Understanding the Sentiment Behind the Statement


Sentiment analysis is invaluable to the marketing function. Sentiment analysis models that understand the sentiment behind customer statements can produce both macro-level analyses, like sifting through social media postings and online reviews, and micro-level insights, like keeping tabs on individual customer interactions. These insights can be used to inform PR and Marketing campaign strategies and used to disseminate those campaigns to specific geographical locales and individual customers.

However, understanding sentiment is a difficult proposition because of the nuanced nature of human speech. The simple statement, “I love this stupid little office”, for example, requires deeper interpretation. For a sentiment analysis model to succeed at this difficult task, it needs high-quality training data. Only then can models deliver true value to the businesses they are created for.

Recently, we assisted a Fortune 500 tech company to create a model that could understand the sentiment behind Korean phrases. To do this, we needed to collect 10,000 input phrases and then we needed to annotate the sentiment behind each of them. This high-quality data would be used to train the model to understand the real meaning behind a person’s statement.