resources/app/plugins/deepmoji_plugin/xvasynth_torchmoji.py

# -*- coding: utf-8 -*-

""" Use torchMoji to score texts for emoji distribution.

The resulting emoji ids (0-63) correspond to the mapping
in emoji_overview.png file at the root of the torchMoji repo.

Returns the result as an array.
"""

from __future__ import print_function, division, unicode_literals


import time

import sys
from os.path import abspath, dirname

import json
import csv
import numpy as np

from torchmoji.sentence_tokenizer import SentenceTokenizer
from torchmoji.model_def import torchmoji_emojis
from torchmoji.global_variables import PRETRAINED_PATH, VOCAB_PATH

def top_elements(array, k):
    ind = np.argpartition(array, -k)[-k:]
    return ind[np.argsort(array[ind])][::-1]

maxlen = 30

print('Tokenizing using dictionary from {}'.format(VOCAB_PATH))
with open(VOCAB_PATH, 'r') as f:
    vocabulary = json.load(f)

st = SentenceTokenizer(vocabulary, maxlen)

print('Loading model from {}.'.format(PRETRAINED_PATH))
model = torchmoji_emojis(PRETRAINED_PATH)

def scoreText(text, scalp_amount=5):
    global st, model

    print('Running predictions.')
    # text
    tokenized, _, _ = st.tokenize_sentences([text])
    print(tokenized)
    prob = model(tokenized)

    for prob in [prob]:
        # Find top emojis for each sentence. Emoji ids (0-63)
        # correspond to the mapping in emoji_overview.png
        # at the root of the torchMoji repo.
        scores = []
        for i, t in enumerate([text]):
            t_tokens = tokenized[i]
            t_score = [t]
            t_prob = prob[i]
            ind_top = top_elements(t_prob, scalp_amount)
            t_score.append(sum(t_prob[ind_top]))
            t_score.extend(ind_top)
            t_score.extend([t_prob[ind] for ind in ind_top])

            return t_score
    return scores