Rostlab / nala

Text mining of natural language mutations mentions

Home Page:https://www.tagtog.net/-corpora/IDP4+

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Number replaced WE

abojchevski opened this issue · comments

With WE replaced numbers nala_training_51

P:0.8162    R:0.7947    F:0.8053    0   exact
P:0.4323    R:0.3879    F:0.4089    1   exact
P:0.4364    R:0.3810    F:0.4068    2   exact
P:0.6718    R:0.6322    F:0.6514    TOTAL   exact

P:0.9219    R:0.8998    F:0.9107    0   overlapping
P:0.9338    R:0.8328    F:0.8804    1   overlapping
P:0.8831    R:0.8000    F:0.8395    2   overlapping
P:0.9222    R:0.8657    F:0.8931    TOTAL   overlapping

vs the baseline nala_training_51

P:0.8157    R:0.7924    F:0.8039    0   exact
P:0.4105    R:0.3645    F:0.3861    1   exact
P:0.4727    R:0.4127    F:0.4407    2   exact
P:0.6687    R:0.6264    F:0.6469    TOTAL   exact

P:0.9267    R:0.9046    F:0.9155    0   overlapping
P:0.9380    R:0.8317    F:0.8816    1   overlapping
P:0.8800    R:0.7952    F:0.8354    2   overlapping
P:0.9262    R:0.8675    F:0.8959    TOTAL   overlapping

OLD WE

fold 0

tp:143  fp:79   fn:49   fpo:33  fno:35  P:0.6441        R:0.7448        F:0.6908        0       exact
tp:30   fp:80   fn:65   fpo:56  fno:49  P:0.2727        R:0.3158        F:0.2927        1       exact
tp:11   fp:15   fn:20   fpo:11  fno:12  P:0.4231        R:0.3548        F:0.3860        2       exact
tp:184  fp:174  fn:134  fpo:100 fno:96  P:0.5140        R:0.5786        F:0.5444        TOTAL   exact

tp:143  fp:79   fn:49   fpo:33  fno:35  P:0.8210        R:0.9378        F:0.8755        0       overlapping
tp:30   fp:80   fn:65   fpo:56  fno:49  P:0.8491        R:0.8940        F:0.8710        1       overlapping
tp:11   fp:15   fn:20   fpo:11  fno:12  P:0.8947        R:0.8095        F:0.8500        2       overlapping
tp:184  fp:174  fn:134  fpo:100 fno:96  P:0.8370        R:0.9091        F:0.8716        TOTAL   overlapping

fold 1

tp:100  fp:52   fn:26   fpo:14  fno:14  P:0.6579        R:0.7937        F:0.7194        0       exact
tp:32   fp:106  fn:80   fpo:80  fno:65  P:0.2319        R:0.2857        F:0.2560        1       exact
tp:17   fp:21   fn:19   fpo:13  fno:13  P:0.4474        R:0.4722        F:0.4595        2       exact
tp:149  fp:179  fn:125  fpo:107 fno:92  P:0.4543        R:0.5438        F:0.4950        TOTAL   exact

tp:100  fp:52   fn:26   fpo:14  fno:14  P:0.7711        R:0.9143        F:0.8366        0       overlapping
tp:32   fp:106  fn:80   fpo:80  fno:65  P:0.8719        R:0.9219        F:0.8962        1       overlapping
tp:17   fp:21   fn:19   fpo:13  fno:13  P:0.8431        R:0.8776        F:0.8600        2       overlapping
tp:149  fp:179  fn:125  fpo:107 fno:92  P:0.8286        R:0.9134        F:0.8689        TOTAL   overlapping

fold 2

tp:124  fp:61   fn:39   fpo:21  fno:18  P:0.6703        R:0.7607        F:0.7126        0       exact
tp:24   fp:61   fn:78   fpo:47  fno:45  P:0.2824        R:0.2353        F:0.2567        1       exact
tp:7    fp:14   fn:17   fpo:10  fno:11  P:0.3333        R:0.2917        F:0.3111        2       exact
tp:155  fp:136  fn:134  fpo:78  fno:74  P:0.5326        R:0.5363        F:0.5345        TOTAL   exact

tp:124  fp:61   fn:39   fpo:21  fno:18  P:0.8030        R:0.8859        F:0.8424        0       overlapping
tp:24   fp:61   fn:78   fpo:47  fno:45  P:0.8923        R:0.7785        F:0.8315        1       overlapping
tp:7    fp:14   fn:17   fpo:10  fno:11  P:0.8750        R:0.8235        F:0.8485        2       overlapping
tp:155  fp:136  fn:134  fpo:78  fno:74  P:0.8411        R:0.8365        F:0.8388        TOTAL   overlapping

