diff --git a/CHANGELOG.md b/CHANGELOG.md
index 03f85dbae1..ceacd2a3ad 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,8 @@
+## 0.18.24-dev0
+
+### Enhancement
+- Optimize `OCRAgentTesseract.extract_word_from_hocr` (codeflash)
+
 ## 0.18.23
 
 ### Enhancement
diff --git a/unstructured/__version__.py b/unstructured/__version__.py
index 9d7aede9fb..eb15f85ed4 100644
--- a/unstructured/__version__.py
+++ b/unstructured/__version__.py
@@ -1 +1 @@
-__version__ = "0.18.23"  # pragma: no cover
+__version__ = "0.18.24-dev0"  # pragma: no cover
diff --git a/unstructured/partition/utils/ocr_models/tesseract_ocr.py b/unstructured/partition/utils/ocr_models/tesseract_ocr.py
index a92540a018..efac81cd23 100644
--- a/unstructured/partition/utils/ocr_models/tesseract_ocr.py
+++ b/unstructured/partition/utils/ocr_models/tesseract_ocr.py
@@ -26,6 +26,8 @@
     from unstructured_inference.inference.elements import TextRegions
     from unstructured_inference.inference.layoutelement import LayoutElements
 
+_RE_X_CONF = re.compile(r"x_conf (\d+\.\d+)")
+
 # -- force tesseract to be single threaded, otherwise we see major performance problems --
 if "OMP_THREAD_LIMIT" not in os.environ:
     os.environ["OMP_THREAD_LIMIT"] = "1"
@@ -152,12 +154,12 @@ def extract_word_from_hocr(
         if len(character_spans) == 0:
             return ""
 
-        word_text = ""
+        chars = []
         for character_span in character_spans:
             char = character_span.text
 
             char_title = character_span.get("title", "")
-            conf_match = re.search(r"x_conf (\d+\.\d+)", char_title)
+            conf_match = _RE_X_CONF.search(char_title)
 
             if not (char and conf_match):
                 continue
@@ -165,9 +167,9 @@ def extract_word_from_hocr(
             character_probability = float(conf_match.group(1)) / 100
 
             if character_probability >= character_confidence_threshold:
-                word_text += char
+                chars.append(char)
 
-        return word_text
+        return "".join(chars)
 
     @requires_dependencies("unstructured_inference")
     def get_layout_elements_from_image(self, image: PILImage.Image) -> LayoutElements: