Characteristics of Multimedia Information Retrieval
- Content-based retrieval
- from pieces of information to the full picture
- "Who is the author of this music?" or "Where does this picture come from?"
- "Is there any design like this?"
- From the viewpoint of subject analysis: A picture is worth a thousand words
- From the viewpoint of a reference librarian: A word is worth a thousand pictures
- Automatic indexing
- "Ask not what the country can do for you, ask what you can do for the country"
- Pieces of information is sometimes more valuable, while such indexing by human is too costly
- similarity matching
- similar contents may have different presentation
- data filtering rather than exact matching (data selection)
- browsing and relevance feedback
- No ideal mathematical model for defining similarity, human feedback is required
- A MIR query may involve more interaction between human and machine
Notes: