Ontology-based Recommendation from Natural Language User Descriptions: Recommending Eco-Innovations to Companies

The paper reports on the development of a recommendation system which offers complex services and green technology solutions for companies. The system must work in a predominantly data sparse environment with respect to traditional evidence of user preference and therefore has to seek other ways of collecting the information necessary for principled recommendations. We utilize and adapt methods from natural language processing, text analysis, information retrieval, information extraction and knowledge engineering in order to overcome the data sparseness problem. The system works with Hungarian language data but with the necessary resources developed the framework is applicable in other languages as well.