Zurvey does automated sentiment analysis on the content uploaded. The system automatically recognizes the positive and negative meaning of the phrases in texts. Every text is given a score, called opinion index that represents the value of the positive or negative opinion. Generally this score falls between -3 and +3 but in extreme circumstances (e.g. using inappropriate language) the score could be between -5 and +5. The calculated phrase scores are summarized for the entire text, showing the opinion of the whole comment.
The basis of the scoring is one of the most precise sentiment analysis algorithms created by Neticle Technologies available in Bulgarian, English and Hungarian languages. It works with 80%-85% precision depending on the domain. You can try the neticle Semantic API out here.
Using the basic Zurvey settings it does document level sentiment analysis which means in a text every phrases and labels are analyzed no matter what these phrases related to inside the text.
Opinion phrases are recognized from text. Every positive and negative phrase is given an opinion index between -3 and +3 based on the opinion’s strength. The sum of these scores indicates how positive or negative the whole text is.
Among the Zurvey settings users can add special categories or keywords to focus during the analyzation. This is called entity oriented sentiment analysis or keyword oriented sentiment analysis: in a text only the phrases and labels related to the set target entity are analyzed. This target keyword is set by it's synonyms, spelling and mispellings (ie: support, Support, SUPPORT).
When analyzing opinions regarding specific keywords (e.g. brand, product, company, person, etc.) by giving synonyms and variants (spellings, misspellings, nicknames) to each keyword only those phrases and labels which are related to the certain keyword are analyzed and not the whole text.
Zurvey’s text analytics also recognizes automatically the key topics and important entities inside the texts. This tagging or labeling function is really useful to understand and summerize the key topics and related sentiments in the texts.
Service and product attributes (e.g.. screen, bandwidth), key topics, places, people, brands, emotions and organizations (e.g.. 3G, mobile payment) are recognized from unstructured text.
The combination of tagging and sentiment analysis helps to understand the trends and important focus points inside the upload texts.
Leonardo da Vinci 41. (2nd floor/14)
Budapest 1082, Hungary