Streaming-based Text Mining using Deep Learning and Semantics

Since the most of the world’s data is unstructured, the mining of required information from text was, is and will be essential. However, the requirements for text mining are becoming more complex, as rare languages or new domains need to be supported, social media content has often a worse grammar or mixed languages, and the amount and velocity of the data growths constantly. To compete with these requirements, the current research progress regarding Deep Learning (DL) for Natural Language Processing seems to promising. Within the LEDS project, Ontos rely on them in order to develop a DL-based MINER in order to extract required information from arbitrary texts. This service is embedded into a streaming platform that allows for 1) flexible text mining with a high throughput and 2) a disambiguation and linking against existing knowledge bases. In this industry talk, we will provide insights regarding concepts, implementations, and the proof-of-concept for news and social media aggregation.