Update ppt_chunker.py

ppt_chunker.py

@@ -7,6 +7,8 @@ from unstructured.cleaners.core import clean_ordered_bullets, clean_bullets, cle
 
 from ordered_multimap import OrderedMultiIndexMapWeakRef
 
+COMMON_WORDS = {"BY", "IN", "ON", "AT", "WITH", "FROM", "TO", "AND", "OR", "BUT"}
+
 def process_chunk(chunk, nlp):
     marked = []
     
@@ -24,7 +26,7 @@ def process_chunk(chunk, nlp):
                 
             if (type(next) is list) and next[1].isupper() and (next[0] == 'Title'):
                 print(f'TOKEN: {current_text}, {tokens[-1]}, {tokens[-1].pos_}')
-                if tokens[-1].pos_ in ["ADP", 'PART', 'PRON', 'DET', "AUX", 'SCONJ', 'CONJ', "CCONJ"]:
+                if (tokens[-1].pos_ in ["ADP", 'ADV', 'PART', 'PRON', 'DET', "AUX", 'SCONJ', 'CONJ', "CCONJ"]) or ((tokens[-1].pos_ == 'PROPN') and (tokens[-1] in COMMON_WORDS)):
                     chunk[1][i+1][1] = current_text + ' ' + next[1]
                     marked.append(i)