Differences between revisions 2 and 58 (spanning 56 versions)

Results for Exercise Sheet 2 (Ranking)

Add your row to the table below, following the examples already there. Please write down all values with exactly two digits of precision. In the first column, write your first name, or name1+name2 if you work in a group.

Name	P@3	P@R	MAP	BM25 Parameters	Refinements
Baseline	0.57	0.38	0.37	k=1.75, b=0.0	None (baseline)
Robin	0.57	0.41	0.35	k=1.25, b=0.75	prefer exact matches and popular documents
Marco	0.60	0.42	0.39	k=0.75, b=0.25	prefer popular documents (in a very primitive way)
Julian	0.60	0.43	0.40	k=0.75, b=0.0	prefer popular documents
Michael+Max	0.60	0.46	0.39	k=1.42, b=0.21	prefer "exact" matches and popular documents
Lord Voldemort	0.63	0.44	0.40	k=1.25, b=0.0	Dark Magic
T-Izzle	0.57	0.37	0.40	k=0.75, b=0.0	Boost "all word hits" and hits in Title
Frank	0.57	0.47	0.39	k=1.75, b=0.5	Prefer popular documents
Zhang	0.63	0.43	0.46	k=1.2, b=0.0	Prefer exact matches and documents with low tf words
Hardware Hans-Peter	0.57	0.39	0.38	k=1.5, b=0.0	Prefer popular documents
Lötkolben Ludwig	0.63	0.42	0.38	k=1.0, b=0.2	Boosted popular documents
MS-DOS Manfred	0.60	0.40	0.38	k=1.5, b=0.2	Used MS-DOS for development
Natalie	0.67	0.46	0.41	k=0.8, b=0.4	prefer popular documents and exact matches
Janosch	0.53	0.41	0.37	k=0.75, b=0.25	dust
Thomas+Frank	0.57	0.42	0.38	k=1.5, b=0.25	prefer popular documents
Lukas	0.63	0.44	0.39	k=1.5, b=0.4	prefer popular documents and boost multiple hit query words
Anup	0.60	0.42	0.39	k=0.6, b=0.0	just parameters (k and b) manipulation
Alex	0.60	0.42	0.41	k=1.75, b=0.25	prefere popular documents and reward the occurence of all query-words
Dou Nut	0.40	0.02	0.03	k=1.00, b=0.65	Special ranking mechanism implemented
IE	0.43	0.23	0.23	k=1.5, b=0.7	An attempt wars worth (thanks google translate!) prefer popular docs
Raghu	0.37	0.23	0.20	k=1.75, b=0.75	Tried multiplicative merging
Sven	0.43	0.34	0.29	k=0.75, b=0.25	prefer some of the popular documents
BIOS Bernard	0.58	0.43	0.40	k=1.2, b=0.0	prefer popular movies
Evgeny+Numair	0.57	0.44	0.40	k=0.75, b=0.0	prefer popular documents
Jan	0.24	0.11	0.10	k=1.2, b=0.75	baseline
KC	0.57	0.34	0.35	k=1.25, b=0.2	prefer popular records, added weight for keys occur in title
Elias+Maximilian	0.5	0.31	0.3	k=1.4, b=0.04	prefer popular records, boost if multiple search terms hit
Matthias	0.5	0.07	0.34	k=1.23, b=0.53	best neighbour optimizer + (far to small) training-set + popularity ranking
Daniel	0.3	0.14	0.13	k=1.75, b=0.4	Only parameter adjustment

-  ⇤ ← Revision 2 as of 2015-10-27 12:00:44 → 
  Size: 327
  Editor: adpult
  Comment:
+   ← Revision 58 as of 2015-11-03 13:57:21 → ⇥
  Size: 3192
  Editor: 10
  Comment:
-Deletions are marked like this.
+Additions are marked like this.
 Line 3:
-Add your row to the table below, following the examples already there. Please write down all values with exactly two digits of precision.
||'''Name''' ||'''Mean P@3''' ||'''Mean P@R'''||'''Mean Average Precision (MAP)'''||
||Baseline || 0.47 ||0.35 ||0.29 ||
+Add your row to the table below, following the examples already there. Please write down all values with '''exactly two digits of precision'''. In the first column, write your first name, or name1+name2 if you work in a group.

