Add alignment cleaning script

src/clean_alignment.py

@@ -0,0 +1,41 @@
+
+import os
+
+
+ENC = "utf-8"
+ALIGNED_FILE_NAME = "aligned"
+ALIGNED_CLEAN_FILE_NAME = "aligned_cleaned.txt"
+OUTPUT_DIR = "output"
+OUT_FILE_NAME = "out_hr.en.txt"
+
+
+def main():
+    outf = os.path.join(OUTPUT_DIR, OUT_FILE_NAME)
+    alf = os.path.join(OUTPUT_DIR, ALIGNED_FILE_NAME)
+    input_sentences = []
+    with open(outf) as f:
+        for l in f:
+            h, e = l.split(" ||| ")
+            input_sentences.append((h, e))
+
+    res_dic = []
+    i = 0
+    with open(alf) as f:
+        for l in f:
+            h, e = input_sentences[i]
+            hs = h.split()
+            es = e.split()
+            for t in l.split():
+                is1, is2 = t.split('-')
+                i1 = int(is1)
+                i2 = int(is2)
+                res_dic.append((hs[i1], es[i2]))
+            i += 1
+    alcf = os.path.join(OUTPUT_DIR, ALIGNED_CLEAN_FILE_NAME)
+    with open(alcf, "w") as f:
+        for h, e in res_dic:
+            f.write(f"{h} - {e}\n")
+
+
+if __name__ == "__main__":
+    main()