Hootsuite Analytics uses a technology partner to provide sentiment. They use general algorithm technology to combine both machine-learning and natural language processing.
The accuracy of classification algorithms like language and sentiment detection can vary. They are built by first analyzing a training set of data. This is a collection of millions of examples that have been categorized for language and sentiment by humans. The algorithm recognizes patterns from the training data, and classifies new data by matching it to the existing patterns that it knows.
In certain industries, a general detection set may not match new data with accuracy, especially for sentiment, based on the subject matter. In a medical context for example, saying "This drug can be deadly if overdosed" is a neutral statement, but might be classified as negative based on the majority of patterns that the algorithm can match it with.
Analytics analyzes mentions and predicts sentiment in 18 languages: Arabic, Chinese, Danish, Dutch, English, French, German, Hebrew, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Russian, Spanish, Swedish, and Turkish.