2010:Symbolic Melodic Similarity Results

Introduction

These are the results for the 2010 running of the Symbolic Melodic Similarity task set. For background information about this task set please refer to the 2010:Symbolic Melodic Similarity page.

Each system was given a query and returned the 10 most melodically similar songs from those taken from the Essen Collection (5274 pieces in the MIDI format; see ESAC Data Homepage for more information). For each query, we made four classes of error-mutations, thus the set comprises the following query classes:

• 0. No errors
• 1. One note deleted
• 2. One note inserted
• 3. One interval enlarged
• 4. One interval compressed

For each query (and its 4 mutations), the returned results (candidates) from all systems were then grouped together (query set) for evaluation by the human graders. The graders were provide with only heard perfect version against which to evaluate the candidates and did not know whether the candidates came from a perfect or mutated query. Each query/candidate set was evaluated by 1 individual grader. Using the Evalutron 6000 system, the graders gave each query/candidate pair two types of scores. Graders were asked to provide 1 categorical score with 3 categories: NS,SS,VS as explained below, and one fine score (in the range from 0 to 100).

Evalutron 6000 Summary Data

Number of evaluators = 6
Number of evaluations per query/candidate pair = 1
Number of queries per grader = 1
Total number of candidates returned = 3900
Total number of unique query/candidate pairs graded = 895
Average number of query/candidate pairs evaluated per grader: 149
Number of queries = 6 (perfect) with each perfect query error-mutated 4 different ways = 30

General Legend

Sub code	Submission name	Abstract	Contributors
HFRA1	SMS simbals	PDF	Pierre Hanna, Pascal Ferraro, Matthias Robine, Julien Allali
JU1	SMS-Domain	PDF	Julián Urbano, Juan Lloréns, Jorge Morato, Sonia Sánchez-Cuadrado
JU2	SMS-PitchDeriv	PDF	Julián Urbano, Juan Lloréns, Jorge Morato, Sonia Sánchez-Cuadrado
JU3	SMS-ParamDeriv	PDF	Julián Urbano, Juan Lloréns, Jorge Morato, Sonia Sánchez-Cuadrado
JU4	SMS-Shape	PDF	Julián Urbano, Juan Lloréns, Jorge Morato, Sonia Sánchez-Cuadrado
LL1	CbrahmsS2	PDF	Mika Laitinen, Kjell Lemström
LL2	CbrahmsW2	PDF	Mika Laitinen, Kjell Lemström
RI1	UAC	PDF	David Rizo, José Manuel Iñesta
RI2	UAT	PDF	David Rizo, José Manuel Iñesta
RI3	UAT3	PDF	David Rizo, José Manuel Iñesta
RI4	UAPR	PDF	David Rizo, José Manuel Iñesta
SU1	NGR5	PDF	Iman Suyoto, Alexandra Uitdenbogerd
SU2	PIOI	PDF	Iman Suyoto, Alexandra Uitdenbogerd

Broad Categories

NS = Not Similar
SS = Somewhat Similar
VS = Very Similar

Table Headings

ADR = Average Dynamic Recall
NRGB = Normalize Recall at Group Boundaries
AP = Average Precision (non-interpolated)
PND = Precision at N Documents

Calculating Summary Measures

Fine⁽¹⁾ = Sum of fine-grained human similarity decisions (0-100).
PSum⁽¹⁾ = Sum of human broad similarity decisions: NS=0, SS=1, VS=2.
WCsum⁽¹⁾ = 'World Cup' scoring: NS=0, SS=1, VS=3 (rewards Very Similar).
SDsum⁽¹⁾ = 'Stephen Downie' scoring: NS=0, SS=1, VS=4 (strongly rewards Very Similar).
Greater0⁽¹⁾ = NS=0, SS=1, VS=1 (binary relevance judgment).
Greater1⁽¹⁾ = NS=0, SS=0, VS=1 (binary relevance judgment using only Very Similar).

⁽¹⁾Normalized to the range 0 to 1.

