Summarizing Online Movie Reviews: A Machine Learning Approach to Big Data Analytics

Table 5

Movie reviews classification accuracy on three tasks: PL04 dataset contains 2,000 movie reviews and is used as standard data collection for the classification problems in sentiment analysis [58], Full IMDB data collection includes 50,000 reviews, and subjectivity data collection comprises of 1000 review sentences [58].

	NB with features	PL04	Full IMDB	Subjectivity
1	Unigrams features	81.5	86.66	90.75
2	Bigrams features	77.7	88.29	76.03
3	Unigrams + bigrams feature set [53]	82.4	88.91	91.22
4	Unigrams + bigrams + trigrams feature set	80.15	89.22	91.18
5	Unigrams occurrence + smooth IDF + cosine norm	82.1	87.36	90.7
6	Bigrams occurrence + smooth IDF + cosine norm	81.15	88.31	76.72
7	Unigrams + bigrams + smooth IDF + cosine norm [53]	83.7	89.28	90.91
8	Unigrams + bigrams + trigrams + smooth IDF + cosine norm	83.15	89.33	90.87
9	Benchmark model [54]	88.90	88.89	88.13