#634: Scoring overhaul

[ericz1803]

biothings/biothings_explorer#634

• Update scoring system to use below formula: https://docs.google.com/presentation/d/1DdsKaQx88ppCM0lNWVKZsk4EwegyAtWW6h7aC-23vXA/edit?usp=sharing
  

• Update logs:

  •      "timestamp": "2023-08-02T20:40:04.476Z",
         "level": "DEBUG",
         "message": "[Template-3]: Scored 320 results with NGD score, scored 258 results without NGD",
         "code": null
     },``` 
    

  • deleted scoring summary

Benchmarks comparison of old scoring system vs new scoring system:

Old:

Benchmark: treats
                        k=1     k=5     k=10    k=20
Precision @ k           0.0000  0.0273  0.0318  0.0295
Recall @ k              0.0000  0.0714  0.1667  0.3095
mAP @ k                 0.0000  0.0273  0.0535  0.0597
Top-k Accuracy          0.0000  0.0714  0.1667  0.3095

Mean Reciprocal Rank    0.06169455544455544

Benchmark: DrugCentral_creative
                        k=1     k=5     k=10    k=20
Precision @ k           0.2600  0.0680  0.0360  0.0200
Recall @ k              0.2600  0.3400  0.3600  0.4000
mAP @ k                 0.2600  0.2867  0.2900  0.2930
Top-k Accuracy          0.2600  0.3400  0.3600  0.4000

Mean Reciprocal Rank    0.2929946524064171

New:

Benchmark: treats
                        k=1     k=5     k=10    k=20    k=50    k=100   k=200   k=500   k=1000
Precision @ k           0.2273  0.1364  0.0864  0.0523  0.0264  0.0145  0.0075  0.0031  0.0015
Recall @ k              0.1190  0.3571  0.4524  0.5476  0.6905  0.7619  0.7857  0.8095  0.8095
mAP @ k                 0.2273  0.2449  0.2550  0.2631  0.2707  0.2722  0.2726  0.2728  0.2728
Top-k Accuracy          0.1190  0.3571  0.4524  0.5476  0.6905  0.7619  0.7857  0.8095  0.8095

Mean Reciprocal Rank    0.314370593945062

Benchmark: DrugCentral_creative
                        k=1     k=5     k=10    k=20    k=50    k=100   k=200   k=500   k=1000
Precision @ k           0.4200  0.1120  0.0640  0.0370  0.0164  0.0088  0.0045  0.0019  0.0009
Recall @ k              0.4200  0.5600  0.6400  0.7400  0.8200  0.8800  0.9000  0.9400  0.9400
mAP @ k                 0.4200  0.4767  0.4890  0.4959  0.4986  0.4995  0.4996  0.4997  0.4997
Top-k Accuracy          0.4200  0.5600  0.6400  0.7400  0.8200  0.8800  0.9000  0.9400  0.9400

Mean Reciprocal Rank    0.4997344748094153