Social Media is here to stay and a major part of every product managers plan and life now. And it is not just limited to products, every person on Social Media is now a brand and takes a lot of effort to build and maintain a brand vibe around her/him. With the advent of this wonderful platform have come a plethora of methods and metrics to measure how one is doing on Social Media and how to track it. One keeps hearing “Social Sentiment” in that context a lot! Let us try and take a better look at this phenomenon here. More specifically, can we/ should we and how do we measure this sentiment on social media.
So what is social sentiment? To analyse that, we need to understand what an opinion is.
An opinion is an expression (a binomial to be precise) that consists of two key components:
- A target (which we shall call “topic”, as referred to by most social Analytics tools);
- A sentiment on the target/topic.
Here the target is the product/service/person/event that the social campaign is talking about. So in ” I love my xyz smartphone” – xyz smartphone is the target and the sentiment as seen in “love” is positive. So it follows that the sentiment here is positive. This sounds really simple, but then we have semantic shifters (less), connectives (but) and modals (should) that change the way you interpret sentiment in a statement.
Other than individual words, expressions and phrases can be used to show sentiment. Take for example “I’m over the moon“. If taken word for word, ‘over‘ and ‘moon‘ are not sentiment words, as they don’t express any positivity or negativity. But in effect, we know that this statement is an expression of extreme satisfaction/happiness. Hence, we can conclude that sentiment cannot be accurately measured from word-by-word analysis, but only on a level that allows for semantic interpretation.
So there are possibilities that if you are using dictionary word based sentiment trackers, you may be way off the target when it comes to reading sentiment. That means that Automatic Analysis needs to be complimented or augmented with NLP and AI to come close to accurately estimating or measuring sentiment.
There are 5 main factors to look at when doing sentiment analysis in social listening:
- topics: what are the main areas of discussion?
- aspects (subtopics and attributes): what about those topics is being talked about?
- sentiment: what is the sentiment of the content and the opinions contained?
- holder: whose opinion is being discussed? Are there multiple in the same content? If so, how do they differ, if at all?
- time: when was this content posted?
If you need to automate, automate correctly, get a tool that can take these factors into account. And when we measure, we need to account for subjectivity and emotion also. Else, the sentiment estimate will be flawed and misleading.
There are four main factors that currently prevent us from blindly using tools for sentiment analysis:
- Context: when I say that my ISP is efficient at over billing me, I am really not giving them a compliment
- Sentiment Ambiguity: when I ask ” can you suggest a good hotel?” it is not a complement and when I say that my browser uses a lot of memory, I don’t mean it in a good way
- Sarcasm: something that has amplified multiple folds on social media, when I say “Sure, I’m glad that my system crashed!” I really am not glad, am I?
- Language: a word can change sentiment and meaning depending on the language used. This is often seen in slang, dialects, and language variations.For example, “cool“, which can change meaning based on context, tone and language, although clear to the target audience.
To put it simply, Can we? Yes we can! But it is fraught with the proverbial traps that we can easily fall into. The current state of sentiment tracking is not yet there to do it automatically with tools, human analysis and interpretation are the key to this and will be for some time to come!
Note: I have based this post on the article here.