Summary Results

Run Times

Participant	Runtime (min)	Machine
HFRA1	16	nema-c-1-2
JU1	7	win-0-1-1
JU2	15	win-0-1-1
JU3	21	win-0-1-1
JU4	20	win-0-1-1
LL1	205	compute-1-0
LL2	286	compute-1-0
RI1	13	nema-c-1-2
RI2	4	nema-c-1-2
RI3	6	nema-c-1-2
RI4	5	nema-c-1-2
SU1	26	compute-1-0
SU2	8	compute-1-0

download these results as csv

Overall Scores (Includes Perfect and Error Candidates)

Overall	HFRA	JU1	JU2	JU3	JU4	LL1	LL2	RI1	RI2	RI3	RI4	SU1	SU2
ADR	0.2472	0.3069	0.3091	0.3168	0.3713	0.3077	0.2105	0.2744	0.1078	0.2234	0.1384	0.2904	0.2487
NRGB	0.2405	0.2974	0.2941	0.2877	0.3278	0.2989	0.1730	0.2751	0.1167	0.2028	0.1581	0.2651	0.2361
AP	0.1914	0.2998	0.2987	0.3012	0.3492	0.2398	0.1816	0.2560	0.0660	0.1448	0.1281	0.2454	0.2265
PND	0.2829	0.3726	0.3726	0.3681	0.3985	0.3410	0.2226	0.3530	0.1390	0.2241	0.2062	0.3353	0.2839
Fine	0.5625	0.5793	0.5832	0.5806	0.6055	0.5799	0.4140	0.5453	0.3794	0.4705	0.5219	0.5356	0.5099
Psum	0.5783	0.6133	0.6200	0.6150	0.6417	0.6000	0.4017	0.5817	0.3583	0.4983	0.5583	0.5433	0.5317
WCsum	0.5122	0.5589	0.5633	0.5589	0.5800	0.5478	0.3489	0.5200	0.3022	0.4289	0.4789	0.5033	0.4844
SDsum	0.4792	0.5317	0.5350	0.5308	0.5492	0.5217	0.3225	0.4892	0.2742	0.3942	0.4392	0.4833	0.4608
Greater0	0.7767	0.7767	0.7900	0.7833	0.8267	0.7567	0.5600	0.7667	0.5267	0.7067	0.7967	0.6633	0.6733
Greater1	0.3800	0.4500	0.4500	0.4467	0.4567	0.4433	0.2433	0.3967	0.1900	0.2900	0.3200	0.4233	0.3900

download these results as csv

Scores by Query Error Types

No_Errors	HFRA	JU1	JU2	JU3	JU4	LL1	LL2	RI1	RI2	RI3	RI4	SU1	SU2
ADR	0.2576	0.3492	0.3492	0.3585	0.3946	0.3213	0.1688	0.2799	0.1536	0.3026	0.1346	0.2811	0.2822
NRGB	0.2420	0.3279	0.3279	0.3279	0.3445	0.2973	0.1428	0.2903	0.1640	0.2688	0.1615	0.2543	0.2388
AP	0.2122	0.3470	0.3512	0.3574	0.3898	0.2708	0.1823	0.2634	0.1107	0.2279	0.1461	0.2457	0.2612
PND	0.2750	0.4097	0.4097	0.4097	0.4264	0.3847	0.2250	0.3556	0.2014	0.2847	0.2000	0.3306	0.2750
Fine	0.5452	0.6293	0.6227	0.6155	0.6413	0.5958	0.4188	0.5450	0.4162	0.5172	0.5545	0.5755	0.5328
Psum	0.5667	0.6667	0.6583	0.6500	0.6750	0.6250	0.4083	0.5750	0.4167	0.5500	0.6000	0.5833	0.5583
WCsum	0.5000	0.6056	0.6000	0.5944	0.6111	0.5722	0.3611	0.5167	0.3667	0.4833	0.5167	0.5389	0.5111
SDsum	0.4667	0.5750	0.5708	0.5667	0.5792	0.5458	0.3375	0.4875	0.3417	0.4500	0.4750	0.5167	0.4875
Greater0	0.7667	0.8500	0.8333	0.8167	0.8667	0.7833	0.5500	0.7500	0.5667	0.7500	0.8500	0.7167	0.7000
Greater1	0.3667	0.4833	0.4833	0.4833	0.4833	0.4667	0.2667	0.4000	0.2667	0.3500	0.3500	0.4500	0.4167

