We work on
In order to understand messages in an automated way, first you have to recognize the words with which they were written.
In writing, users vary word endings, make writing and typing mistakes, or intentionally deform words.
Our technology recognizes words in more than 57 million possible variations –no matter how they are written– by registering their roots: it’s not necessary to keep an endless list of all possible word combinations in order to search for them.
Keepcon automatically labels each comment, identifying the theme mentioned in it with its respective feeling.
Comments with several themes get multi-labeled, allowing an integral and exhaustive comprehension of the specific thematic.
Once the words are recognized, our technology looks for the specific combination of nearby words found in the text.
Capillarity and precision in the automatic classification
Our advanced semantic technology, together with the optical recognition of characters (ORC), enables to extract and extend in a few minutes the whole content of hundreds of pages, regardless of their technicities or formats.
We know how to identify sections and obtain specific data.
Not only do we take advantage of automatic classification in order to understand an isolated message, but we also understand the “conversation” linguistic element in order to detect aspects inside the conversation dynamic. For example, we can know if an operator is or isn’t answering the customer’s enquiries, if he or she is administering correctly the pauses or the anger.
Conversation as a complex linguistic element
We can also carry out a conversation in real time through a chatbot architecture that combines the advantages of static navigation menus with the possibility of the customer expressing in a natural way, as he or she would do with a person.