llm-ie 0.3.3__py3-none-any.whl → 0.3.5__py3-none-any.whl

llm_ie/extractors.py

@@ -1,6 +1,7 @@
 import abc
 import re
 import json
+import json_repair
 import inspect
 import importlib.resources
 import warnings
@@ -117,7 +118,12 @@ class Extractor:
                 dict_obj = json.loads(dict_str)
                 out.append(dict_obj)
             except json.JSONDecodeError:
-                warnings.warn(f'Post-processing failed:\n{dict_str}', RuntimeWarning)
+                dict_obj = json_repair.repair_json(dict_str, skip_json_loads=True, return_objects=True)
+                if dict_obj:
+                    warnings.warn(f'JSONDecodeError detected, fixed with repair_json:\n{dict_str}', RuntimeWarning)
+                    out.append(dict_obj)
+                else:
+                    warnings.warn(f'JSONDecodeError could not be fixed:\n{dict_str}', RuntimeWarning)
         return out
     
 
@@ -181,19 +187,22 @@ class FrameExtractor(Extractor):
         Returns : Tuple[Tuple[int, int], float]
             a tuple of 2-tuple span and Jaccard score.
         """
+        if not text or not pattern:
+            return None, 0
+
         text_tokens, text_spans = self._get_word_tokens(text)
         pattern_tokens, _ = self._get_word_tokens(pattern)
         pattern_tokens_set = set(pattern_tokens)
         window_size = len(pattern_tokens)
-        window_size_min = int(window_size * (1 - buffer_size))
-        window_size_max = int(window_size * (1 + buffer_size))
+        window_size_min = max(1, int(window_size * (1 - buffer_size)))
+        window_size_max = int(window_size * (1 + buffer_size)) + 1
         closest_substring_span = None
         best_score = 0
         
         for i in range(len(text_tokens) - window_size_max):
             for w in range(window_size_min, window_size_max): 
                 sub_str_tokens = text_tokens[i:i + w]
-                if sub_str_tokens[0] == pattern_tokens[0]:
+                if len(sub_str_tokens) > 0 and sub_str_tokens[0] == pattern_tokens[0]:
                     score = self._jaccard_score(set(sub_str_tokens), pattern_tokens_set)
                     if score > best_score:
                         best_score = score
@@ -238,7 +247,7 @@ class FrameExtractor(Extractor):
 
             # Exact match  
             match = re.search(re.escape(entity), text)
-            if match:
+            if match and entity:
                 start, end = match.span()
                 entity_spans.append((start, end))
                 # Replace the found entity with spaces to avoid finding the same instance again
@@ -246,7 +255,7 @@ class FrameExtractor(Extractor):
             # Fuzzy match
             elif fuzzy_match:
                 closest_substring_span, best_score = self._get_closest_substring(text, entity, buffer_size=fuzzy_buffer_size)
-                if best_score >= fuzzy_score_cutoff and closest_substring_span:
+                if closest_substring_span and best_score >= fuzzy_score_cutoff:
                     entity_spans.append(closest_substring_span)
                     # Replace the found entity with spaces to avoid finding the same instance again
                     text = text[:closest_substring_span[0]] + ' ' * (closest_substring_span[1] - closest_substring_span[0]) + text[closest_substring_span[1]:]

{llm_ie-0.3.3.dist-info → llm_ie-0.3.5.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: llm-ie
-Version: 0.3.3
+Version: 0.3.5
 Summary: An LLM-powered tool that transforms everyday language into robust information extraction pipelines.
 License: MIT
 Author: Enshuo (David) Hsu
@@ -10,6 +10,7 @@ Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Requires-Dist: colorama (>=0.4.6,<0.5.0)
+Requires-Dist: json_repair (>=0.30,<0.31)
 Requires-Dist: nltk (>=3.8,<4.0)
 Description-Content-Type: text/markdown
 

{llm_ie-0.3.3.dist-info → llm_ie-0.3.5.dist-info}/RECORD

@@ -16,8 +16,8 @@ llm_ie/asset/prompt_guide/SentenceFrameExtractor_prompt_guide.txt,sha256=oKH_QeD
 llm_ie/asset/prompt_guide/SentenceReviewFrameExtractor_prompt_guide.txt,sha256=oKH_QeDgpw771ZdHk3L7DYz2Jvfm7OolUoTiJyMJI30,9541
 llm_ie/data_types.py,sha256=hPz3WOeAzfn2QKmb0CxHmRdQWZQ4G9zq8U-RJBVFdYk,14329
 llm_ie/engines.py,sha256=PTYs7s_iCPmI-yFUCVCPY_cMGS77ma2VGoz4rdNkODI,9308
-llm_ie/extractors.py,sha256=yBdIcevjMfwto85Jb0KkRMN-AjIMk92fD5yWB3Qm8MY,64408
+llm_ie/extractors.py,sha256=WM-9ZmhGYCYzXLPiShfF42grezjRz42JbgXwueixZRI,64870
 llm_ie/prompt_editor.py,sha256=Xc5ZHsEnM8-YYITokIsM6BVsf2Ec_8ajJDaldPf-P8U,8577
-llm_ie-0.3.3.dist-info/METADATA,sha256=CeTsMNtWhEWCvOqHWSXu0KqOgDp3kMwN2WtBF4N-4zE,41266
-llm_ie-0.3.3.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-llm_ie-0.3.3.dist-info/RECORD,,
+llm_ie-0.3.5.dist-info/METADATA,sha256=y5nABgPeeMPEkZ58WoOBR3sgFrKOE_mF4fqaHK59K1w,41308
+llm_ie-0.3.5.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
+llm_ie-0.3.5.dist-info/RECORD,,