get_errors script; edit_dist file

scripts/get_errors.py

@@ -0,0 +1,14 @@
+import sys
+
+import pandas as pd
+
+if __name__ == "__main__":
+    in_path = sys.argv[1]
+    in_path = sys.argv[1]
+    out_path = sys.argv[2]
+    number = int(sys.argv[3])
+
+    df = pd.read_csv(in_path, sep='\t', keep_default_na=True)
+    correct = df['correct@5']
+    out_df = df[~correct].sample(n=number)
+    out_df.to_csv(out_path, sep='\t', index=None)

sound_law/evaluate/edit_dist.py

@@ -0,0 +1,32 @@
+from typing import Sequence, Union, overload
+
+import numpy as np
+from lingpy.align.pairwise import pw_align
+
+from dev_misc.devlib import NDA
+from editdistance import eval as ed_eval
+from editdistance import eval_all as ed_eval_all
+
+Number = Union[int, float]
+
+
+def edit_dist(seq_0: str, seq_1: str, mode: str) -> Number:
+    """A master function for dispatching different methods of computing edit distance."""
+    if mode == 'ed':
+        return ed_eval(seq_0, seq_1)
+    elif mode == 'global':
+        l0 = len(seq_0)
+        l1 = len(seq_1)
+        return max(l0, l1) - pw_align(seq_0, seq_1, mode='global')[-1]
+    else:
+        raise ValueError(f'Unrecognized value "{mode}" for mode.')
+
+
+def edit_dist_all(seqs_0: Sequence[str], seqs_1: Sequence[str], mode: str) -> NDA:
+    if mode == 'ed':
+        return ed_eval_all(seqs_0, seqs_1)
+
+    ret = list()
+    for seq_0 in seqs_0:
+        ret.append([edit_dist(seq_0, seq_1, mode) for seq_1 in seqs_1])
+    return np.asarray(ret)