Multiple internal studies based on twitter corpusses indicate that the Sentimentics algorithm scores high on accuracy with values between 0.92 and 0.94 for classifying the tweets in categories positive, negative and neutral. The data supporting these results are available at request. Furthermore, we are testing the Sentimentics algorithm on within category performance. Early findings indicate that the algorithm can also discriminate within categories positive and negative with high accuracy. More information about these findings will be published as soon as possible.
We are continuously optimizating the algorithms to keep them state of the art within their field.
Because the algorithm can interpret text accurately in terms of sentiment, and in terms of extent, a complete new field of possible applications is created. Think of human-computer interaction, where the computer can process unstructured information in a structured manner, and can intelligently react to the human. Think of accurately processing huge amounts of opinions of consumers of your product, giving high value meaning to the results.
Also, the algorithm is able to generate likert-scale scores, such as used in questionnaires, again using the ability to processing unstructured information in a structured manner, creating possibilities in domains such as clinical information gathering, lifestyle information gathering, and consumer information gathering.
Continuous innovation ensures state of the art features within the field of sentiment analysis. Latest developments make the algorithms able to interpret temporal relation statements.
Multiple languages are supported by the algorithm. At this moment English and Dutch are active. We are currently finalizing French, German, and Spanish (Europe) language.
The continuous optimization progress always insures the best performance possible for each language. It is also possible to optimize the algorithms specifically to your domain of application.
