Sentiment analysis is widely applied to voice of the customer materials such as reviews and survey responses, online and social media. The tools extract associated emotions or polarity of the associated unstructured information.
We employ several methodologies, depending on the use cases. There are trade off’s between accuracy and capabilities. For example, we can infer sentiment polarity (positive or negative view point) from texts with up to 94% accuracy. Other techniques can identify other emotions, however these are often less accurate: depending on the use case a deeper emotional isight may be preferred, trading this for accuracy.
Other models are more suited to social media sources (particularly Twitter) when sentence structure is often inconsistent (or non-existent). In such cases we can fall back on more straightforward analysis of trends in the average word ‘valence’ or warmth.
Public Sentiment Use Case
Shown below, analysis of 500,000 tweets containing the word Brexit. Each purple dot represents the average sentiment of a batch of 100 tweets. Tweets are processed in real time as they come off the live Twitter API feed.
Gwendra AI can provision similar analysis on all social media feeds as stand alone signals and analysis for your organisation or as part of the input into more complex advanced analytics.