Natural Language Understanding

Natural Language Understanding

Faktion has made it its mission to push the boundaries of what is possible with Natural Language Processing. We make a difference by applying computational linguistics to solve big, real world problems. From classification, prediction, information extraction to personality, demographical and mood analysis.

Why now?

One of the most interesting subfields of Artificial Intelligence is Natural Language Understanding. Modern processing power and new techniques have made an enormous amount of use cases possible that were science fiction up to a couple of years ago. Faktion has made it its mission to push the boundaries of what is possible with NLU. Our conversational intelligence platform is the most publicly visible result of that.

We especially love doing things with it that are useful in the real world. Examples include

  • Intent recognition and entity extraction. We have benchmarked our intent recognition models and have found that our models are between 2 and 10% more accurate than the major technology providers. Our models can also run on premise, in private cloud and can be adapted to suit specific use cases.
  • Classifying and tagging text. Whether the data source is tweets, e-mails, documents, website or books, you want to be able to better search, quantify and gain insight in your existing and future collections. At Faction XYZ, we do research on benchmark breaking NLP models.
  • Regression and prediction. Unstructured text can be used as an additional input for regular predictive models. Predictions from unstructured text are used in insurance claim processing, AI jury members, predictive maintenance on inspection logs, etc.
  • Information extraction. It is a sad given that most of the information in the world is in an unstructured form. Recent advances in information extraction however have made it feasible to automate large parts of the information extraction process. By using ontologies and parsing trees, Machine Learning models can find connections and patterns between words with a high degree of accuracy. This saves companies immense amounts of time by having the right information in a structured format handy.
  • Automatic response generation. Knowing what people are saying is great, but being able to respond semi-automatically or fully automatically is where you can save time and money. Depending on the use case, AI can be used to suggest answers, choose the correct answer from a number of predefined templates, or fully generate text itself.
  • Document similarity and script adherence. We have experience in analysing call center transcripts for script adherence metrics and quantifying the impact of that.
  • Voice, translation, transcribing. We have speech recognition models for Dutch, French and English that are among the most accurate in the world. These can be customised to your use case and deployed anywhere from your private cloud to your own devices.
  • Personality, demographical and stress analysis. All people write and talk in different ways. By analysing their voice timbre, language patterns, emoji use, vocabulary, etc., you can reasonably estimate a person’s age, gender, level of education, or stress level at the moment. This data can be used for personalised advertising, gaining more insight or reacting in a more personal way.