Keyword extraction

The phrases that actually characterise a document, ranked, without you reading it first.

curl -X POST "$DOATHING_API/v1/nlp/keyword-extract" \
  -H "x-api-key: $DOATHING_KEY" \
  -H "content-type: application/json" \
  -d '{"text": "Ada Lovelace worked in London for Charles Babbage. She wrote the first algorithm intended for a machine. Her notes described a general-purpose computer a century before one was built."}'

The phrases that actually characterise a document, ranked, without you reading it first.

Extract ranked key phrases using RAKE or TF-IDF scoring over candidate n-grams.

Input

Send either a text string or a file object. A file is decoded as UTF-8 and must be one of text/plain, text/markdown, text/html, text/csv.

Response

The result carries your remaining balance alongside it, so you can track spend without a second call.

{
  "keywords": [
    {
      "phrase": "analytical engine",
      "score": 4.0,
      "relative": 1.0
    }
  ],
  "count": 10,
  "method": "rake",
  "request_id": "37f01edb-0163-42a1-ac51-0acaef979800",
  "credits_remaining": 96
}

Cost

3 credits per call, whether it is run from the site or from the API — the credential differs, the price does not. A new account starts with 20 credits.

A rejected request still costs a credit: the authorizer decrements before the tool validates. A call rejected for a missing or invalid key is free.

Parameters

Generated from the endpoint’s own validation, so this is exactly what it accepts. A body field goes at the top level; an option goes inside options.

Name	In	Type	Default	Notes
method	options	string	"rake"	rake for multi-word phrases, tfidf for single terms. one of rake, tfidf
language	options	string	"english"	Stopword list used to find phrase boundaries. one of albanian, arabic, azerbaijani, basque, belarusian, bengali, catalan, chinese, danish, dutch, english, finnish, french, german, greek, hebrew, hinglish, hungarian, indonesian, italian, kazakh, nepali, norwegian, portuguese, romanian, russian, slovene, spanish, swedish, tajik, tamil, turkish, uzbek
count	options	integer	10	How many to return, up to 100. 1 to 100