Add formatted output for error-counts

seqscore/conll.py

@@ -2,7 +2,17 @@
 from collections import defaultdict
 from itertools import chain
 from statistics import mean, stdev
-from typing import Any, DefaultDict, Iterable, List, Optional, Sequence, TextIO, Tuple
+from typing import (
+    Any,
+    Counter,
+    DefaultDict,
+    Iterable,
+    List,
+    Optional,
+    Sequence,
+    TextIO,
+    Tuple,
+)
 
 from attr import attrib, attrs
 from tabulate import tabulate
@@ -466,13 +476,46 @@ def score_conll_files(
         all_acc_scores.append(class_scores)
 
         if error_counts:
-            headers = ("False positives", "False negatives")
-            counters = (class_scores.false_pos_examples, class_scores.false_neg_examples)
-            for error_header, counter in zip(headers, counters):
-                print(error_header)
-                for item, count in counter.items():
-                    print(" ".join(item.tokens), item.type, count, sep="\t")
-                print()
+            if multi_files:
+                raise ValueError(
+                    "Outputting error counts is only available for a single prediction file"
+                )
+
+            if output_format == FORMAT_CONLL:
+                raise ValueError(
+                    f"Format {repr(output_format)} is not supported with error counts"
+                )
+            elif output_format in (FORMAT_PRETTY, FORMAT_DELIM):
+                header = ["Count", "Error", "Type", "Tokens"]
+
+                # Combine counts across the two counters
+                combined_counts: Counter[Tuple[str, str, str]] = Counter()
+                for counter, error_type in zip(
+                    (class_scores.false_pos_examples, class_scores.false_neg_examples),
+                    ("FP", "FN"),
+                ):
+                    for item, count in counter.items():
+                        combined_counts[
+                            (error_type, item.type, " ".join(item.tokens))
+                        ] = count
+
+                rows = [
+                    [count, error_type, mention_type, token_str]
+                    for (
+                        error_type,
+                        mention_type,
+                        token_str,
+                    ), count in combined_counts.most_common()
+                ]
+
+                if output_format == FORMAT_PRETTY:
+                    print(tabulate(rows, header, tablefmt="github"))
+                else:
+                    # Delimited output
+                    score_summaries.append(delim.join(header))
+                    score_summaries.extend(_join_delim(row, delim) for row in rows)
+                    print("\n".join(score_summaries))
+
             # Exit early since all the following logic is for printing scores
             return