. Author manuscript; available in PMC: 2010 Dec 1.

Published in final edited form as: Int J Med Inform. 2009 Jul 9;78(12):e59–e67. doi: 10.1016/j.ijmedinf.2009.05.003

Table III.

Lucene retrieval results (IR) for topics included and excluded from machine learning (ML) experiments. The IR results for caption retrieval with query expansion (text) are shown.

Topics	Features	IR		ML
Topics	Features	MAP	P@5	Precision	P@5
14 included in ML	Text	0.198	0.471	0.341	0.443
14 included in ML	Text + image	0.041	0.129	0.372	0.457
7 best in ML	Text	0.202	0.400	0.588	0.714
7 best in ML	Text + image	0.043	0.200	0.744	0.771
7 worst in ML	Text	0.098	0.271	0.095	0.171
7 worst in ML	Text + image	0.019	0.029	0	0.143
16 excluded from ML	Text	0.112	0.375	Topics excluded from machine learning experiments due to lacking or insufficient positive examples
16 excluded from ML	Text + image	0.033	0.100
All topics	Text	0.153	0.420
All topics	Text + image	0.039	0.119