Four reasons why sentiment analysis matters to you – part 1

There are many reasons for you to carry out sentiment analysis and word studies, but those listed in this and part two are four of the best.

Last time I gave you an overview of why word sentiment and word count matter, how it adds to our understanding of text, and can give pointers where to focus. This time we’ll look at specifics.

Rorschach test

Top reasons for sentiment analysis

It’s easy to measure some factors for understanding why a piece performs well or not – it has typos, is too long, too short, or wrong for the audience. Yet sentiment analysis offers an extra layer.

Much as contour lines are added to a map to give relevance, sentiment analysis can be added to traffic data to see how word use hits website traffic and other key measurements. Here are ways it can help:

1. To give us a wider understanding about writing styles and performance.

In an ideal world we have access to all data to get an idea of how it affects traffic – does a joyful piece cause people to read more of your site or work, share it with others, and does a negative, angry piece cause readers to comment more?

As media mogul David Ogilvy was very much aware of this on the copy for advertising, negative words were avoided as to avoid the associated negativity with it, even if the ad copy was being negative about a rival. As he put in Confessions of an advertising man:

Research shows that it is dangerous to use negatives in headlines. If, for example, you write OUR SALT CONTAINS NO ARSENIC, many readers will miss the negative and go away with the impression that you wrote OUR SALT CONTAINS ARSENIC.

This then is the biggest use – what immediate effect does the sentiment, tone and word use have on your audience and their behaviour?

2. Who is the author really addressing?

Word count and sentiment tools can be used to find out who the reader is taking to. At its simplest, it is pronoun user, whether they tend to us I, we, us, you or them.

As with Freud and his slips, we can also get ideas on a subconscious level what they are thinking about or their psychological state.

For example, dwelling on death is indicated not just by words associated with mortality, but self-obsession and negative words. This is true for other topics – if money is on your mind  you tend to talk more about cash, dollars and so on, and if you are religious your words are more divine (or at least associated with those upstairs). As Gordon Gekko said in Wall Street:

It’s all about bucks, kid. The rest is conversation.

Take this piece by Netflix CEO, a half-arsed apology. It is all about Reed Hastings – how he felt, what he wanted. But going by my own experience of company-wide emails, he’s not the alone in being a self-obsessed top man. And they say it’s lonely at the top.

So we can use sentiment analysis to judge an author or character’s true audience and what is on their mind, we can reveal their subtle thoughts.

3. Sentiment analysis as an idea of the author’s relationship with others.

Word use can tell us a lot about their relationship with others, chiefly through word choice and the amount of words used.

This means analysis usually works best for group situations, as who speaks the most and how they address others is a big clue. Pronoun use is one such way. Oddly the results aren’t what you may expect – talking about yourself can be a submissive type in a lower ranking relationship.

This sounds contradictory to what I just said about CEOs, doesn’t it? I’m not alone in noting this, but think of it this way. For example, when writing to someone you want a favour off of you may put”I’m sorry to bother you but it’d help me if I could borrow your mower as mine is broken.” I, me, mine. Yet if you were the CEO in desperate need of a mower (say his gardener rode off with it) it is more likely you would put “send the mower round this evening.” It’s true these examples are exaggerations, but they are also true.

Think of the last time you applied for a job, the amount of “I’s” and ‘me’s’ in the cover letter, were you to study a transcript of a job interview you would see who holds the power and who is the interviewee.

So why did we look at the Netflix apology? Well that was what it was – an apology and in an apology the person showing regret is (meant to) be in the lower quality relationship.

We can also look at terms that put people off. For instance, using “you” too much is not personal, it’s off-putting and creates distance. In fact relatives who used this in one study were deemed to be ‘over involved’ and scored high for criticism.

So while imperfect, we can make some judgments of the author’s relationship with others, though we have to take the reading with others.

Is the use of “I” and “me” a sign that they are in a lower ranking relationship to you (or pretending to be in one in order to flatter), or are they self involved? One study I carried out at the BBC years ago found that financial editor Robert Peston wrote about himself about 50% more than Stephanie Flanders, the economics editor. The amount of times Robert Peston’s name is linked with self-important is a lot more than her.

Google predicts what you're going to type

Google predicts what you’re going to type

Indeed higher positive emotion suggested a stronger relationship, or at least a happier one. We can also look at the topics and their thoughts.

For instance – and let’s hope we don’t find many in my studies – depressed and even suicidal writers tend to be more self-focused, show more negative emotion and use more death-related words. In fact the study by Pennebaker and colleagues linked to above  suggest that studies of writing are a good way to determine the mental health of an individual.

“This individual difference may show an attentional difference, that is, more self-focus in response to emotional pain, or it may indicate a thinking pattern that is a predilection for experiencing depression.”

So we can sum up what the reader is thinking about – you, me, them, everybody.

Next time – acceptance and summary