cand. gen. with stemmed emb. + s-gram cos. (data)

config

@@ -14,7 +14,8 @@ prediction_fn = ${rootpath}/runs/predictions/${timestamp}.tsv
 detailed_fn = ${rootpath}/runs/detailed/${timestamp}.{}.tsv
 
 [candidates]
-generator = PhraseVecFixedSet(10, "mean", "emb_stem")
+generator = SGramCosine(.5, 10)
+	PhraseVecFixedSet(10, "mean", "emb_stem")
 oracle = none
 workers = 0
 

log

@@ -1,46 +1,33 @@
-2018-07-31 07:38:07,635 - 'pattern' package not found; tag filters are not available for English
-2018-07-31 07:38:07,641 - loading terminology...
-2018-07-31 07:38:07,856 - loading pretrained embeddings...
-2018-07-31 07:38:07,857 - loading projection weights from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
-2018-07-31 07:38:53,703 - loaded (2231686, 200) matrix from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
-2018-07-31 07:38:57,688 - loading vectorizer...
-2018-07-31 07:38:57,688 - loading candidate generator...
-2018-07-31 07:38:57,689 - loading pretrained embeddings...
-2018-07-31 07:38:57,689 - loading projection weights from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
-2018-07-31 07:39:43,628 - loaded (2231686, 200) matrix from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
-2018-07-31 07:41:06,475 - loading vectorizer...
-2018-07-31 07:41:17,449 - preprocessing validation data...
-2018-07-31 07:41:17,449 - loading corpus...
-2018-07-31 07:41:17,459 - generating candidates with 0 workers...
-2018-07-31 07:41:19,905 - generated 3680 pair-wise samples (7870 with duplicates)
-2018-07-31 07:41:19,907 - compiling model architecture...
-2018-07-31 07:41:31,438 - preprocessing training data...
-2018-07-31 07:41:31,438 - loading corpus...
-2018-07-31 07:41:31,486 - generating candidates with 0 workers...
-2018-07-31 07:41:43,159 - generated 17100 pair-wise samples (51450 with duplicates)
-2018-07-31 07:41:43,165 - training CNN...
-2018-07-31 07:42:06,823 - Ranking accuracy: 0.632783
-2018-07-31 07:42:33,161 - Ranking accuracy: 0.672173
-2018-07-31 07:42:59,333 - Ranking accuracy: 0.695044
-2018-07-31 07:43:25,963 - Ranking accuracy: 0.70521
-2018-07-31 07:43:52,621 - Ranking accuracy: 0.700127
-2018-07-31 07:44:15,388 - Ranking accuracy: 0.736976
-2018-07-31 07:44:41,879 - Ranking accuracy: 0.7446
-2018-07-31 07:45:09,455 - Ranking accuracy: 0.750953
-2018-07-31 07:45:36,947 - Ranking accuracy: 0.747141
-2018-07-31 07:46:00,250 - Ranking accuracy: 0.756036
-2018-07-31 07:46:26,797 - Ranking accuracy: 0.759848
-2018-07-31 07:46:53,455 - Ranking accuracy: 0.767471
-2018-07-31 07:47:19,997 - Ranking accuracy: 0.770013
-2018-07-31 07:47:46,816 - Ranking accuracy: 0.775095
-2018-07-31 07:48:13,728 - Ranking accuracy: 0.775095
-2018-07-31 07:48:36,227 - Ranking accuracy: 0.776366
-2018-07-31 07:49:02,801 - Ranking accuracy: 0.782719
-2018-07-31 07:49:29,332 - Ranking accuracy: 0.778907
-2018-07-31 07:49:52,294 - Ranking accuracy: 0.778907
-2018-07-31 07:49:52,295 - Epoch 00019: early stopping
-2018-07-31 07:49:52,295 - done training.
-2018-07-31 07:49:52,301 - load best model...
-2018-07-31 07:50:04,107 - predict scores for validation data...
-2018-07-31 07:50:05,411 - evaluate and/or serialize...
-2018-07-31 07:50:05,439 - done.
+2018-07-31 08:07:42,706 - 'pattern' package not found; tag filters are not available for English
+2018-07-31 08:07:42,713 - loading terminology...
+2018-07-31 08:07:42,922 - loading pretrained embeddings...
+2018-07-31 08:07:42,922 - loading projection weights from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
+2018-07-31 08:08:28,042 - loaded (2231686, 200) matrix from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
+2018-07-31 08:08:32,130 - loading vectorizer...
+2018-07-31 08:08:32,131 - loading candidate generator...
+2018-07-31 08:08:43,654 - loading pretrained embeddings...
+2018-07-31 08:08:43,654 - loading projection weights from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
+2018-07-31 08:09:29,441 - loaded (2231686, 200) matrix from /mnt/storage/karr/users/furrer/prlnk/data/embeddings/wvec_200_win-30_chiu-et-al.bin
+2018-07-31 08:10:53,552 - loading vectorizer...
+2018-07-31 08:11:04,731 - preprocessing validation data...
+2018-07-31 08:11:04,731 - loading corpus...
+2018-07-31 08:11:04,741 - generating candidates with 0 workers...
+2018-07-31 08:11:09,634 - generated 5578 pair-wise samples (11853 with duplicates)
+2018-07-31 08:11:09,636 - compiling model architecture...
+2018-07-31 08:11:21,755 - preprocessing training data...
+2018-07-31 08:11:21,755 - loading corpus...
+2018-07-31 08:11:21,803 - generating candidates with 0 workers...
+2018-07-31 08:11:44,819 - generated 26131 pair-wise samples (77122 with duplicates)
+2018-07-31 08:11:44,828 - training CNN...
+2018-07-31 08:12:20,410 - Ranking accuracy: 0.757306
+2018-07-31 08:12:59,131 - Ranking accuracy: 0.776366
+2018-07-31 08:13:37,762 - Ranking accuracy: 0.767471
+2018-07-31 08:14:12,247 - Ranking accuracy: 0.778907
+2018-07-31 08:14:50,669 - Ranking accuracy: 0.776366
+2018-07-31 08:15:26,574 - Ranking accuracy: 0.777637
+2018-07-31 08:15:26,574 - Epoch 00006: early stopping
+2018-07-31 08:15:26,575 - done training.
+2018-07-31 08:15:26,582 - load best model...
+2018-07-31 08:15:40,322 - predict scores for validation data...
+2018-07-31 08:15:42,189 - evaluate and/or serialize...
+2018-07-31 08:15:42,224 - done.

results

@@ -1,6 +1,6 @@
-accuracy     0.7827191867852605
-correct        616
+accuracy     0.7789072426937739
+correct        613
 total          787
-unreachable    133
+unreachable    104
 nocandidates     0
 ambiguous        2