import subprocess import shlex import os.path import sys import pandas as pd from os import getcwd class PySentiStr: def __init__(self): pass def setSentiStrengthPath(self, ss_Path): self.SentiStrengthLocation = ss_Path def setSentiStrengthLanguageFolderPath(self, sslf_Path): # Ensure it has a forward slash at the end if sslf_Path[-1] != '/': sslf_Path += '/' self.SentiStrengthLanguageFolder = sslf_Path def getSentiment(self, df_text, score='scale'): if not hasattr(self, 'SentiStrengthLocation'): assert False, "Set path using setSentiStrengthPath(path) function." if not hasattr(self, 'SentiStrengthLanguageFolder'): assert False, "Set path using setSentiStrengthLanguageFolderPath(path) function." if type(df_text) != pd.Series: df_text = pd.Series(df_text) df_text = df_text.str.replace('\n','') df_text = df_text.str.replace('\r','') conc_text = '\n'.join(df_text) p = subprocess.Popen(shlex.split("java -jar '" + self.SentiStrengthLocation + "' stdin sentidata '" + self.SentiStrengthLanguageFolder + "' trinary"),stdin=subprocess.PIPE,stdout=subprocess.PIPE,stderr=subprocess.PIPE) b = bytes(conc_text.replace(" ","+"), 'utf-8') stdout_byte, stderr_text = p.communicate(b) stdout_text = stdout_byte.decode("utf-8") stdout_text = stdout_text.rstrip().replace("\t"," ") stdout_text = stdout_text.replace('\r\n','') senti_score = stdout_text.split(' ') try: senti_score = list(map(float, senti_score)) except: raise Exception(stdout_text) senti_score = [int(i) for i in senti_score] if score == 'scale': # Returns from -4 to 4 senti_score = [sum(senti_score[i:i+2]) for i in range(0, len(senti_score), 3)] elif score == 'binary': # Return 1 if positive and -1 if negative senti_score = [1 if senti_score[i] >= abs(senti_score[i+1]) else -1 for i in range(0, len(senti_score), 3)] elif score == 'trinary': # Return Positive and Negative Score and Neutral Score senti_score = [tuple(senti_score[i:i+3]) for i in range(0, len(senti_score), 3)] elif score == 'dual': # Return Positive and Negative Score senti_score = [tuple(senti_score[i:i+2]) for i in range(0, len(senti_score), 3)] else: return "Argument 'score' takes in either 'scale' (between -1 to 1) or 'binary' (two scores, positive and negative rating)" return senti_score