actor-sentiment-analysis/NLPProject.py at master · Tyler-Shepherd/actor-sentiment-analysis · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
import csv
import sys
from pprint import pprint
import NameTagging
import LexiconSentimentAnalysis
import NLPParsing
import ReadDNNResults
import os
from subprocess import call

sys.path.insert(0, './lib/DNN-Sentiment')

import getSentiment


sys.path.insert(0, './lib/LSTM/reviews')
import actorReview

if __name__ == "__main__":
    if len(sys.argv) <= 2:
        print("Improper number of arguments. Please run with python3 NLPProject.py <review filename> <algo choice>")
        print("-l for lexicon, -r for RNN, -c for CNN, -a for AT-LSTM")
        sys.exit()

    review_filename = sys.argv[1]
    algo_type = sys.argv[2]

    review_file = open("./data/examples/"+review_filename, "r")

    review = ""

    for line in review_file:
        review += line[:line.find('\n')] + " "

    review_file.close()

    # Name tagging
    # Gets all named entities of type PER in the review

    per_entities = NameTagging.name_tagging(review)

    print("PER entities", per_entities, flush = True)

    named_entities = [i for i in per_entities if NameTagging.is_actor(i)]

    print("named entity actors", named_entities, flush = True)


    # Could use entity linking to determine whether each of these are an actor


    # Write actor sentences to file needed for RNN and CNN
    if algo_type == "-r" or algo_type == "-c":
        # write to file for DNN-Sentiment to read
        output_file = open("./lib/DNN-Sentiment/data/actor_sentences.txt", "w")
        for entity in named_entities:
            actor_sentences = NLPParsing.get_actor_sentences(entity, review)

            for sen in actor_sentences:
                output_file.write(sen + ".")
        output_file.close()

        if algo_type == "-r":
            # Old: 1524375413
            # New: 1524617308

            all_predictions,all_scores = getSentiment.getSentimentRNN("actor_sentences",'1524617308')
            getSentiment.saveSentiment("actor_sentences",all_predictions,all_scores)
        else:
            all_predictions,all_scores = getSentiment.getSentimentCNN("actor_sentences",'1524270913')
            getSentiment.saveSentiment("actor_sentences",all_predictions,all_scores)
    elif algo_type == "-a":
        output_file = open("./lib/LSTM/data/actors/new_review.txt", "w")

    for entity in named_entities:
        if algo_type == "-l":
            sentiment = LexiconSentimentAnalysis.GetSentiment(entity, review)
            print(entity, ": ", NLPParsing.int_to_sentiment(sentiment), flush=True)
        elif algo_type == "-r" or algo_type == "-c":
            actor_sentences = NLPParsing.get_actor_sentences(entity, review)
            sentiment = ReadDNNResults.GetSentiment("./lib/DNN-Sentiment/data/actor_sentences.csv", actor_sentences)
            print(entity, ": ", NLPParsing.int_to_sentiment(sentiment), flush=True)
        elif algo_type == "-a":
            rev_line = review.replace(entity,"$T$")
            rev_line = rev_line.replace(entity.lower(), "$T$")

            # Replace all mentions of just one word in the actors name (i.e. their last name)
            for word in entity.split():
                if len(word) < 2:
                    # Ignore single letters (like an initial)
                    continue
                rev_line = rev_line.replace(word, "$T$")
                rev_line = rev_line.replace(word, "$T$")

            while "$T$ $T$" in rev_line:
                rev_line = rev_line.replace("$T$ $T$", "$T$")
            rev_line = rev_line.split(".")
            for sen in rev_line:
                if "$T$" in sen:
                    sen += '\n'
                    aspect_line = entity + "\n"
                    sentiment_line = "0"
                    output_file.write(sen + aspect_line + sentiment_line +"\n")
    #To ensure that the attention network LSTM model works, add three random reviews to end of the test dataset.
    if algo_type == "-a":
        positive = "$T$ did a phenomenal job in black panther which overall was a solid movie but a little bit overhyped..\n"
        positive += "chadwick boseman\n" + "1\n"
        output_file.write(positive)
        neutral = "$T$ did a mediocre job in black panther which overall was a solid movie but a little bit overhyped.\n"
        neutral += "mark jones\n" +  "0\n"
        output_file.write(neutral)
        terrible = "$T$ did a horrible job in black panther which overall was a solid movie but a little bit overhyped.\n"
        terrible += "johnny adams\n" + "-1\n"
        output_file.write(terrible)
        output_file.close()
        os.chdir("lib/LSTM/reviews")
        call(["py", "-3", "actorReview.py"])