iGNSSMM

In its heart iGNSSMM consists of a chain of interchangeable NLP components, specially designed and trained for application in an on demand or ad hoc way. Within this work we concentrated on two main components, an ad hoc Named Entity Recogniton (NER) approach that identifies named entities that are related to a given topic using snippets coming from a search engine in a full unsupervised way. Second a Relation Extraction (RE) approach which extracts relations between the named entities again fully unsupervised. We furthermore equipped the system by a Concept Extraction (CE) component and adapted it for usage on mobile devices. The CE basically identifies and clusters descriptive sentences for the node of a topic graph that has been selected by the user.

All components are extremely shallow, data-driven and fast.