Module auto_osint_v.sentiment_analyser
This module's primary role is to analyse the semantics of the intelligence statement.
This module will likely be reused/modified within source aggregation.
Expand source code
"""This module's primary role is to analyse the semantics of the intelligence statement.
This module will likely be reused/modified within source aggregation.
"""
from transformers import pipeline
class SentimentAnalyser:
"""This class provides methods for conducting sentiment analysis on a given document.
This document could be the intelligence statement or a different document.
"""
def __init__(self, read_statement, statement_title, file_handler_object):
"""Initialises variables to be used in this object.
Args:
read_statement: This is the statement input in __main__.py and read by
FileHandler.read_file()
statement_title: This is the source title or filename.
file_handler_object: The file handler object passed to the class for reuse
"""
self.statement = read_statement
self.file_name = statement_title
self.file_handler = file_handler_object
# Trying a variety of models. Need one with 3 labels for +ve, -ve and neutral.
# We want intelligence statements to be neutral and not too +ve or -ve
self.sentiment_analysis = pipeline("sentiment-analysis",
model="Souvikcmsa/BERT_sentiment_analysis")
def set_statement(self, new_statement):
"""Setter for the self.statement initial variable"""
self.statement = new_statement
def statement_analyser(self):
"""This method analyses the overall sentiment of the intelligence statement.
(it could be used to analyse other statements)
This code is all from the Hugging Face documentation.
Returns:
Nothing - outputs to file
"""
classification = self.sentiment_analysis(self.statement)
# print(classification)
# create a sentiment threshold for the intel statement
# If the threshold is exceeded add extra information to warn user that their statement is
# likely bias or sensational, see Calvo et al. (2021).
threshold = 0.9
# If you want a different threshold depending on source type,
# add as a parameter to statement_analyser
# Thought to offer better readability to user
evidence_type = "sentiment-analysis-of-" + self.file_name
classification_label = classification[0]["label"]
classification_score = classification[0]["score"]
if classification_label != 'neutral' and classification_score > threshold:
# write analysis info and a warning as extra info
warning = ("Warning: analysis of your statement indicates it is likely biased or "
"sensational")
self.file_handler.open_evidence_file([evidence_type, classification_label +
' sentiment, score: ' + str(classification_score),
warning])
else:
# write just the analysis info
self.file_handler.open_evidence_file([evidence_type, classification_label +
' sentiment, score: ' + str(classification_score)]
)
def headline_analyser(self, headline):
"""Runs sentiment analysis on a given headline.
Returns:
the sentiment label (positive, negative, neutral) and the confidence score.
"""
headline = headline.strip()
classification = self.sentiment_analysis(headline)
# get the label and score
classification_label = classification[0]["label"]
classification_score = classification[0]["score"]
return classification_label, classification_scoreClasses
class SentimentAnalyser (read_statement, statement_title, file_handler_object)-
This class provides methods for conducting sentiment analysis on a given document.
This document could be the intelligence statement or a different document.
Initialises variables to be used in this object.
Args
read_statement- This is the statement input in main.py and read by FileHandler.read_file()
statement_title- This is the source title or filename.
file_handler_object- The file handler object passed to the class for reuse
Expand source code
class SentimentAnalyser: """This class provides methods for conducting sentiment analysis on a given document. This document could be the intelligence statement or a different document. """ def __init__(self, read_statement, statement_title, file_handler_object): """Initialises variables to be used in this object. Args: read_statement: This is the statement input in __main__.py and read by FileHandler.read_file() statement_title: This is the source title or filename. file_handler_object: The file handler object passed to the class for reuse """ self.statement = read_statement self.file_name = statement_title self.file_handler = file_handler_object # Trying a variety of models. Need one with 3 labels for +ve, -ve and neutral. # We want intelligence statements to be neutral and not too +ve or -ve self.sentiment_analysis = pipeline("sentiment-analysis", model="Souvikcmsa/BERT_sentiment_analysis") def set_statement(self, new_statement): """Setter for the self.statement initial variable""" self.statement = new_statement def statement_analyser(self): """This method analyses the overall sentiment of the intelligence statement. (it could be used to analyse other statements) This code is all from the Hugging Face documentation. Returns: Nothing - outputs to file """ classification = self.sentiment_analysis(self.statement) # print(classification) # create a sentiment threshold for the intel statement # If the threshold is exceeded add extra information to warn user that their statement is # likely bias or sensational, see Calvo et al. (2021). threshold = 0.9 # If you want a different threshold depending on source type, # add as a parameter to statement_analyser # Thought to offer better readability to user evidence_type = "sentiment-analysis-of-" + self.file_name classification_label = classification[0]["label"] classification_score = classification[0]["score"] if classification_label != 'neutral' and classification_score > threshold: # write analysis info and a warning as extra info warning = ("Warning: analysis of your statement indicates it is likely biased or " "sensational") self.file_handler.open_evidence_file([evidence_type, classification_label + ' sentiment, score: ' + str(classification_score), warning]) else: # write just the analysis info self.file_handler.open_evidence_file([evidence_type, classification_label + ' sentiment, score: ' + str(classification_score)] ) def headline_analyser(self, headline): """Runs sentiment analysis on a given headline. Returns: the sentiment label (positive, negative, neutral) and the confidence score. """ headline = headline.strip() classification = self.sentiment_analysis(headline) # get the label and score classification_label = classification[0]["label"] classification_score = classification[0]["score"] return classification_label, classification_scoreMethods
def headline_analyser(self, headline)-
Runs sentiment analysis on a given headline.
Returns
the sentiment label (positive, negative, neutral) and the confidence score.
Expand source code
def headline_analyser(self, headline): """Runs sentiment analysis on a given headline. Returns: the sentiment label (positive, negative, neutral) and the confidence score. """ headline = headline.strip() classification = self.sentiment_analysis(headline) # get the label and score classification_label = classification[0]["label"] classification_score = classification[0]["score"] return classification_label, classification_score def set_statement(self, new_statement)-
Setter for the self.statement initial variable
Expand source code
def set_statement(self, new_statement): """Setter for the self.statement initial variable""" self.statement = new_statement def statement_analyser(self)-
This method analyses the overall sentiment of the intelligence statement.
(it could be used to analyse other statements) This code is all from the Hugging Face documentation.
Returns
Nothing - outputs to file
Expand source code
def statement_analyser(self): """This method analyses the overall sentiment of the intelligence statement. (it could be used to analyse other statements) This code is all from the Hugging Face documentation. Returns: Nothing - outputs to file """ classification = self.sentiment_analysis(self.statement) # print(classification) # create a sentiment threshold for the intel statement # If the threshold is exceeded add extra information to warn user that their statement is # likely bias or sensational, see Calvo et al. (2021). threshold = 0.9 # If you want a different threshold depending on source type, # add as a parameter to statement_analyser # Thought to offer better readability to user evidence_type = "sentiment-analysis-of-" + self.file_name classification_label = classification[0]["label"] classification_score = classification[0]["score"] if classification_label != 'neutral' and classification_score > threshold: # write analysis info and a warning as extra info warning = ("Warning: analysis of your statement indicates it is likely biased or " "sensational") self.file_handler.open_evidence_file([evidence_type, classification_label + ' sentiment, score: ' + str(classification_score), warning]) else: # write just the analysis info self.file_handler.open_evidence_file([evidence_type, classification_label + ' sentiment, score: ' + str(classification_score)] )