Extract nouns from Japanese text.

noun-extraction.clj

(ns kuro.core
  (:import (com.atilika.kuromoji.ipadic Token Tokenizer))
  (:gen-class))

(defn noun? [token]
  (let [pos (first (:part-of-speech token))]
    (or (= pos "名詞"))))

(defn ->part-of-speech [token]
  (letfn [(na? [x] (= x "*"))]
    (remove na?
           `(~(.getPartOfSpeechLevel1 token)
             ~(.getPartOfSpeechLevel2 token)
             ~(.getPartOfSpeechLevel3 token)
             ~(.getPartOfSpeechLevel4 token)))))

(defn ->token [obj]
  {
    :surface (.getSurface obj)
    :part-of-speech (->part-of-speech obj)
  })

(defn morphological-analysis [text]
  (let [tokenizer (new Tokenizer)
        tokens (.tokenize tokenizer text)]
    (map ->token tokens)))

(defn extract-nouns [text]
  (let [tokens (morphological-analysis text)]
    (map (fn [token] (:surface token))
         (filter noun? tokens))))

(defn -main
  [& args]
  (let [text "お寿司が食べたい。"]
    (println (extract-nouns text))))