Harald Bosch

About


I am an entrepreneur, founder of ScatterBlogs Visual Social Media Analytics, machine learning consultant at NovaTec and former PostDoc researcher at the Institute for Visualization and Interactive Systems of the University of Stuttgart. My research focus is Visual Analytics, the topic of combining artificial intelligence, information visualization and human-computer-interaction to facilitate complex domain specific analysis task, especially in massive unstructured datasets.

 

Projects


ScatterBlogs - Visual Social Media Analytics

ScatterBlogs detects breaking news and events from their footprint in social media and alarms journalists, risk managers, asset managers and public safety agents prior to the traditional media coverage. To achieve this, it employs statistical models and artificial intelligence to detect, locate, verify and rate potential current events in a massive data stream in real-time.

Movement Patterns from Social Media

This was a fun project to deduct aggregated movement patterns from georeferenced social media. From a global sample of georeferenced tweets, we collect movements by linking all GPS locations that were visited by the same person during one week in the summer holidays. This revealed interessting patterns like centralization of movements to hubs in France and a network of connections in Germany.

PESCaDO - Personalized Environmental Decission Support and Delivery Orchestration

The goal of PESCaDO was to make the abundance of environemental data, published in varying formats, resolution, quality, available for supporting professional as well as personal decisions. Data was searched for with spiced queries, classified using machine learning, transformed into semantic structures and made availabe for fuzzy inference in the context of task and problem definitions. These definitions could vary from leisure activities (e.g. ozone concentrations to high for sport activities) to urban governance (e.g. how high is the risk for black ice).

PatExpert

Patent documents combine the incomprehensibly of technical documents with the legalese of intelectual property rights documents. Research concerning the patentability of an inventions, the freedom to operate in a market, or searching for white spots in the patent domain are challenging and time consuming. PatExpert facilitates these research tasks by creating and combining several powerfull search capabilities for related images, semantic relations, similar documents and metadata with ad-hoc result summarization and several feedback cycles. Please have a look at the TVCG paper about PatViz, the user interface of PatExpert.

Publications


2015

2014

2013

2012

2011

2010

2009

2008