Merge remote-tracking branch 'origin/master'

AMR-Visualization-master/AMR-Visualization-master/amr-slide-simple.py

@@ -123,4 +123,4 @@ def load_amr(fpath):
 
 
 if __name__ == "__main__":
-    load_amr(os.path.join("C:\\Users\\Wesenberg\\PycharmProjects\\rl-sentence-compression-wesenberg\\AMR-Visualization-master\\AMR-Visualization-master\\source.txt"))
+    load_amr('test_single.txt')

AMR-Visualization-master/AMR-Visualization-master/source.txt

-(o / open-01
-      :ARG0 (p / person
-            :name (n / name
-                  :op1 "Fernando"
-                  :op2 "Henrique"
-                  :op3 "Cardoso")
-            :ARG0-of (s / scandalize-01
-                  :ARG2 (p2 / peddle-01
-                        :ARG1 (ii / influence-01))
-                  :ARG1-of (m / mushroom-02)))
-      :ARG1 (p3 / probe-01))
\ No newline at end of file
+(d / describe-01
+      :ARG0 (h / he)
+      :ARG1 (s / she)
+      :ARG2 (g / genius))
\ No newline at end of file

AMR-Visualization-master/AMR-Visualization-master/test.txt

-# ::snt export of high-tech products has frequently been in the spotlight , making a significant contribution to the growth of foreign trade in guangdong .
-# ::tok export of high - tech products has frequently been in the spotlight , making a significant contribution to the growth of foreign trade in guangdong .
-# ::alignments 24-25|0 22-23|0.0 21-22|0.0.1.0.1.0 19-20|0.0.1 16-17|0.0.1.0 15-16|0.0.1.0.1 11-12|0.0.0.0 7-8|0.0.0 5-6|0.0.1.0.0.0 4-5|0.0.1.0.0.0.0 2-3|0.0.1.0.0.0.0.0 0-1|0.0.1.0.0 ::annotator JAMR dev v0.3 ::date 2019-07-08T23:42:17.243
-(g / guangdong
-      :ARG2-of (t / trade-01
-            :ARG1 (f2 / frequent-02
-                  :ARG1 (s2 / spotlight))
-            :ARG1-of (g2 / grow-01
-                  :ARG2-of (c / contribute-01
-                        :ARG1 (e / export-01
-                              :ARG1 (p / product
-                                    :mod (t2 / tech
-                                          :ARG1-of (h / high-02))))
-                        :ARG1-of (s / significant-02
-                              :compared-to (f / foreign))))))
+(d / describe-01
+      :ARG0 (h / he)
+      :ARG1 (s / she)
+      :ARG2 (g / genius))
\ No newline at end of file

AMR-Visualization-master/AMR-Visualization-master/test_single.txt

-gold :op1 interrogative
-know :arg0 you :arg1 it :beneficiary interrogative :mod ( country :name ( name :op1 wouldn )  )  :polarity -
-(g / guangdong :ARG2-of (t / trade-01 :ARG1 (f2 / frequent-02 :ARG1 (s2 / spotlight)) :ARG1-of (g2 / grow-01 :ARG2-of (c / contribute-01 :ARG1 (e / export-01 :ARG1 (p / product :mod (t2 / tech :ARG1-of (h / high-02)))) :ARG1-of (s / significant-02 :compared-to (f / foreign))))))
+(d / describe-01
+      :ARG0 (h / he)
+      :ARG1 (s / she)
+      :ARG2 (g / genius))
\ No newline at end of file

amr_source.txt

-
+she was a genius, according to his description
+his description of her: genius
\ No newline at end of file

config/gigaword-L10_1500.json

+{
+  "loader": "loaders/gigaword.py",
+  "dataset": "data/train-data/gigaword",
+  "indices": "data/train-data/gigaword/indices.npy",
+  "model_dir": "data/models/gigaword-L10_1500",
+  "verbose": true,
+  "print_every": 1,
+  "eval_every": 500,
+  "save_every": 500,
+  "max_val_steps": 1500,
+  "max_train_seconds": null,
+  "max_train_steps": 1500,
+  "batch_size": 4,
+  "learning_rate": 1e-05,
+  "k_samples": 100,
+  "sample_aggregation": "max",
+  "loss": "pgb",
+  "encoder_model_id": "distilroberta-base",
+  "rewards": {
+    "Fluency": {
+      "weight": 1,
+      "type": "masked",
+      "model_id": "distilroberta-base",
+      "max_score": 40.0,
+      "norm": "max"
+    },
+    "BiEncoderSimilarity": {
+      "weight": 1,
+      "model_id": "all-distilroberta-v1"
+    },
+    "GaussianLength": {
+      "weight": 1,
+      "mean": 10,
+      "std": 3.2
+    }
+  }
+}

