OSM as an NLP Resource: A new Corpus for Semantic Parsing & Beyond Current corpora for semantic parsing are either not big enough, not complex enough or cannot be used with supervised approaches. I created a new corpus which is based on the database of OpenStreetMap.org (OSM), the SPOC corpus (SPatial Open Corpus). Its basis are 900 hand-crafted sentences in German and English with corresponding machine readable formulas pertaining to landmarks and areas in the world. One example is the following: "How many hotels in Paris have wheelchair access?" The corpus was then extended automatically making use of the queryable nature of the OSM database. This resulting set includes over 700,000 questions about 30 cities in Europe. I will detail the corpus' creation, especially the designing of a machine readable language for the corpus based on OSM's query language Overpass, and how its statistics compare to other semantic parsing corpora. I will then proceed to present the semantic parser that was adapted to be able to run with SPOC, the various extensions and improvements and the experimental results. In a final step, I will demonstrate how I intend to use the resulting semantic parser: The parser will be used to give feedback to an SMT tuning algorithm, REBOL. The feedback given to the SMT system will be positive if the translation of a question leads to the same answer as the gold standard machine readable formula of that question, otherwise the feedback will be negative. REBOL will then use this feedback to improve the translations.