We describe our runs and results for the fixed label Wild
Wild Web Tagging Task at MediaEval 2010. Our experiments indicate that including all words in the ASR transcripts of the document set results in better labeling accuracy than restricting the index to only words with recognition confidence above a fixed level. Additionally our results
show that tagging accuracy can be improved by incorporating additional metadata describing the documents where it is available.
Item Type:
Conference or Workshop Item (Paper)
Event Type:
Workshop
Refereed:
Yes
Uncontrolled Keywords:
information retrieval; automatic speech recognition; text tagging