SentiSquare s.r.o.

Create rules:
human-made rules and vocabularies
+precise for individual examples
+easy to understand and interpret
-domain and language specific
-impossible to cover all examples

Build mathematical model:
semantic space with high dimension
+great at covering vast majority of examples
+language and domain independent
-some cases are hard to cover
-hard to analyze and interpret
 
Our solution?
Combining the best of two worlds

• Named entity / keyword extraction
  extracting the most important parts of the text

• Distributional semantics
  extracting essential information

Machine learning
to combine the best of the two approaches

Regardless of the sector, every business today relies on large volumes of text information. Composed of internally stored organizational information such as emails, call centre transcriptions, customer information and feedback as well as external open source information and social media, this big data is largely unstructured and in a state of constant growth.

Our work starts when it is not in human capacity to read and sort data anymore. People should do what they are good at and leave machine work to machines.

Our algorithms can handle any written text. It does not matter in which language it’s written or if it contains grammatical errors or typos.

We understand the specific language of your industry, your clients and the channel they are using to communicate with you (emails, SMS, comments, posts, etc.).