Anticipating Corporate Financial Performance from CEO Letters Utilizing Sentiment Analysis

Table 1

Related work of lexicon-based approach in sentiment analysis.

Authors	Objectives	Models	Data source	Evaluation method	Data set	Accuracy (percent)	Precision (percent)	Recall	F1
Hatzivassiloglou and Mckeown [24]	Assign adjective ±	Nonhierarchical clustering	WSJ corpus	N/A	657 adj (+) 679 adj (−)	78.1–92.4	N/A	N/A	N/A
Turney [23]	Assign docs sentiments	PMI-IR	Automobile, bank, movie, travel reviews	N/A	240 (+) 170 (−)	65.8–84	N/A	N/A	N/A
Turney and Littman [25]	Assign docs sentiments	PMI LSA	AV-ENG corpus AV-CA corpus TASA corpus	N/A	1614 (+) 1982 (−)	82.8–95	N/A	N/A	N/A
Taboada et al. [26]	Assign adjectives ±	SO-PMI	Reviews	N/A	521 adj	49.5–56.75	N/A	N/A	N/A
Ding et al. [27]	Assign adjectives, adverbs, verbs, and nouns ±	Opinion Observer	445 customer reviews of products from amazon.com	Macroaveraging (macroaveraging means given a set of confusion tables, a set of values are generated, and each value represents the precision or recall of an automatic classifier for each category)	N/A	N/A	91	90	90
		No context dependency				N/A	92	83	87
		Without using equation				N/A	90	85	87
		FBS				N/A	92	74	82
Taboada et al. [4]	Assign adjectives, verbs, adverbs, nouns ±	SO-CAL	Reviews	Across various domains	N/A	80	N/A	N/A	N/A
Dey et al. [28]	Assign docs sentiments	SO-CAL, Senti-N-Gram	Movie, books, cars, cookware, phones, hotels, music, and computers reviews	3-fold cross validation	Books	68	66	76	71
					Cars	84	76	100	86
					Computers	80	73	96	83
					Cookware	74	68	92	78
					Hotels	74	67	96	79
					Movies	66	64	72	68
					Music	64	62	72	67
					Phones	86	85	88	86