scripts/sa/sentexicon.py

#! /usr/bin/python
# -*- coding: utf-8 -*-

"""	
	This module defines classes to load and manipulate sentiment lexicons.
	
	Version 0.2									(28-Sep-2015)
	- collapseScores() is added to Sentexicon.
	- collapseScore() is added to SentexEntry.
	
	Version 0.1									(03-Sep-2015 to 04-Sep-2015)
	- Sentexicon and SentexEntry are added.
"""


class Sentexicon:
	'''
	Sentiment lexicon class
	'''
	
	
	def __init__(self):
		'''
		Constructor
		'''
		
		## lexicon entry dictionary: keys are words and values SentexEntry
		## objects; dictionary is used for search efficiency
		## NOTE: words is stored both in the keys and the objects, which
		## may sound odd. However, being in dictionary keys makes the search
		## efficient and being in the entry objects make them self-contained.
		self.entries = {}
		
		# character case of entries: (lower)cased, (upper)cased or (true)cased (i.e. original case).
		self.case = "lower"
		
	
	
	def loadFromCaraCSV(self, pFilename, pCase, pflgAppend = False):
		'''
		Loads the lexicon from file in Cara CSV format
		
		pCase specifies the case used in the lexicon. That is whether the
		entries are (lower)cased, (upper)cased or (true)cased (i.e. original
		case).
		
		For format details, see a lexicon sample in https://github.com/CNGL-repo/Cara/tree/master/src/feature_extraction/lexicon
		For more description, see https://github.com/CNGL-repo/Cara/wiki/Lexicon augmentation.
		'''
		
		vLines = open(pFilename).read().strip().split('\n')
		
		
		self.setCase(pCase) 
		
		if not pflgAppend:
			self.entries = {}
		
		for vLine in vLines[1:]:
			vEntry = SentexEntry()
			vEntry.loadFromCaraCSV(vLine)
			self.entries[vEntry.word] = vEntry
		
	
	
	def setCase(self, pCase):
		'''
		Sets the character case of the lexicon entries
		
		Values are:
		- (lower)cased
		- (upper)cased
		- (true)cased (i.e. original case)
		'''
		
		if pCase.lower().startswith("lower"):
			self.case = "lower"
		elif pCase.lower().startswith("upper"):
			self.case = "upper"
		elif pCase.lower().startswith("true"):
			self.case = "true"
		else:
			raise Exception("Unknown character case: %s" % pCase)
		
	
	
	def getScore(self, pWord):
		'''
		Returns the sentiment score of the given word if it is found in the
		lexicon, otherwise None
		'''
		
		if self.case == "lower":
			vWord = pWord.lower()
		elif self.case == "upper":
			vWord = pWord.upper()
		elif self.case == "true":
			vWord = pWord
		
		try:
			return self.entries[vWord].score
		except KeyError:
			return None
			
		
	
	
	def collapseScores(self):
		'''
		Collapses the sentiment scores to a coarse scale comprising three 
		scores for negative (-1), neutral(0) and positive (+1) sentiments
		'''
		
		for vEntry in self.entries:
			self.entries[vEntry].collapseScore()
		
	
	


class SentexEntry:
	'''
	Sentiment lexicon entry class
	'''
	
	
	def __init__(self):
		'''
		Constructor
		'''
		
		self._word = ''
		self._score = None                   # sentiment score
		
	
	
	def loadFromCaraCSV(self, pLine):
		'''
		Loads the lexicon entry from a line in Cara CSV format
		
		For details, see a lexicon sample in https://github.com/CNGL-repo/Cara/tree/master/src/feature_extraction/lexicon
		For more description, see https://github.com/CNGL-repo/Cara/wiki/Lexicon augmentation.
		'''
		
		vLineSplit = pLine.split(',')
		self._word = vLineSplit[0]
		try:
			self._score = float(vLineSplit[1])
		except ValueError:
			raise Exception("The sentiment score in not valid in: %s" % pLine)
		
	
	
	@property
	def word(self):
		'''
		Returns the entry word
		'''
		
		return self._word
		
	
	
	@property
	def score(self):
		'''
		Returns the sentiment score of the entry
		'''
		
		return self._score
		
	
	
	def collapseScore(self):
		'''
		Collapses the sentiment score to a coarse scale comprising three
		scores for negative (-1), neutral(0) and positive (+1) sentiments
		'''
		
		if self._score > 0:
			self._score = 1
		elif self._score < 0:
			self.score = -1