clean up code

app.py

@@ -98,9 +98,7 @@ def chat(
     partial_text += f"1. Text Repetition Score: {repetition_score:.3f}\n"
     partial_text += f"1. Total Repetitions: {total_repetitions:.3f}\n"
     rr = total_repetitions / len(answer) if len(answer) > 0 else 0
-    partial_text += (
-        f"1. Repetition Ratio: {rr:.3f}\n"
-    )
+    partial_text += f"1. Repetition Ratio: {rr:.3f}\n"
 
     if index >= 0:  # RAG
         key = (
@@ -114,9 +112,9 @@ def chat(
         partial_text += f'1. BLEU-1: {scores["bleu_scores"]["bleu"]:.3f}\n'
         partial_text += f'1. RougeL: {scores["rouge_scores"]["rougeL"]:.3f}\n'
         perf = scores["bert_scores"]["f1"][0]
-        partial_text += f'1. BERT-F1: {perf:.3f}\n'
+        partial_text += f"1. BERT-F1: {perf:.3f}\n"
         nrr = 1 - rr
-        partial_text += f'1. RAP-BERT-F1: {perf * nrr * nrr * nrr:.3f}\n'
+        partial_text += f"1. RAP-BERT-F1: {perf * nrr * nrr * nrr:.3f}\n"
 
         partial_text += f"\n\nGround truth: {questions[index][key][0]}\n"
 

eval_modules/utils.py

@@ -7,6 +7,11 @@ import pandas as pd
 
 print(f"loading: {__file__}")
 
+
+bleu = evaluate.load("bleu")
+rouge = evaluate.load("rouge")
+bert_score = evaluate.load("bertscore")
+
 # pattern_non_word_char_repetition = re.compile(r"\s{5,}")
 # pattern_text_repetitions = re.compile(r"(.{5}.*)\s*((\1)\s*)+", re.M | re.DOTALL)
 
@@ -81,12 +86,6 @@ def detect_repetitions(text, debug=False):
     return result
 
 
-
-bleu = evaluate.load("bleu")
-rouge = evaluate.load("rouge")
-bert_score = evaluate.load("bertscore")
-
-
 def calc_perf_scores(predictions, references, debug=False):
     if debug:
         print("predictions:", predictions)
@@ -112,73 +111,3 @@ def calc_perf_scores(predictions, references, debug=False):
         print("result:", result)
 
     return result
-
-
-def calc_metrics(df):
-    predictions = [df["answer"][i] for i in range(len(df))]
-    references = [df["ground_truth"][i] for i in range(len(df))]
-
-    return calc_bleu_rouge_scores(predictions, references)
-
-
-pattern_abnormal_newlines = re.compile(r"\n{5,}")
-pattern_text_repetitions = re.compile(r"\b(\w.+?)\b(\1+)", re.M | re.DOTALL)
-exception_pattern = re.compile(r"(\w+\.)\1")
-
-
-# final version for repetition detection
-def detect_repetitions(
-    text, debug=False, pattern_text_repetitions=pattern_text_repetitions
-):
-    subtotals = [0, 0]
-
-    if isinstance(text, str):
-        patterns = [pattern_abnormal_newlines, pattern_text_repetitions]
-        for i, pattern in enumerate(patterns):
-            if debug:
-                print(
-                    f"----detect {'abnormal newlines' if i == 0 else 'text repetitions'}----"
-                )
-            matches = pattern.finditer(text)
-            for match in matches:
-                if debug:
-                    print(match)
-                    for groupNum in range(0, len(match.groups())):
-                        groupNum = groupNum + 1
-                        print(
-                            "Group {groupNum} found at {start}-{end}: `{group}`".format(
-                                groupNum=groupNum,
-                                start=match.start(groupNum),
-                                end=match.end(groupNum),
-                                group=match.group(groupNum),
-                            )
-                        )
-
-                if exception_pattern.match(match[0]):
-                    if debug:
-                        print("ignored: ", match[0])
-                    continue
-
-                start, end = match.span()
-                subtotals[i] += end - start
-
-    result = (subtotals[0], subtotals[1], subtotals[0] + subtotals[1])
-
-    if debug:
-        print(result)
-    return result
-
-
-def detect_abnormal_newlines(text, debug=False):
-    return detect_repetitions(text, debug=debug)[0]
-
-
-def detect_text_repetitions(text, debug=False):
-    return detect_repetitions(text, debug=debug)[1]
-
-
-def detect_repetition_scores(text, debug=False):
-    newline_score, repetition_score, total_repetitions = detect_repetitions(
-        text, debug=debug
-    )
-    return pd.Series([newline_score, repetition_score, total_repetitions])