My client built a besoke text mining web-based Java application based on weka library. The application analyses medical free text captured by users during interaction with doctors. Application is mature and proved successful and its value is proven in a single country and single language (English). There is a need to scale out the solution to serve many countries, many languages (including European, Chinese, and Japanese) and hundreds of users. This raised the question if we are better off redeveloping the solution using a hadoop architecture.
I am looking for examples/options for haddop-based architetures that deliver text mining/NLP capabilities. Appreciate any help in this regards.