.txt is a file saving format (for plain text), and XML is a mark-up format. An XML can also be saved as a .txt file.
To analyse collocation in Chinese text using Xaira, you must first of all tokenise/segment the text using software such ICTCLAS, the result would look like: word1/tag1 word2/tag2... Then you can convert the wor-tag pair frmt he backslash style to XML style - I uploaded some perl scripts for such jobs to the site last year (just search the site). It is easy to make the whole file XML-compliant: just add a tag at the beginning and end of the file, by yourself, or usng Preprocessing in Xaira Indexer tool.
After indexing your corpus, you can then use the Xaira client with your corpus.