||'''Name''' ||'''P@3''' ||'''P@R'''||'''MAP'''||'''BM25 Parameters'''||'''Refinements'''||
||Baseline || 0.57 ||0.38 ||0.37 ||k=1.75, b=0.0 ||None (baseline) ||
|| Robin || 0.57 || 0.41 || 0.35 || k=1.25, b=0.75 || prefer exact matches and popular documents ||
|| Marco || 0.60 || 0.42 || 0.39 || k=0.75, b=0.25 || prefer popular documents (in a very primitive way) ||
|| Julian || 0.60 || 0.43 || 0.40 || k=0.75, b=0.0 || prefer popular documents ||
|| Michael+Max || 0.60 || 0.46 || 0.39 || k=1.42, b=0.21 || prefer "exact" matches and popular documents ||
|| Lord Voldemort ||0.63 || 0.44 || 0.40 || k=1.25, b=0.0 || Dark Magic ||
|| T-Izzle ||0.57 || 0.37 || 0.40 || k=0.75, b=0.0 || Boost "all word hits" and hits in Title ||
|| Frank ||0.57 || 0.47 || 0.39 || k=1.75, b=0.5 || Prefer popular documents ||
|| Zhang ||0.63 || 0.43 || 0.46 || k=1.2, b=0.0 || Prefer exact matches and documents with low tf words ||
|| Hardware Hans-Peter || 0.57 || 0.39 || 0.38 || k=1.5, b=0.0 || Prefer popular documents  ||
|| Lötkolben Ludwig || 0.63 || 0.42 || 0.38 || k=1.0, b=0.2 || Boosted popular documents  ||
|| MS-DOS Manfred ||0.60 || 0.40 || 0.38 || k=1.5, b=0.2 || Used MS-DOS for development ||
|| Natalie || 0.67 || 0.46 || 0.41 || k=0.8, b=0.4 || prefer popular documents and exact matches ||
|| Janosch || 0.53 || 0.41 || 0.37 || k=0.75, b=0.25 || *dust* ||
|| Thomas+Frank || 0.57 || 0.42 || 0.38 || k=1.5, b=0.25 || prefer popular documents ||
|| Lukas || 0.63 || 0.44 || 0.39 || k=1.5, b=0.4 || prefer popular documents and boost multiple hit query words||
|| Anup|| 0.60 || 0.42 || 0.39 || k=0.6, b=0.0 || just parameters (k and b) manipulation ||
|| Alex || 0.60 || 0.42 || 0.41 || k=1.75, b=0.25 || prefere popular documents and reward the occurence of all query-words ||
|| Dou Nut || 0.40 || 0.02 || 0.03 || k=1.00, b=0.65 || Special ranking mechanism implemented ||
|| IE || 0.43 || 0.23 || 0.23 || k=1.5, b=0.7 || An attempt wars worth (thanks google translate!) prefer popular docs||
|| Raghu || 0.37 || 0.23 || 0.20 || k=1.75, b=0.75 || Tried multiplicative merging||
|| Sven || 0.43 || 0.34 || 0.29 || k=0.75, b=0.25 || prefer some of the popular documents ||
|| BIOS Bernard ||0.58 || 0.43 || 0.40 || k=1.2, b=0.0 || prefer popular movies ||
|| Evgeny+Numair ||0.57 || 0.44 || 0.40 || k=0.75, b=0.0 || prefer popular documents ||
|| Jan ||0.24 || 0.11 || 0.10 || k=1.2, b=0.75 || baseline ||
|| KC ||0.57 || 0.34 || 0.35 || k=1.25, b=0.2 || prefer popular records, added weight for keys occur in title ||
|| Elias+Maximilian || 0.5 || 0.31 || 0.3 || k=1.4, b=0.04 || prefer popular records, boost if multiple search terms hit ||
|| Matthias || 0.5 || 0.07 || 0.34 || k=1.23, b=0.53 || best neighbour optimizer + (far to small) training-set + popularity ranking ||
|| Daniel || 0.3 || 0.14 || 0.13 || k=1.75, b=0.4 || Only parameter adjustment ||