The data files have a who attribute with value "#interviewer" for full utterances by the interviewer. The code for generating the full transcript captured that attribute and generated the correct xhtml class attribute. The javascript code generating the subtitles texts did not. I added an isInterviewer key to the caption object in js_captions.xsl and then in captions.js I test for the value of the isInterviewer key to determine whether or not to add a class="interviewer" to the output element. I tried doing it all in the js_captions.xsl (to save more processing in the js loop in captions.js, but couldn't get it to go.