What did the Tweetosphere think of 2017? Twitters year in review.

2017 was really terrible, wasn’t it? We all thought that 2016 was bad, then boom, here comes 2017 with a concoction of sexual misconduct scandal, an election that completely fucked my PhD data, and a toddler hell-bent on nuclear war while continuing his temper tantrum by pissing up and all over American democracy.

But then again, are we coming from this ‘year in review’ stuff from the wrong perspective? After all there is a tendency for humans (and animals) to place more emphasis on bad emotions rather than the good. You’re going to be more caught up on losing £50 than you are on gaining £50. In addition, I’m another cynical brit – whose experience of the news dominating Brexit negotiations and general poor political landscape probably leaves me more exposed to bad events rather than the good.

It turns out rather than being the negative shithole we thought 2017 was, it turns out Twitter thought it was a bit ‘Meh’ or neutral. With the majority of tweets not having a massive deviation. If anything sentiment reminiscing about 2017 was somewhat more positive than negative, but not by a large amount.

The Juicy Analysis

I did this test during the holidays for a little fun (yes, sometimes playing with Twitter data can be fun), but also to examine what Twitter thinks of 2017. Seeing if my experience is in the minority or shared throughout the Twitter population. Plus, I thought it would be a good time to put my PhD methodical learnings to the test for a little fun for one answer, was 2017 more negative or positive.

To start, I did a quick data collection of a selection of tweets during the 31st December 2017 and 1st January 2018. It’s quite typical at this time of year for people to start posting tweets reminiscing about their year – you know them annoying ‘my year in review’ type tweets. To collect these tweets, I ran a search containing any tweets relating to 2017 or words/phrases which are often used in people reminiscing about the previous year. These phrases included: ‘2017’, ‘#2017’, ‘remember’, ‘memory’, ‘year’, etc. However, it’s also the time of year where people start posting ‘happy new year’, which would completely ball up any sentiment analysis. So I made sure any tweet containing ‘happy new’ as a phrase was removed from the data set. For good measure, I also made sure every tweet was written entirely in English, because my sentiment analysis tool is apparently incredibly xenophobic and sees any other language as inferior.

After collecting 20,179 tweets, I removed any and all Tabs in the tweets using a visual basic script (because my xenophobic sentiment analysis tool also hates anything other than a Tab delimited text file), then sent it off for sentiment analysis using SentiStrength. This takes each tweet and gives it a positive score (1 to 5) and a negative score (-1 to -5) based on the language used in the text. It doesn’t do well with sarcasm, and it’s accuracy is about 70%. However, on such a large amount of tweets it’s mostly accurate. Which is the basis of academia, right? (inb4 I get slaughtered by all political scientists). Anyway, it’s mostly a bit of fun, so please put the pitchforks down.

The tweets collected had a wide variance of sentimentality regarding 2017.

Erik here had some explicitly negative things to say about 2017:

Yeah! You go King, you go get 2018.


I will not link to the Tweet that included, 2017, the poop emoji, and an accompanying picture of a pile of faeces.

While others had a much more positive view on 2017, and well was generally a lot more wholesome:



But this is all anecdotal, you didn’t come here for the anecdotal, you came here for cold hard data.

Overall in terms of net sentiment, most tweets were neutral in terms of sentiment score. 17,036 (84.42%) of the tweets collected could be found in the -1 to 1 net sentiment range. With 1,344 (6.67%) showing a negative sentiment (-2 to -4), and 1,804 (8.94%) more positive (2 to 4). However, it should be noted that ranges of -3 or lower and 3 or higher are somewhat outliers in across the data set.

Another way of looking at sentiment regarding 2017 is by creating average positive and average scores. By netting sentiment analysis, you’re ignoring the fact that one tweet can have elements of extreme positivity and negativity within the same sentence. At the same time zero doesn’t really exist to SentiStrength, as the lowest scores are either 1 or -1. When we average the scores we get an average positive sentiment of +1.541 (SD=0.711) and a negative of -1.3431 (SD=0.7499). So once again, we can see that Twitter rated 2017 only slightly more positively than negatively.

So, 2017 according to Twitter was a bit meh, really.

Or in Emoji – 😐