download these results as csv

Deleted	HFRA	JU1	JU2	JU3	JU4	LL1	LL2	RI1	RI2	RI3	RI4	SU1	SU2
ADR	0.2440	0.3305	0.3305	0.3691	0.4068	0.2773	0.2174	0.2745	0.0329	0.1251	0.1396	0.3135	0.2400
NRGB	0.2423	0.3313	0.3313	0.3299	0.3577	0.2450	0.1855	0.2487	0.0567	0.1200	0.1443	0.2973	0.2362
AP	0.2339	0.3547	0.3547	0.3741	0.4036	0.2194	0.2305	0.2555	0.0095	0.0542	0.0805	0.2808	0.2177
PND	0.3241	0.4278	0.4278	0.4278	0.4444	0.2889	0.2370	0.3222	0.0500	0.1333	0.1852	0.3870	0.2704
Fine	0.5533	0.6165	0.6165	0.6198	0.6323	0.5487	0.4182	0.5242	0.2983	0.3795	0.4598	0.5328	0.5168
Psum	0.5667	0.6583	0.6583	0.6667	0.6750	0.5500	0.4167	0.5500	0.2333	0.3667	0.4833	0.5417	0.5167
WCsum	0.5111	0.6056	0.6056	0.6111	0.6167	0.5056	0.3611	0.5000	0.1833	0.3056	0.4167	0.5056	0.4722
SDsum	0.4833	0.5792	0.5792	0.5833	0.5875	0.4833	0.3333	0.4750	0.1583	0.2750	0.3833	0.4875	0.4500
Greater0	0.7333	0.8167	0.8167	0.8333	0.8500	0.6833	0.5833	0.7000	0.3833	0.5500	0.6833	0.6500	0.6500
Greater1	0.4000	0.5000	0.5000	0.5000	0.5000	0.4167	0.2500	0.4000	0.0833	0.1833	0.2833	0.4333	0.3833

download these results as csv

Inserted	HFRA	JU1	JU2	JU3	JU4	LL1	LL2	RI1	RI2	RI3	RI4	SU1	SU2
ADR	0.2324	0.2751	0.3156	0.3193	0.3555	0.3131	0.3225	0.2516	0.0488	0.1556	0.1487	0.2978	0.2253
NRGB	0.2253	0.2568	0.2635	0.2735	0.3119	0.3202	0.2693	0.2354	0.0530	0.1406	0.1598	0.2671	0.2303
AP	0.1719	0.2679	0.2899	0.2803	0.3217	0.2098	0.2216	0.2209	0.0218	0.0963	0.1139	0.2360	0.1886
PND	0.2656	0.3587	0.3754	0.3587	0.3921	0.2899	0.2751	0.3288	0.0685	0.1942	0.2146	0.3365	0.3198
Fine	0.5512	0.5880	0.6032	0.5932	0.6205	0.5545	0.4202	0.5315	0.3300	0.4363	0.5157	0.5088	0.5070
Psum	0.5500	0.6417	0.6583	0.6417	0.6750	0.5667	0.4083	0.5750	0.3083	0.4750	0.5500	0.5167	0.5500
WCsum	0.4889	0.5778	0.5944	0.5778	0.6056	0.5111	0.3556	0.5000	0.2389	0.4000	0.4667	0.4778	0.4944
SDsum	0.4583	0.5458	0.5625	0.5458	0.5708	0.4833	0.3292	0.4625	0.2042	0.3625	0.4250	0.4583	0.4667
Greater0	0.7333	0.8333	0.8500	0.8333	0.8833	0.7333	0.5667	0.8000	0.5167	0.7000	0.8000	0.6333	0.7167
Greater1	0.3667	0.4500	0.4667	0.4500	0.4667	0.4000	0.2500	0.3500	0.1000	0.2500	0.3000	0.4000	0.3833

download these results as csv

Enlarged	HFRA	JU1	JU2	JU3	JU4	LL1	LL2	RI1	RI2	RI3	RI4	SU1	SU2
ADR	0.2349	0.2665	0.2609	0.2352	0.3300	0.3171	0.1763	0.3018	0.1346	0.2593	0.1379	0.2691	0.2548
NRGB	0.2289	0.2804	0.2637	0.2350	0.2961	0.3428	0.1381	0.3171	0.1403	0.2394	0.1670	0.2514	0.2489
AP	0.1564	0.2257	0.2095	0.1907	0.2693	0.2563	0.1179	0.2733	0.0730	0.1450	0.1618	0.2106	0.2315
PND	0.2389	0.3111	0.2944	0.2889	0.3389	0.3889	0.1778	0.3833	0.1444	0.2444	0.2444	0.2944	0.2667
Fine	0.5760	0.5027	0.4973	0.5047	0.5455	0.6010	0.3932	0.5692	0.4278	0.5153	0.5333	0.5248	0.4858
Psum	0.6000	0.5167	0.5083	0.5250	0.5667	0.6417	0.3667	0.5917	0.4167	0.5833	0.5750	0.5250	0.5000
WCsum	0.5222	0.4722	0.4611	0.4722	0.5056	0.5944	0.3167	0.5333	0.3556	0.5056	0.4944	0.4833	0.4556
SDsum	0.4833	0.4500	0.4375	0.4458	0.4750	0.5708	0.2917	0.5042	0.3250	0.4667	0.4542	0.4625	0.4333
Greater0	0.8333	0.6500	0.6500	0.6833	0.7500	0.7833	0.5167	0.7667	0.6000	0.8167	0.8167	0.6500	0.6333
Greater1	0.3667	0.3833	0.3667	0.3667	0.3833	0.5000	0.2167	0.4167	0.2333	0.3500	0.3333	0.4000	0.3667

