Annotate.io

Tweet example with brands

You can HTTP POST the text in this tweet to our API:

"Snubbed by Apple, foursquare makes its maps experience better on iPhone http://tinyurl.com/d37w2ue by @JBruin #mapping"

and we'll recognise the brands, users, URLs and hashtags:

"Snubbed by Apple, foursquare makes its maps experience better on iPhone http://tinyurl.com/d37w2ue by @JBruin #mapping"

and we'll send back the following:

{"Brands":
 {"Apple":      {"span": [11, 15], "raw": "Apple"},
  "FourSquare": {"span": [18, 27], "raw": "foursquare"},
  "Apple":      {"span": [65, 71], "raw": "iPhone"}
 },
 "URLs":
 {"URL":        {"span": [72, 98], "raw": "http://tinyurl.com/d37w2ue"}
 },
 "Users":
 {"jbruin":     {"span": [102, 109], "raw": "@JBruin"}
 },
 "Hashtags":
 {"mapping":    {"span": [110, 118], "raw": "#mapping"}
 },
 "BrandLinks":
 {"Apple":      {"URL": "http://apple.com", "Wikipedia": "http://en.wikipedia.org/wiki/Apple_Inc."},
  "FourSquare": {"URL": "http://foursquare.com", "Wikipedia": "http://en.wikipedia.org/wiki/Foursquare"}
 }
}

The clever part is our use of Natural Language Processing and Machine Learning to provide an Entity Recognition engine that correctly recognises brand terms from non-brand usages. Examples include:

• Apple the brand vs apple the fruit (or the Big Apple location or Fiona Apple the singer)
• Paris the Latin American brand vs Paris in France or the USA (or the celebrity)
• Coke the brand vs Coke the fuel (or the slang name for a drug)
• Palm the brand vs palm the tree (or the oil)
• Just Do It the phrase versus a usage like "if you just do it then it'll be done"