I would like to propose an idea to improve phonetic strings handling for multilingual environments. I believe Tika should not concatenate phonetic strings because text with phonetic strings is recognized as noisy text in most situations of natural language processing.
Excel files include phonetic strings in some languages such as Japanese, Chinese and so on. Apache POI concatenates phonetic strings onto the shared strings when Tika extract text from Excel files.
Recent Apache POI has an switch flag for phonetic strings concatination as follows:
Tika should set the 2nd argument "includePhoneticRuns" as false. Here is the simple patch for my idea.