id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_5lcn66xezff75mzhcf2ufszade	Robert West	Exploiting Social Network Structure for Person-to-Person Sentiment Analysis	2014	13	.pdf	application/pdf	8737	986	72	Linguistic sentiment analysis suggests another path forward: one could leverage textual features to predict the valence of evaluative texts Separate sentiment or signed-network models will miss or misread these signals. Many social networks encode person-to-person sentiment information via signed edges between users signs that (1) agree with the predictions of the sentiment model, and (2) form triangles that agree with respectively, and pe is the probability of edge e being positive according to the sentiment model alone. Intuitively, the more the inferred edge sign xe deviates from the prediction pe of the sentiment model, an HL-MRF to predict edge signs based on triangle sentiment model for all blue edges, and the signs of During testing, the network structure of all yellow edges, the sentiment predictions for all yellow Figure 5: Normalized cost λ(i) (defined in Sec. 4.3; logarithmic scale) for deviating from sentiment-model predictions pe, for bins i = 1,...,10 (Wikipedia).	./cache/work_5lcn66xezff75mzhcf2ufszade.pdf	./txt/work_5lcn66xezff75mzhcf2ufszade.txt