... subfiles of the DOCX. If this doesn't work a second attempt is made where the corrupt XML subfiles are truncated at the first error, and the correct ending tags are again added with xmllint.
If all else fails, SilverCoder's DocToText is used to extract text.
Try also http://wordcorruptdocchecker.codeplex.com/ and https://support.microsoft.com/en-us/kb/2528942 and my other SF projects: Corrupt Extractor for Microsoft Office, Corrupt DOCX Salvager, S2 Recovery Tools for Microsoft Word.