• Semantic search
  • text mining
  • agile software development

AI helps Járókelő

Fri 22 September, 2017

Járókelő is getting more and more notification and the affected organizations are getting more active, that’s why it is vitally important to provide them continuous feedback. Users can up- or down vote answers on the site for a long time, but this method yields only a few reactions so one cannot count on this method as a reliable source of objective feedback. Our new development provides automatic feedback to stakeholders so they can adjust their style according to the norms of the community.

How does it learn?

Answers have been evaluated by volunteers from Járókelő and annotators from Precognox to make a training dataset. This dataset have been used to teach a classifier algorithm which takes various features (e.g. expressions of politeness,  length of the answer or the grammatical complexity of sentences) of the text into account.

Classification is a common thing in our everyday life

Classification is one of the most common things in our everyday life. From marketing to healthcare, we have to face the deluge of information in the form of texts, mails, pictures, and etc. There are two main barriers to the automation of classification. The first is the lack of training data, which means the lack of data with labels. The second of the imbalanced nature of the available data, it is very common to have huge amount of examples from one class of our interest while having only a few examples from an equally important class.

Precognox helps its clients through the whole process of making an automatic classifier, from making an appropriate and balanced training dataset to integrating the trained algorithm into a software system.

We are ready to help you in similar problems, reach us by mail or phone.