Performance

You can get DBpedia Spotlight to run faster if:

• you put your index in a SSD (Solid State Drive) or a RAM drive -- anything that makes disk reads to the context index very fast
• put the entire context index in memory (not yet released as of 0.6, but possible from our latest source code)
• warm up the index so that the caches are built with the most common words (you can do this by setting up several queries to the system with the kinds of text you will want to annotate later)

The CoOccurrenceBasedSelector will also speed up the whole annotation process by removing some unnecessary disambiguations at spotting time, but it needs you to download and point the software to some extra models (see configuration file). The files are available here: http://spotlight.dbpedia.org/download/release-0.5/spot_selector.tgz

Another idea is to focus your index. If you're only looking for sports things, you can build a much smaller index with only sports data that will perform much faster.