Figure 7
Google Books sentiment time series from each sentiment dictionary, with stop values of 0.5, 1.0, and 1.5 from the dictionaries with word scores in the 1-9 range. To normalize the sentiment score, we subtract the mean and divide by the absolute range. We observe that each time series has increased variance, with a few pronounced negative time periods, and trending positive towards the end of the corpus. The score of labMT varies substantially with different stop values, although remaining highly correlated, and finds absolute lows near the World Wars. The LIWC and OL dictionaries trend down towards 1990, dipping as low as the war periods.