Here at Social Report, we are often asked if we do “sentiment analysis” and we always find it difficult to answer this question with a simple ‘Yes’ or ‘No’.
Let’s first try to define the problem space.
Sentiment analysis is a phrase that is thrown around a lot. Almost like the word ‘Cloud’ – people tend to think of it as panacea for all everyday problems. Sentiment identification implies ability to identify feelings or an attitudes. In social media world – it is often thought about in terms of tweets or comments having a ‘sentiment’. Is this tweets positive? Is this reply negative?
There is one other important dimension to this issue. The sentiment is not something isolated. Attitudes and feelings are always expressed toward a certain target – a person, a topic, a place, etc. Therefore when we talk about ‘sentiment analysis’ – we must also think that it is the sentiment toward something specific that we need to be looking for.
On the surface it seems like a fairly simple problem to solve. If you, the human, look at a sentence – you can quite easily tell if it it s positive or negative. Did you ever think about what criteria you used to determine that? Did presence of certain words tell you that it was a positive or negative sentence? Most likely it was a combination of factors: words, phrases, their position in a sentence, discussion topic, timing, author’s name, attached images, etc.
Over the years I’ve seen several platforms attempt to deliver this type of analysis – when you scratch the surface however – they all fail. These tools are typically really good at simple sentences and they fail miserably for more complicated conversations and discussions.
What’s worse? Not having sentiment analysis or have it be inaccurate in 20%, 30%, 50% of the time?
We thought about this challenge here at Social Report and decided that instead will not go the route that is guaranteed to work in some cases, instead we decided to empower our users to create their own sentiment analytics by giving them a set of very simple tools.
- Our users can track social content using keywords. This can be content located globally or within your own social media profiles (i.e. comments, replies, etc)
- Use any keywords you like – keywords that you will feel imply attitude or feelings.
- Have ability to include and exclude keywords for more refined searching capabilities.