In this paper, we take a new look at the problem of analyzing course evaluations. We examine ten years of undergraduate course evaluations from a large Engineering faculty. To the best of our knowledge, our data set is an order of magnitude larger than those used by previous work on this topic, at over 250,000 student evaluations of over 5,000 courses taught by over 2,000 distinct instructors. We build linear regression models to study the factors affecting course and instructor appraisals, and we perform a novel information-theoretic study to determine when some classmates rate a course and/or its instructor highly but others poorly. In addition to confirming the results of previous regression studies, we report a number of new observations that can help improve teaching and course quality.

Electric Vehicles (EVs) are envisioned to play a large role in the transition from fossil fuel to renewables based transportation. However, their sales thus far are nominal compared to traditional car sales. It has been difficult for manufacturers to measure owners' initial perceptions in order to build improved vehicles more drivers are likely to adopt. Sentiments towards EVs have mostly been determined using either field trials or large surveys of drivers, both of which are problematic. We build a system that mines EV owners' sentiments from online forums. Our system has three main uses. First, it graphs the percentage of positive and negative opinions for each vehicle feature of interest, e.g., battery capacity, giving the user a high level product overview. There is currently no easily-consumable review system for EVs. Second, it allows the user to read opinions about the specific features they are most interested in without searching though irrelevant text. In our case study, we find only 3% of the comments on EV ownership forums express opinions on the features. The system therefore reduces the space of text the user must read by 97%, even assuming they wish to read all opinions about all features. Finally, in addition to mining the same perceptions found during expensive field trials, our system finds perceptions that were only realized after the owners possessed their EVs for an extended period of time, i.e., perceptions not available during shorter trials. The system extracts and classifies opinions with a precision and recall of 60%, which is on par or better than previous opinion mining systems.