download these results as csv

Compressed	HFRA	JU1	JU2	JU3	JU4	LL1	LL2	RI1	RI2	RI3	RI4	SU1	SU2
ADR	0.2671	0.3132	0.2891	0.3017	0.3694	0.3096	0.1673	0.2643	0.1691	0.2745	0.1313	0.2903	0.2413
NRGB	0.2642	0.2908	0.2842	0.2723	0.3288	0.2891	0.1293	0.2838	0.1697	0.2451	0.1580	0.2552	0.2264
AP	0.1828	0.3036	0.2884	0.3036	0.3617	0.2430	0.1555	0.2670	0.1149	0.2009	0.1383	0.2540	0.2334
PND	0.3110	0.3555	0.3555	0.3555	0.3907	0.3527	0.1983	0.3751	0.2305	0.2638	0.1870	0.3282	0.2877
Fine	0.5868	0.5602	0.5762	0.5700	0.5877	0.5997	0.4197	0.5567	0.4247	0.5040	0.5462	0.5360	0.5068
Psum	0.6083	0.5833	0.6167	0.5917	0.6167	0.6167	0.4083	0.6167	0.4167	0.5167	0.5833	0.5500	0.5333
WCsum	0.5389	0.5333	0.5556	0.5389	0.5611	0.5556	0.3500	0.5500	0.3667	0.4500	0.5000	0.5111	0.4889
SDsum	0.5042	0.5083	0.5250	0.5125	0.5333	0.5250	0.3208	0.5167	0.3417	0.4167	0.4583	0.4917	0.4667
Greater0	0.8167	0.7333	0.8000	0.7500	0.7833	0.8000	0.5833	0.8167	0.5667	0.7167	0.8333	0.6667	0.6667
Greater1	0.4000	0.4333	0.4333	0.4333	0.4500	0.4333	0.2333	0.4167	0.2667	0.3167	0.3333	0.4333	0.4000

download these results as csv

Friedman Test with Multiple Comparisons Results (p=0.05)

The Friedman test was run in MATLAB against the Fine summary data over the 30 queries.
Command: [c,m,h,gnames] = multcompare(stats, 'ctype', 'tukey-kramer','estimate', 'friedman', 'alpha', 0.05);

