2014:Symbolic Melodic Similarity Results

From MIREX Wiki
Revision as of 20:25, 17 October 2014 by Kahyun Choi (talk | contribs) (General Legend)

Introduction

These are the results for the 2014 running of the Symbolic Melodic Similarity task set. For background information about this task set please refer to the 2014: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 unique query/candidate pairs graded = 436
Average number of query/candidate pairs evaluated per grader: 73
Number of queries = 6 (perfect) with each perfect query error-mutated 4 different ways = 30

General Legend

Sub code Submission name Abstract Contributors
JU1 ShapeH PDF Julián Urbano
JU2 ShapeTime PDF Julián Urbano
JU3 Time PDF Julián Urbano
YO1 YOkuboSMS PDF Yoshiaki OKUBO

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

Overall Scores (Includes Perfect and Error Candidates)

SCORE JU1 JU2 JU3 YO1
ADR 0.7089 0.7962 0.7997 0.6912
NRGB 0.6786 0.7493 0.7602 0.6378
AP 0.7344 0.7534 0.7992 0.5535
PND 0.7361 0.7444 0.7611 0.5611
Fine 53.7767 54.5967 51.1933 36.9633
PSum 1.1167 1.13 1.1267 0.69
WCSum 1.5033 1.5133 1.5433 0.93667
SDSum 1.89 1.8967 1.96 1.1833
Greater0 0.73 0.74667 0.71 0.44333
Greater1 0.38667 0.38333 0.41667 0.24667

download these results as csv

Scores by Query Error Types

No Errors

SCORE JU1 JU2 JU3 YO1
ADR 0.7089 0.7962 0.7997 0.6912
NRGB 0.6786 0.7493 0.7602 0.6378
AP 0.7344 0.7534 0.7992 0.5535
PND 0.7361 0.7444 0.7611 0.5611
Fine 58.2333 59.0667 53.3 38.3
PSum 1.2167 1.2333 1.1833 0.71667
WCSum 1.6167 1.6333 1.6167 0.96667
SDSum 2.0167 2.0333 2.05 1.2167
Greater0 0.81667 0.83333 0.75 0.46667
Greater1 0.4 0.4 0.43333 0.25

download these results as csv

Note Deletions

SCORE JU1 JU2 JU3 YO1
ADR 0.6419 0.8038 0.7728 0.5673
NRGB 0.6268 0.7623 0.7349 0.5436
AP 0.6510 0.7414 0.7364 0.4675
PND 0.6125 0.7139 0.6889 0.4694
Fine 57.1 58.2667 53.65 34.6167
PSum 1.2167 1.2333 1.1833 0.63333
WCSum 1.65 1.6667 1.6333 0.86667
SDSum 2.0833 2.1 2.0833 1.1
Greater0 0.78333 0.8 0.73333 0.4
Greater1 0.43333 0.43333 0.45 0.23333

download these results as csv

Note Insertions

SCORE JU1 JU2 JU3 YO1
ADR 0.6550 0.7943 0.8162 0.5889
NRGB 0.6326 0.7343 0.8015 0.5854
AP 0.6940 0.7282 0.7750 0.5167
PND 0.7361 0.6611 0.7361 0.5444
Fine 55.85 57.4667 53.05 37.35
PSum 1.15 1.1833 1.15 0.68333
WCSum 1.5333 1.5667 1.5833 0.93333
SDSum 1.9167 1.95 2.0167 1.1833
Greater0 0.76667 0.8 0.71667 0.43333
Greater1 0.38333 0.38333 0.43333 0.25

download these results as csv

Enlarged Intervals

SCORE JU1 JU2 JU3 YO1
ADR 0.6807 0.7895 0.7727 0.6957
NRGB 0.6378 0.7304 0.7135 0.6360
AP 0.6694 0.7299 0.7296 0.5757
PND 0.6278 0.7389 0.7000 0.5917
Fine 45.4667 45.4167 47.0167 33.9833
PSum 0.91667 0.91667 1.0167 0.61667
WCSum 1.25 1.25 1.3667 0.83333
SDSum 1.5833 1.5833 1.7167 1.05
Greater0 0.58333 0.58333 0.66667 0.4
Greater1 0.33333 0.33333 0.35 0.21667

download these results as csv

Compressed Intervals

SCORE JU1 JU2 JU3 YO1
ADR 0.6125 0.7076 0.7076 0.5701
NRGB 0.5579 0.6655 0.6655 0.5486
AP 0.6407 0.6711 0.6247 0.4728
PND 0.5833 0.6250 0.6250 0.4306
Fine 52.2333 52.7667 48.95 40.5667
PSum 1.0833 1.0833 1.1 0.8
WCSum 1.4667 1.45 1.5167 1.0833
SDSum 1.85 1.8167 1.9333 1.3667
Greater0 0.7 0.71667 0.68333 0.51667
Greater1 0.38333 0.36667 0.41667 0.28333

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);

Row Labels JU1 JU2 JU3 YO1
q01 63.6 63.6 53.7 37.3
q01_1 63.6 63.6 53.7 27.2
q01_2 47.6 49.5 47.1 29.1
q01_3 25.2 24.9 32.8 30.5
q01_4 37.7 38.4 23.2 29.1
q02 72.5 72.5 72 56.5
q02_1 73.5 72.5 74 56
q02_2 70.5 70.5 70.5 60.5
q02_3 70 70 68.5 56.5
q02_4 72.5 70 72.5 56
q03 49 49 31.5 31.8
q03_1 45 45 31.8 31.8
q03_2 49 49 38.5 35.8
q03_3 49 49 34.5 31.8
q03_4 49 49 31.5 32
q04 67.7 67.7 64.8 29.2
q04_1 62.9 62.9 69.1 29.2
q04_2 70.9 73.7 68.9 33.2
q04_3 64.5 64.5 60 21.6
q04_4 62.6 62.6 69.7 47.3
q05 63 68 63 43.5
q05_1 64 72 58.5 32
q05_2 63 68 58.5 34
q05_3 30.5 30.5 50 34
q05_4 63 68 63 47.5
q06 33.6 33.6 34.8 31.5
q06_1 33.6 33.6 34.8 31.5
q06_2 34.1 34.1 34.8 31.5
q06_3 33.6 33.6 36.3 29.5
q06_4 28.6 28.6 33.8 31.5

download these results as csv

TeamID TeamID Lowerbound Mean Upperbound Significance
JU2 JU1 -0.6669 0.1500 0.9669 FALSE
JU2 JU3 -0.4002 0.4167 1.2336 FALSE
JU2 YO1 0.9498 1.7667 2.5836 TRUE
JU1 JU3 -0.5502 0.2667 1.0836 FALSE
JU1 YO1 0.7998 1.6167 2.4336 TRUE
JU3 YO1 0.5331 1.3500 2.1669 TRUE

download these results as csv

2014 sms fine scores friedmans.png