complete

Align.py

@@ -0,0 +1,68 @@
+from metric import Correct_Rate, Accuracy, Align
+from jiwer import wer
+import pandas as pd
+
+
+
+#Align ref_our, human_our, ref_human
+
+test = pd.read_csv("./test.csv")
+f = open("./ref_human_detail", 'a')
+cor_cnt = 0
+sub_cnt = 0
+ins_cnt = 0
+del_cnt = 0
+for i in range(len(test)):
+    # f.write("000" + str(test['Path'][i]) + " ")
+    path = test['Path'][i]
+    path = str(path)
+    seq1 = test['Canonical'][i]
+    seq2 = test['Transcript'][i]
+    seq1, seq2 = Align(seq1.split(" "), seq2.split(" "))
+    REF = ''
+    HYP = ''
+    OP = ''
+    cor = 0
+    sub = 0
+    ins = 0
+    dell = 0
+    for i in range(len(seq1)):
+        REF = REF  + seq1[i] + " "
+        HYP = HYP  + seq2[i] + " "
+        if seq1[i]!="<eps>" and seq2[i]=="<eps>":
+            OP = OP + "D" + " "
+            dell = dell + 1
+            del_cnt +=1
+        elif seq1[i] == "<eps>" and seq2[i]!="<eps>" :
+            OP = OP + "I" + " "
+            ins = ins + 1
+            ins_cnt+=1
+        elif (seq1[i]!=seq2[i]) and seq2[i]!="<eps>" and seq1[i]!="<eps>":
+            OP = OP + "S" + " "
+            sub = sub + 1
+            sub_cnt+=1
+        else:
+            OP = OP + "C" + " "
+            cor = cor + 1
+            cor_cnt+=1
+    # print(REF)
+    # print(HYP)
+    # print(OP)
+    cor = str(cor)
+    sub = str(sub)
+    ins = str(ins)
+    dell = str(dell)
+
+    # print(cor)
+    # print(sub)
+    # print(ins)
+    # print(dell)
+    f.write(path + " " + "ref" + " " + REF + "\n")
+    f.write(path + " " + "hyp" + " " + HYP + "\n")
+    f.write(path + " " + "op" + " " + OP + "\n")
+    f.write(path + " " + "#csid" + " " + cor + " " + sub + " " + ins + " " + dell + "\n")
+
+print(cor_cnt)
+print(sub_cnt)
+print(ins_cnt)
+print(del_cnt)