\nDok\u00e1\u017ee rozozna\u0165 prvky z videa, \u010dasy, v ktor\u00fdch sa dan\u00e9 prvky vo videu nach\u00e1dzaj\u00fa, ako i pravdepodobnos\u0165, s akou si je API ist\u00e1, \u017ee sa nem\u00fdli.<\/p>\n

\"\"<\/p>\n

Tie\u017e dok\u00e1\u017ee ur\u010di\u0165, kde bolo video nasn\u00edman\u00e9.<\/p>\n

Ako je to vyu\u017eite\u013en\u00e9?<\/h3>\n

Predstavte si, \u017ee m\u00e1me blog, kam m\u00f4\u017eu prispieva\u0165 v\u0161etci registrovan\u00ed pou\u017e\u00edvatelia. M\u00f4\u017eu p\u00edsa\u0165 \u010dl\u00e1nky, av\u0161ak rovnako m\u00f4\u017eu prid\u00e1va\u0165 aj svoje vlastn\u00e9 vide\u00e1. Pomocou Google Video Intelligence API\u00a0skontrolujeme, \u010di nie je obsah videa nevhodn\u00fd<\/strong>.<\/p>\n

Dok\u00e1\u017eeme tie\u017e analyzova\u0165 to, \u010do sa na v\u0161etk\u00fdch vide\u00e1ch nach\u00e1dza a ich po\u010detnos\u0165. Ak niekto vo videu rozpr\u00e1va, vieme ur\u010di\u0165, v ktorom \u010dase sa tak deje. Taktie\u017e vieme\u00a0vyextrahova\u0165, kto \u010do hovor\u00ed<\/strong>. Teda m\u00f4\u017eeme analyzova\u0165 video z poh\u013eadu obsahu, ako aj z poh\u013eadu\u00a0objektov a farieb<\/strong>.<\/p>\n

Predstavte si, \u017ee na va\u0161ej str\u00e1nke s\u00fa stovky r\u00f4znych blogov\u00fdch vide\u00ed. Ako spr\u00e1vca webu nem\u00e1te \u0161ancu v\u0161etky tieto\u00a0<\/span>vide\u00e1 pozera\u0165 a vyhodnoti\u0165<\/b>. Av\u0161ak pomocou anal\u00fdzy a \u0161tatistiky dok\u00e1\u017eete presne zisti\u0165,\u00a0<\/span>\u010do sa na va\u0161ej str\u00e1nke nach\u00e1dza<\/b>. Ke\u010f viete, \u010do obsahuj\u00fa vide\u00e1 na va\u0161om webe , pozn\u00e1te aj to, ak\u00e9 t\u00e9my s\u00fa zauj\u00edmav\u00e9 pre va\u0161e publikum.<\/span><\/p>\n

No a ke\u010f lep\u0161ie pozn\u00e1te va\u0161e publikum, dok\u00e1\u017eete mu\u00a0<\/span>pon\u00faknu\u0165 relevantnej\u0161ie slu\u017eby<\/b>.<\/span><\/p>\n

Technick\u00fd v\u00fdstup<\/h2>\n

D\u00e1ta z nov\u00fdch Google API treba pred anal\u00fdzou spracova\u0165. Tieto\u00a0d\u00e1ta s\u00fa vo form\u00e1te JSON<\/strong>, ktor\u00fd je \u013eahko pr\u00edstupn\u00fd pre \u010fal\u0161iu d\u00e1tov\u00fa analytiku napr\u00edklad v Pythone<\/a>, R<\/a> alebo REST API.<\/p>\n

\"\"<\/p>\n

Z\u00e1le\u017e\u00ed len na V\u00e1s, ktor\u00fd n\u00e1stroj preferujete.<\/p>\n

Isto je e\u0161te\u00a0ve\u013ea in\u00fdch mo\u017enost\u00ed na vyu\u017eitie<\/strong>\u00a0t\u00fdchto API. E\u0161te viac zauj\u00edmav\u00fdmi sa st\u00e1vaj\u00fa, ke\u010f sa spoja viacer\u00e9 z nich do jedn\u00e9ho produktu.<\/p>\n

\n

Nov\u00e9 Google API pohromade:<\/strong><\/h3>\n

Video Intelligence API:<\/strong>\u00a0https:\/\/cloud.google.com\/video-intelligence\/<\/a>
\nVision API:<\/strong>\u00a0https:\/\/cloud.google.com\/vision\/<\/a>
\nSpeech API:<\/strong>\u00a0https:\/\/cloud.google.com\/speech\/<\/a>
\nNatural language API:<\/strong>\u00a0https:\/\/cloud.google.com\/natural-language\/<\/a>
\nTranslate API:<\/strong>\u00a0https:\/\/cloud.google.com\/translate\/<\/a><\/p><\/blockquote>\n","protected":false},"excerpt":{"rendered":"

Google len predned\u00e1vnom predstavil nov\u00e9 API. Ide hlavne o vylep\u0161enie mo\u017enost\u00ed anal\u00fdzy d\u00e1t z obr\u00e1zkov, vide\u00ed alebo textu….<\/p>\n","protected":false},"author":64,"featured_media":13803,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":[],"categories":[340,200],"tags":[603],"_links":{"self":[{"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/posts\/13801"}],"collection":[{"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/users\/64"}],"replies":[{"embeddable":true,"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/comments?post=13801"}],"version-history":[{"count":4,"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/posts\/13801\/revisions"}],"predecessor-version":[{"id":15221,"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/posts\/13801\/revisions\/15221"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/media\/13803"}],"wp:attachment":[{"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/media?parent=13801"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/categories?post=13801"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.dase-analytics.com\/blog\/sk\/wp-json\/wp\/v2\/tags?post=13801"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}