|
|
|
|
@ -8,7 +8,9 @@
|
|
|
|
|
[com.owoga.phonetics :as phonetics]
|
|
|
|
|
[com.owoga.phonetics.syllabify :as syllabify]
|
|
|
|
|
[com.owoga.phonetics.stress-manip :as stress-manip]
|
|
|
|
|
[clojure.math.combinatorics :as combinatorics]))
|
|
|
|
|
[clojure.math.combinatorics :as combinatorics]
|
|
|
|
|
[com.owoga.prhyme.nlp.core :as nlp]
|
|
|
|
|
[com.owoga.prhyme.data-transform :as data-transform]))
|
|
|
|
|
|
|
|
|
|
(defn gen-artist []
|
|
|
|
|
(->> [(rand-nth (seq dict/adjectives))
|
|
|
|
|
@ -319,12 +321,59 @@
|
|
|
|
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
(defn rhymes-by-quality
|
|
|
|
|
[seed-phrase]
|
|
|
|
|
(->> seed-phrase
|
|
|
|
|
(prhyme/phrase->all-phones)
|
|
|
|
|
(map first)
|
|
|
|
|
(map
|
|
|
|
|
(fn [phones]
|
|
|
|
|
[phones (markov/rhymes
|
|
|
|
|
models/rhyme-trie-unstressed-trailing-consonants
|
|
|
|
|
phones)]))
|
|
|
|
|
(map (fn append-quality-of-rhyme [[phones1 words]]
|
|
|
|
|
[phones1 (->> (mapcat
|
|
|
|
|
prhyme/phrase->all-phones
|
|
|
|
|
(reduce into #{} (map second words)))
|
|
|
|
|
(map (fn [[phones2 word]]
|
|
|
|
|
[phones2
|
|
|
|
|
word
|
|
|
|
|
(prhyme/quality-of-rhyme-phones
|
|
|
|
|
phones1
|
|
|
|
|
phones2)])))]))
|
|
|
|
|
(map (fn sort-by-quality-of-rhyme [[phones1 words]]
|
|
|
|
|
[phones1 (sort-by (fn [[_ _ quality]]
|
|
|
|
|
(- quality))
|
|
|
|
|
words)]))
|
|
|
|
|
(mapcat second)
|
|
|
|
|
(sort-by #(- (nth % 2)))
|
|
|
|
|
(take 20)
|
|
|
|
|
(map second)))
|
|
|
|
|
|
|
|
|
|
(comment
|
|
|
|
|
(rhymes-by-quality "bother me")
|
|
|
|
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
(defn open-nlp-perplexity
|
|
|
|
|
"Returns the perplexity of the parse tree using OpenNLP.
|
|
|
|
|
This is an alternative to the perplexity of the Markov model.
|
|
|
|
|
Normalized per word because long sentences are naturally more perplex."
|
|
|
|
|
[phrase]
|
|
|
|
|
(->> phrase
|
|
|
|
|
nlp/tokenize
|
|
|
|
|
(string/join " ")
|
|
|
|
|
(nlp/most-likely-parse)
|
|
|
|
|
((fn [[line perplexity]]
|
|
|
|
|
[line (/ perplexity (count (string/split line #" ")))]))
|
|
|
|
|
second))
|
|
|
|
|
|
|
|
|
|
(defn lyric-suggestions [seed-phrase trie database]
|
|
|
|
|
(let [realize-seed (fn [seed]
|
|
|
|
|
(string/join " " (-> (map database (reverse seed))
|
|
|
|
|
butlast
|
|
|
|
|
rest)))]
|
|
|
|
|
(data-transform/untokenize
|
|
|
|
|
(-> (map database (reverse seed))
|
|
|
|
|
butlast
|
|
|
|
|
rest)))]
|
|
|
|
|
(loop [seed (vec (reverse (map #(get database % 0) (string/split seed-phrase #" "))))]
|
|
|
|
|
(cond
|
|
|
|
|
(< 20 (count seed)) (realize-seed seed)
|
|
|
|
|
@ -335,7 +384,28 @@
|
|
|
|
|
(partial remove (fn [child]
|
|
|
|
|
(= (.key child) (database prhyme/EOS)))))))))))
|
|
|
|
|
|
|
|
|
|
(defn phrase->quality-of-rhyme
|
|
|
|
|
"Gets the quality of rhyme of the thie highest quality pronunciation of all
|
|
|
|
|
combinations of phrases."
|
|
|
|
|
[phrase1 phrase2]
|
|
|
|
|
(let [phones1 (prhyme/phrase->all-phones phrase1)
|
|
|
|
|
phones2 (prhyme/phrase->all-phones phrase2)
|
|
|
|
|
all-possible-rhyme-combinations (combinatorics/cartesian-product
|
|
|
|
|
phones1
|
|
|
|
|
phones2)]
|
|
|
|
|
(->> all-possible-rhyme-combinations
|
|
|
|
|
(map (partial map first))
|
|
|
|
|
(map (juxt identity
|
|
|
|
|
(partial apply quality-of-rhyme-phones)))
|
|
|
|
|
(sort-by (comp - second))
|
|
|
|
|
first)))
|
|
|
|
|
|
|
|
|
|
(comment
|
|
|
|
|
(lyric-suggestions "bother me </s>" models/markov-trie models/database)
|
|
|
|
|
(phrase->quality-of-rhyme "boss hog" "brain fog")
|
|
|
|
|
|
|
|
|
|
(->> #(lyric-suggestions "bother me </s>" models/markov-trie models/database)
|
|
|
|
|
repeatedly
|
|
|
|
|
(take 5)
|
|
|
|
|
(map (juxt identity open-nlp-perplexity (partial phrase->quality-of-rhyme "bother me"))))
|
|
|
|
|
|
|
|
|
|
)
|
|
|
|
|
|
