Before I read this interesting presentation by Simone Teufel at the Text Mining for Scholarly Communications and Repositories Joint Workshop held at the NACTEM, I did not realized how useful may be Sentiment Analysis and Opinion Mining applied to scientific literature.
Scientific papers are read by many people, with different profiles. You can be editing a journal, reviewing it for a conference, or informing yourself with respect to your research (among other things...). In each role, you may want to check an specific part of a paper: the one in which the technique is described, the related work one, the one which summarizes (and sells) the novel contribution, etc. Opinion Mining (with the specifics of scientific texts) can detect those sections you should read first, or which just act as a discharge summary of relevance for investing more time on the paper, for instance. The following paper is annotated with the semantic (scientific) functions:
The work by Teufel also focus on the analysis of citations. Used to support your work or mine, in contrast with it, etc? For instance, the following graph shows a paper, some of the citations and if they are supporting or contrasting:
I find it very interesting and potentially work-saving. Although as tools able to do this analysis go to e.g. Scientific Publishing market, there will be "adversaries" trying to take advantage of them in order to get some low quality papers accepted in journals and conferences... Another adversarial text classification problem!