wezleysherman

def compare_bars(input_bar, artists_bars):
  '''
    input_bars are the fire bars our AI generates
    artists_bars are the original bars for the artist

    The lower the score the better! We want unique bars
  '''
  # Converts sentences to matrix of token counts
  avg_dist = 0
  total_counted = 0

wezleysherman

''' Rhyme density is calculated by taking the number of rhymed syllables and divide it by total number of syllables'''
def calc_rhyme_density(bars):
  total_syllables = 0
  rhymed_syllables = 0
  for bar in bars:
    for word in bar.split():
      p = pronouncing.phones_for_word(word)
      if len(p) == 0:
        break
      syllables = pronouncing.syllable_count(p[0])

wezleysherman

def score_bar(input_bar, artists_bars, artists_avg_readability, artists_avg_rhyme_idx):
  gen_readability = textstat.automated_readability_index(input_bar)
  gen_rhyme_idx = calc_rhyme_density(input_bar)
  comp_bars = compare_bars(input_bar, artists_bars)

  # Scores based off readability, rhyme index, and originality. The lower the score the better.
  bar_score = (artists_avg_readability - gen_readability) + (artists_avg_rhyme_idx - gen_rhyme_idx) + comp_bars
  return bar_score

wezleysherman

def generate_bar(seed_phrase, model, length_of_bar):
  for i in range(length_of_bar):
    seed_tokens = pad_sequences(tokenizer.texts_to_sequences([seed_phrase]), maxlen=29)
    output_p = model.predict(seed_tokens)
    output_word = np.argmax(output_p, axis=1)[0]-1
    seed_phrase += " " + str(list(tokenizer.word_index.items())[output_word][0])
  return seed_phrase

wezleysherman

def generate_rap(model, artists_bars, length_of_bar=10, length_of_rap=20, min_score_threshold=-0.2, max_score_threshold=0.2, tries=5):
  artists_avg_readability = calc_readability(artists_bars)
  artists_avg_rhyme_idx = calc_rhyme_density(artists_bars)
  fire_rap = []
  cur_tries = 0
  candidate_bars = []

  while len(fire_rap) < length_of_rap:
    seed_phrase = markov_model.make_sentence(tries=500).split(" ")
    seed_phrase = " ".join(seed_phrase[:3])

wezleysherman

# CNN+LSTM
D = 512

T = train_X.shape[1]
i = Input(shape=(T,))
x = Embedding(V + 1, D)(i)
x = Dropout(0.2)(x)
x = Conv1D(filters=512, kernel_size=15)(x)
x = Dropout(0.2)(x)
x = Conv1D(filters=256, kernel_size=8)(x)

wezleysherman

# LSTM
D = 768

T = train_X.shape[1]
i = Input(shape=(T,))
x = Embedding(V + 1, D)(i)
x = Dropout(0.1)(x)
x = LSTM(100, return_sequences=True)(x)
x = LSTM(100)(x)
x = Dropout(0.1)(x)

wezleysherman

# GRU
D = 512

T = train_X.shape[1]
i = Input(shape=(T,))
x = Embedding(V + 1, D)(i)
x = Dropout(0.2)(x)
x = Bidirectional(GRU(150))(x)
x = Dense(V, activation="softmax")(x)
gru_model = Model(i, x)

wezleysherman

D = 512

#Simple RNN
T = train_X.shape[1]
i = Input(shape=(T,))
x = Embedding(V, D)(i)
x = Dropout(0.2)(x)
x = SimpleRNN(150)(x)
x = Dense(V, activation="softmax")(x)
rnn_model = Model(i, x)

wezleysherman

function csrfSafeMethod(method) {
    return (/^(GET|HEAD|OPTIONS|TRACE)$/.test(method));
}

$.ajaxSetup({
    beforeSend: function(xhr, settings) {
        if (!csrfSafeMethod(settings.type) && !this.crossDomain) {
            xhr.setRequestHeader("X-CSRFToken", csrftoken);
        }
    }