Our blog

Today's world pushes towards a paperless society. Even in this environment, automating the classification of documents has a role to play.

A number of business processes can benefit from both automatic document classIfication and specific information extraction, converting human-readable, unstructured data into traditional structured information.

We consider how the design of a state-of-the-art conversational design platform should be to allow for wider adoption of this technology.

Some challenges to be addressed are maximum usability, high flexibility in configuration, easy and secure deployment and scalability, without forgetting fast feedback from production into retraining to achieve better user experience as the conversations evolve.

We mix good old lexical analysis with Javascript and CSS3 tools to provide a nice to read appearance to spreadheet-like formulas as you write them.

The HTML5/CSS/js code used in the article is available on Github.

We explore the possibility of identifying the author of a book just looking at three or more sentences. And all this just considering the syntactical constructions without even examining the actual words.

The Python code used in the article is available on Github.

We play with StanfordNLP, a Python package that solves several common tasks in Natural Language Processing in 53 human languages. We demonstrate the usage of the package in finding out parts-of-speech and lexical features in sentences and paragraphs.