# number of reviews a token has to appear to be kept hardthreshold = 2 print "> Loading data" alltoken = data.loadFile(root + '/computed/alltoken.pkl') print "> Scanning data" print "Loading file", filename reviews_feature = dict() reviews_score = dict() tok = Tokenizer(preserve_case=True) # extracting tokens for line in data.generateLine(filename): review = json.loads(line) reviewid = review['review_id'] text = tok.ngrams(review['text'], 1, 3) score = int(review['stars']) # filtering tokens by the ones in the model text = filter(lambda k: k in alltoken, text) reviews_feature[reviewid] = Counter(text) reviews_score[reviewid] = score print "> End of full scan" print "> Saving" data.saveFile(reviews_feature, root + "/computed/reviews_feature.pkl") data.saveFile(reviews_score, root + "/computed/reviews_score.pkl")
