i think it fails because I try searching in .doc MS Word files...

Right. The "old" MS-Word files (*.doc) are binary garbage, with fragments of the text you have entered. The "new" MS-Word files (*.docx) are renamed ZIP files containing several XML files and some helper files.

The embedded XML files should be searchable. Basic concept: Open the *.docx file as a ZIP file (e.g. using Archive::Zip), extract all *.xml files (or, if you know the format better than I currently do [HINT: Use a search engine to find the spec!], just the one that actually contains the text), extract the text from the XML file(s) (e.g. using XML::LibXML, apply the textContent() method to the root node), search the extracted text like you would do with a plain text document.

The old format should be readable, too. There are several libraries to make sense of the binary garbage, one of them is wv, formerly known as mswordview. Try to find perl bindings for it, or use one of the available command line tool to convert the binary garbage to a readable format (HTML via wvHtml, RTF via wvRtf, plain text via wvText). Then proceed as before. For a grep-like search, plain text is probably the most useful format.

Alexander