fold 3

tp:236  fp:80   fn:53   fpo:34  fno:34  P:0.7468        R:0.8166        F:0.7802        0       exact
tp:49   fp:95   fn:88   fpo:75  fno:64  P:0.3403        R:0.3577        F:0.3488        1       exact
tp:7    fp:18   fn:15   fpo:7   fno:8   P:0.2800        R:0.3182        F:0.2979        2       exact
tp:292  fp:193  fn:156  fpo:116 fno:106 P:0.6021        R:0.6518        F:0.6259        TOTAL   exact

tp:236  fp:80   fn:53   fpo:34  fno:34  P:0.8686        R:0.9412        F:0.9034        0       overlapping
tp:49   fp:95   fn:88   fpo:75  fno:64  P:0.9038        R:0.8868        F:0.8952        1       overlapping
tp:7    fp:18   fn:15   fpo:7   fno:8   P:0.6667        R:0.7586        F:0.7097        2       overlapping
tp:292  fp:193  fn:156  fpo:116 fno:106 P:0.8697        R:0.9113        F:0.8900        TOTAL   overlapping

fold 4

tp:185  fp:84   fn:58   fpo:34  fno:33  P:0.6877        R:0.7613        F:0.7227        0       exact
tp:57   fp:106  fn:92   fpo:88  fno:69  P:0.3497        R:0.3826        F:0.3654        1       exact
tp:19   fp:34   fn:31   fpo:22  fno:25  P:0.3585        R:0.3800        F:0.3689        2       exact
tp:261  fp:224  fn:181  fpo:144 fno:127 P:0.5381        R:0.5905        F:0.5631        TOTAL   exact

tp:185  fp:84   fn:58   fpo:34  fno:33  P:0.8344        R:0.9097        F:0.8705        0       overlapping
tp:57   fp:106  fn:92   fpo:88  fno:69  P:0.9224        R:0.9030        F:0.9126        1       overlapping
tp:19   fp:34   fn:31   fpo:22  fno:25  P:0.8462        R:0.9167        F:0.8800        2       overlapping
tp:261  fp:224  fn:181  fpo:144 fno:127 P:0.8693        R:0.9078        F:0.8881        TOTAL   overlapping

number replaced

fold 0

tp:143  fp:79   fn:48   fpo:34  fno:36  P:0.6441        R:0.7487        F:0.6925        0       exact
tp:28   fp:78   fn:67   fpo:56  fno:51  P:0.2642        R:0.2947        F:0.2786        1       exact
tp:14   fp:12   fn:17   fpo:8   fno:9   P:0.5385        R:0.4516        F:0.4912        2       exact
tp:185  fp:169  fn:132  fpo:98  fno:96  P:0.5226        R:0.5836        F:0.5514        TOTAL   exact

tp:143  fp:79   fn:48   fpo:34  fno:36  P:0.8256        R:0.9467        F:0.8820        0       overlapping
tp:28   fp:78   fn:67   fpo:56  fno:51  P:0.8599        R:0.8940        F:0.8766        1       overlapping
tp:14   fp:12   fn:17   fpo:8   fno:9   P:0.8857        R:0.7949        F:0.8378        2       overlapping
tp:185  fp:169  fn:132  fpo:98  fno:96  P:0.8422        R:0.9133        F:0.8763        TOTAL   overlapping

fold 1

tp:101  fp:46   fn:25   fpo:13  fno:14  P:0.6871        R:0.8016        F:0.7399        0       exact
tp:34   fp:103  fn:78   fpo:79  fno:65  P:0.2482        R:0.3036        F:0.2731        1       exact
tp:19   fp:25   fn:17   fpo:12  fno:12  P:0.4318        R:0.5278        F:0.4750        2       exact
tp:154  fp:174  fn:120  fpo:104 fno:91  P:0.4695        R:0.5620        F:0.5116        TOTAL   exact

tp:101  fp:46   fn:25   fpo:13  fno:14  P:0.7950        R:0.9209        F:0.8533        0       overlapping
tp:34   fp:103  fn:78   fpo:79  fno:65  P:0.8812        R:0.9319        F:0.9059        1       overlapping
tp:19   fp:25   fn:17   fpo:12  fno:12  P:0.7679        R:0.8958        F:0.8269        2       overlapping
tp:154  fp:174  fn:120  fpo:104 fno:91  P:0.8329        R:0.9233        F:0.8758        TOTAL   overlapping

fold 2

tp:124  fp:55   fn:39   fpo:20  fno:17  P:0.6927        R:0.7607        F:0.7251        0       exact
tp:25   fp:57   fn:77   fpo:45  fno:42  P:0.3049        R:0.2451        F:0.2717        1       exact
tp:7    fp:14   fn:17   fpo:9   fno:10  P:0.3333        R:0.2917        F:0.3111        2       exact
tp:156  fp:126  fn:133  fpo:74  fno:69  P:0.5532        R:0.5398        F:0.5464        TOTAL   exact

