One common challenge many communication professionals are faced with is knowing how not to be biased. We get it; avoiding bias can be hard when we feel so attached to a plan that we’ve nurtured from the beginning. Although it can be painful to admit and say “this plan just isn’t working”, sometimes that is exactly what is needed in order to move onwards and upwards! A simple online news and social media brand sentiment analysis is a great way to understand our current situation from our audience’s point of view, objectively.
Brand sentiment analysis helps us to gain an understanding of how a person, a brand, a product, or a campaign (amongst other things) is perceived by those we’re trying to target. We can also filter sentiment by trending themes in order to gain a high-level view of exactly what it is our audience likes and doesn’t like. We can then optimise our brand strategy accordingly.
Any good media monitoring tool can provide such insights. Meltwater’s tool offers both social media and editorial brand sentiment analysis and is currently supported by 16 languages including Chinese, Japanese, Portuguese, German, English, Spanish, Arabic, Hindi and Korean.
Sentiment Analysis is a research area examining people’s opinions, attitudes and emotions in written language through the method of Natural Language Processing (NLP). The Oxford Dictionary explains this as:
The process of computationally identifying and categorising opinions expressed in a piece of text, especially in order to determine whether the writer’s attitude towards a particular topic, product, etc. is positive, negative, or neutral.
Opinions are central to almost all human activities. They are key influencers of our behaviours. Seeing as marketing, comms and PR activities aim to influence opinions and behaviours, having a clear understanding of how our audience – be is prospects, clients or influencers – feels is important when trying to move the needle.
The growing need for sentiment analysis coincides with the ever-expanding importance of social media presence of brands and companies. This is especially the case since the introduction of social media, with conversations spiking across the social networks rising by the minute.
As a global leader in media intelligence, we pride ourselves on our sentiment analysis capabilities. So how do we go about offering over 26,000 client’s sentiment insights?
The fundamental task in brand sentiment analysis is text classification – classifying the separation of a given text or whether the expressed opinion in a document is positive, negative, or neutral. Around 800 documents pass through our platform per second from different media sources and providers. We use Natural Language Processing (NLP) to judge which group (positive, negative or neutral) the content belongs to.
Meltwater’s Natural Language Processing model is supported by AI and machine learning algorithms. Using this model, we take individual words into account. Each document, for example, a tweet, is analysed based on the words it contains. Then we map the words to a set of predefined data to see the number of occurrences where they match up.
Let’s use the following sentence as a simple example: ‘Tesla is the best car.’
We are applying a classification algorithm to determine the probability of how positive or negative the word is in the context of the wider text. At the end, we pick the value with the largest probability score.
We think that transparency is key and like many other brand sentiment analysis providers will tell you (at least we hope they will), the metric isn’t foolproof. The Natural Language Processing model Meltwater uses isn’t guaranteed to identify limits of human language – for example, sarcasm, irony, context interpretation, use of slang, cultural differences and the different ways in which opinion can be expressed (subjective vs comparative, explicit vs implicit).
With that being said, Meltwater has a huge repository of exploitable data, one of the largest in industry, making us well placed to investigate different ways brand sentiment and tonality can be drawn, and therefore pulling highly accurate insights.