Headline Splitter Exceptions problems

src/ragas/testset/transforms/splitters/headline.py

@@ -1,10 +1,40 @@
+import re
 import typing as t
 from dataclasses import dataclass
+from unidecode import unidecode
 
 from ragas.testset.graph import Node, NodeType, Relationship
 from ragas.testset.transforms.base import Splitter
 
 
+def normalize_text(text):
+    return unidecode(re.sub(r'\s+', '', text).lower())
+
+
+def remove_indices(text):
+    cleaned_text = re.sub(r'(\d+\.)+ *', '', text)
+    return cleaned_text
+
+
+def adjust_indices(original_text, indices):
+    last_index = 0
+    count = 0
+
+    indices = sorted(indices)
+    new_indices = []
+    for index in indices:
+        while last_index < len(original_text):
+            if not original_text[last_index].isspace():
+                count += 1
+            if count == index + 1:
+                new_indices.append(last_index)
+                last_index += 1
+                break
+            last_index += 1
+
+    return new_indices
+
+
 @dataclass
 class HeadlineSplitter(Splitter):
     async def split(self, node: Node) -> t.Tuple[t.List[Node], t.List[Relationship]]:
@@ -16,14 +46,41 @@ async def split(self, node: Node) -> t.Tuple[t.List[Node], t.List[Relationship]]
         if headlines is None:
             raise ValueError("'headlines' property not found in this node")
 
+        if len(headlines) == 0:
+            return [], []
+
         # create the chunks for the different sections
         indices = []
+        normalized_text = normalize_text(text)
+
         for headline in headlines:
-            indices.append(text.find(headline))
+            if headline is not None and not headline.isspace():
+                indice = normalized_text.find(normalize_text(headline))
+                if indice == -1:
+                    text_search = remove_indices(headline)
+                    text_search = normalize_text(text_search)
+                    indice = normalized_text.find(text_search)
+
+                if indice != -1:
+                    indices.append(indice)
+
+        if len(indices) == 0:
+            return [], []
+
+        indices = adjust_indices(text, indices)
+
         indices.append(len(text))
-        chunks = [text[indices[i] : indices[i + 1]] for i in range(len(indices) - 1)]
 
-        # create the nodes
+        chunks = []
+        for i in range(len(indices) - 1):
+            aux = text[indices[i] : indices[i + 1]]
+
+            if not aux.isspace():
+                chunks.append(aux)
+
+        if len(chunks) == 0:
+            return [], []
+
         nodes = [
             Node(type=NodeType.CHUNK, properties={"page_content": chunk})
             for chunk in chunks