tp:124  fp:55   fn:39   fpo:20  fno:17  P:0.8214        R:0.8798        F:0.8496        0       overlapping
tp:25   fp:57   fn:77   fpo:45  fno:42  P:0.9032        R:0.7619        F:0.8266        1       overlapping
tp:7    fp:14   fn:17   fpo:9   fno:10  P:0.8387        R:0.7879        F:0.8125        2       overlapping
tp:156  fp:126  fn:133  fpo:74  fno:69  P:0.8519        R:0.8237        F:0.8375        TOTAL   overlapping

fold 3

tp:234  fp:85   fn:55   fpo:38  fno:37  P:0.7335        R:0.8097        F:0.7697        0       exact
tp:49   fp:97   fn:88   fpo:75  fno:64  P:0.3356        R:0.3577        F:0.3463        1       exact
tp:8    fp:17   fn:14   fpo:7   fno:8   P:0.3200        R:0.3636        F:0.3404        2       exact
tp:291  fp:199  fn:157  fpo:120 fno:109 P:0.5939        R:0.6496        F:0.6205        TOTAL   exact

tp:234  fp:85   fn:55   fpo:38  fno:37  P:0.8680        R:0.9450        F:0.9048        0       overlapping
tp:49   fp:97   fn:88   fpo:75  fno:64  P:0.8952        R:0.8868        F:0.8910        1       overlapping
tp:8    fp:17   fn:14   fpo:7   fno:8   P:0.6970        R:0.7931        F:0.7419        2       overlapping
tp:291  fp:199  fn:157  fpo:120 fno:109 P:0.8681        R:0.9155        F:0.8912        TOTAL   overlapping

fold 4

tp:187  fp:80   fn:56   fpo:32  fno:32  P:0.7004        R:0.7695        F:0.7333        0       exact
tp:56   fp:107  fn:93   fpo:92  fno:74  P:0.3436        R:0.3758        F:0.3590        1       exact
tp:21   fp:32   fn:29   fpo:20  fno:23  P:0.3962        R:0.4200        F:0.4078        2       exact
tp:264  fp:219  fn:178  fpo:144 fno:129 P:0.5466        R:0.5973        F:0.5708        TOTAL   exact

tp:187  fp:80   fn:56   fpo:32  fno:32  P:0.8395        R:0.9127        F:0.8746        0       overlapping
tp:56   fp:107  fn:93   fpo:92  fno:74  P:0.9367        R:0.9212        F:0.9289        1       overlapping
tp:21   fp:32   fn:29   fpo:20  fno:23  P:0.8421        R:0.9143        F:0.8767        2       overlapping
tp:264  fp:219  fn:178  fpo:144 fno:129 P:0.8775        R:0.9164        F:0.8965        TOTAL   overlapping

NO REPLACEMENT:

tp:788  fp:356  fn:225  fpo:136 fno:134 P:0.6888    R:0.7779    F:0.7306    0   exact
tp:192  fp:448  fn:403  fpo:346 fno:292 P:0.3000    R:0.3227    F:0.3109    1   exact
tp:61   fp:102  fn:102  fpo:63  fno:69  P:0.3742    R:0.3742    F:0.3742    2   exact
tp:1041 fp:906  fn:730  fpo:545 fno:495 P:0.5347    R:0.5878    F:0.5600    TOTAL   exact
tp:788  fp:356  fn:225  fpo:136 fno:134 P:0.8279    R:0.9208    F:0.8719    0   overlapping
tp:192  fp:448  fn:403  fpo:346 fno:292 P:0.8906    R:0.8820    F:0.8863    1   overlapping
tp:61   fp:102  fn:102  fpo:63  fno:69  P:0.8319    R:0.8540    F:0.8428    2   overlapping
tp:1041 fp:906  fn:730  fpo:545 fno:495 P:0.8522    R:0.8985    F:0.8747    TOTAL   overlapping

YES REPLACEMENT:

tp:789  fp:345  fn:223  fpo:137 fno:136 P:0.6958    R:0.7796    F:0.7353    0   exact
tp:192  fp:442  fn:403  fpo:347 fno:296 P:0.3028    R:0.3227    F:0.3124    1   exact
tp:69   fp:100  fn:94   fpo:56  fno:62  P:0.4083    R:0.4233    F:0.4157    2   exact
tp:1050 fp:887  fn:720  fpo:540 fno:494 P:0.5421    R:0.5932    F:0.5665    TOTAL   exact
tp:789  fp:345  fn:223  fpo:137 fno:136 P:0.8362    R:0.9243    F:0.8780    0   overlapping
tp:192  fp:442  fn:403  fpo:347 fno:296 P:0.8978    R:0.8864    F:0.8921    1   overlapping
tp:69   fp:100  fn:94   fpo:56  fno:62  P:0.8095    R:0.8539    F:0.8311    2   overlapping
tp:1050 fp:887  fn:720  fpo:540 fno:494 P:0.8573    R:0.9022    F:0.8791    TOTAL   overlapping

Both NL & ST classes improve. We choose with replacement