TeamID	TeamID	Lowerbound	Mean	Upperbound	Significance
JU4	JU2	-1.6410	1.6667	4.9743	FALSE
JU4	JU3	-1.8077	1.5000	4.8077	FALSE
JU4	LL1	-1.9577	1.3500	4.6577	FALSE
JU4	JU1	-1.6743	1.6333	4.9410	FALSE
JU4	HFRA	-2.1243	1.1833	4.4910	FALSE
JU4	RI1	-1.8910	1.4167	4.7243	FALSE
JU4	SU1	-0.8077	2.5000	5.8077	FALSE
JU4	RI4	0.1590	3.4667	6.7743	TRUE
JU4	SU2	0.6423	3.9500	7.2577	TRUE
JU4	RI3	0.5757	3.8833	7.1910	TRUE
JU4	LL2	2.7757	6.0833	9.3910	TRUE
JU4	RI2	3.3757	6.6833	9.9910	TRUE
JU2	JU3	-3.4743	-0.1667	3.1410	FALSE
JU2	LL1	-3.6243	-0.3167	2.9910	FALSE
JU2	JU1	-3.3410	-0.0333	3.2743	FALSE
JU2	HFRA	-3.7910	-0.4833	2.8243	FALSE
JU2	RI1	-3.5577	-0.2500	3.0577	FALSE
JU2	SU1	-2.4743	0.8333	4.1410	FALSE
JU2	RI4	-1.5077	1.8000	5.1077	FALSE
JU2	SU2	-1.0243	2.2833	5.5910	FALSE
JU2	RI3	-1.0910	2.2167	5.5243	FALSE
JU2	LL2	1.1090	4.4167	7.7243	TRUE
JU2	RI2	1.7090	5.0167	8.3243	TRUE
JU3	LL1	-3.4577	-0.1500	3.1577	FALSE
JU3	JU1	-3.1743	0.1333	3.4410	FALSE
JU3	HFRA	-3.6243	-0.3167	2.9910	FALSE
JU3	RI1	-3.3910	-0.0833	3.2243	FALSE
JU3	SU1	-2.3077	1.0000	4.3077	FALSE
JU3	RI4	-1.3410	1.9667	5.2743	FALSE
JU3	SU2	-0.8577	2.4500	5.7577	FALSE
JU3	RI3	-0.9243	2.3833	5.6910	FALSE
JU3	LL2	1.2757	4.5833	7.8910	TRUE
JU3	RI2	1.8757	5.1833	8.4910	TRUE
LL1	JU1	-3.0243	0.2833	3.5910	FALSE
LL1	HFRA	-3.4743	-0.1667	3.1410	FALSE
LL1	RI1	-3.2410	0.0667	3.3743	FALSE
LL1	SU1	-2.1577	1.1500	4.4577	FALSE
LL1	RI4	-1.1910	2.1167	5.4243	FALSE
LL1	SU2	-0.7077	2.6000	5.9077	FALSE
LL1	RI3	-0.7743	2.5333	5.8410	FALSE
LL1	LL2	1.4257	4.7333	8.0410	TRUE
LL1	RI2	2.0257	5.3333	8.6410	TRUE
JU1	HFRA	-3.7577	-0.4500	2.8577	FALSE
JU1	RI1	-3.5243	-0.2167	3.0910	FALSE
JU1	SU1	-2.4410	0.8667	4.1743	FALSE
JU1	RI4	-1.4743	1.8333	5.1410	FALSE
JU1	SU2	-0.9910	2.3167	5.6243	FALSE
JU1	RI3	-1.0577	2.2500	5.5577	FALSE
JU1	LL2	1.1423	4.4500	7.7577	TRUE
JU1	RI2	1.7423	5.0500	8.3577	TRUE
HFRA	RI1	-3.0743	0.2333	3.5410	FALSE
HFRA	SU1	-1.9910	1.3167	4.6243	FALSE
HFRA	RI4	-1.0243	2.2833	5.5910	FALSE
HFRA	SU2	-0.5410	2.7667	6.0743	FALSE
HFRA	RI3	-0.6077	2.7000	6.0077	FALSE
HFRA	LL2	1.5923	4.9000	8.2077	TRUE
HFRA	RI2	2.1923	5.5000	8.8077	TRUE
RI1	SU1	-2.2243	1.0833	4.3910	FALSE
RI1	RI4	-1.2577	2.0500	5.3577	FALSE
RI1	SU2	-0.7743	2.5333	5.8410	FALSE
RI1	RI3	-0.8410	2.4667	5.7743	FALSE
RI1	LL2	1.3590	4.6667	7.9743	TRUE
RI1	RI2	1.9590	5.2667	8.5743	TRUE
SU1	RI4	-2.3410	0.9667	4.2743	FALSE
SU1	SU2	-1.8577	1.4500	4.7577	FALSE
SU1	RI3	-1.9243	1.3833	4.6910	FALSE
SU1	LL2	0.2757	3.5833	6.8910	TRUE
SU1	RI2	0.8757	4.1833	7.4910	TRUE
RI4	SU2	-2.8243	0.4833	3.7910	FALSE
RI4	RI3	-2.8910	0.4167	3.7243	FALSE
RI4	LL2	-0.6910	2.6167	5.9243	FALSE
RI4	RI2	-0.0910	3.2167	6.5243	FALSE
SU2	RI3	-3.3743	-0.0667	3.2410	FALSE
SU2	LL2	-1.1743	2.1333	5.4410	FALSE
SU2	RI2	-0.5743	2.7333	6.0410	FALSE
RI3	LL2	-1.1077	2.2000	5.5077	FALSE
RI3	RI2	-0.5077	2.8000	6.1077	FALSE
LL2	RI2	-2.7077	0.6000	3.9077	FALSE

download these results as csv

Raw Scores

The raw data derived from the Evalutron 6000 human evaluations are located on the 2010:Symbolic Melodic Similarity Raw Data page.