config/gigaword-L10_1500_then_AMR.json

@@ -5,11 +5,11 @@
   "model_dir": "data/models/gigaword-L10_1500_then_AMR",
   "verbose": true,
   "print_every": 1,
-  "eval_every": 50,
-  "save_every": 50,
-  "max_val_steps": 2100,
+  "eval_every": 100,
+  "save_every": 100,
+  "max_val_steps": 2000,
   "max_train_seconds": null,
-  "max_train_steps": 2100,
+  "max_train_steps": 2000,
   "batch_size": 4,
   "learning_rate": 1e-05,
   "k_samples": 100,

wesenberg/Konstanten.py

@@ -30,6 +30,7 @@ TOKENIZER = "distilroberta-base"
 MODEL_NAME_GIGAWORD_L8 = "gigaword-L8"
 MODEL_NAME_GIGAWORD_L8_test = "gigaword-L8_test"
 MODEL_NAME_GIGAWORD_L10_AMR = "gigaword-L10_AMR"
+MODEL_NAME_GIGAWORD_L10_1500 = "gigaword-L10_1500"
 MODEL_NAME_GIGAWORD_L10_2000 = "gigaword-L10_2000"
 MODEL_NAME_GIGAWORD_L10_2500 = "gigaword-L10_2500"
 MODEL_NAME_GIGAWORD_L10_3000 = "gigaword-L10_3000"
@@ -56,7 +57,7 @@ MODEL_NAME_P75 = "newsroom-P75"
 # TODO The Model-Names in this List will be trained
 
 # MODEL_TO_TRAIN_LIST = [MODEL_NAME_GIGAWORD_L10_2000_THEN_AMR]
-MODEL_TO_TRAIN_LIST = [MODEL_NAME_GIGAWORD_L10_2000_THEN_AMR]
+MODEL_TO_TRAIN_LIST = [MODEL_NAME_GIGAWORD_L10_1500_THEN_AMR]
 
 # TODO The Model-Names in this List will be used for Baseline
 LIST_MODEL_NAME = [MODEL_NAME_GIGAWORD_L10_2000_THEN_AMR_300, MODEL_NAME_GIGAWORD_L10_2000_THEN_AMR_600, MODEL_NAME_GIGAWORD_L10_1500_THEN_AMR_300, MODEL_NAME_GIGAWORD_L10_1500_THEN_AMR_600]

wesenberg/util.py

@@ -281,12 +281,12 @@ def write_string_to_file(path, name, model_name, amr_results_summaries, list_rou
         tmp += "\n\t\tAMR:\t\t\t" + str(np.mean(amr_score_gold_baseline))
 
         tmp += "\n\n2. Summary/Source -- Macht das Model was es soll?"
-        tmp += "\n\tModel AMR:"
+        tmp += "\n\tModel AMR+SMATCH:"
         tmp += "\n\t\tF1:\t\t\t\t" + str(np.mean(summe_amr_summary_source_f1))
         tmp += "\n\t\tPrecision:\t\t" + str(np.mean(summe_amr_summary_source_precision))
         tmp += "\n\t\tRecall:\t\t\t" + str(np.mean(summe_amr_summary_source_recall))
 
-        tmp += "\n\n\tBaseline AMR:"
+        tmp += "\n\n\tBaseline AMR+SMATCH:"
         tmp += "\n\t\tF1:\t\t\t\t" + str(np.mean(summe_amr_baseline_source_f1))
         tmp += "\n\t\tPrecision:\t\t" + str(np.mean(summe_amr_baseline_source_precision))
         tmp += "\n\t\tRecall:\t\t\t" + str(np.mean(summe_amr_baseline_source_recall))
@@ -298,7 +298,7 @@ def write_string_to_file(path, name, model_name, amr_results_summaries, list_rou
         tmp += "\n\t\tst.dev.:\t\t" + str(stdev(summe_rouge_1_fmeasure_summary_baseline))
         tmp += "\n\t\tvariance:\t\t" + str(variance(summe_rouge_1_fmeasure_summary_baseline))
 
-        tmp += "\n\n\tAMR:"
+        tmp += "\n\n\tAMR+SMATCH:"
         tmp += "\n\t\tmean:\t\t\t" + str(np.mean(amr_score_summary_baseline))
         tmp += "\n\t\tmedian:\t\t\t" + str(median(amr_score_summary_baseline))
         tmp += "\n\t\tst.dev.:\t\t" + str(stdev(amr_score_summary_baseline))