better recall strategy

trafilatura/external.py

@@ -88,7 +88,6 @@ def compare_extraction(tree, backup_tree, body, text, len_text, options):
         LOGGER.debug('using custom extraction: %s', options.source)
     # override faulty extraction: try with justext
     if body.xpath(SANITIZED_XPATH) or len_text < options.min_extracted_size:  # body.find(...)
-    # or options.focus == "recall" is True ?
         LOGGER.debug('unclean document triggering justext examination: %s', options.source)
         # tree = prune_unwanted_sections(tree, {}, options)
         body2, text2, len_text2, jt_result = justext_rescue(tree, options, body, 0, '')

trafilatura/htmlprocessing.py

@@ -63,17 +63,16 @@ def tree_cleaning(tree, options):
 
     # prevent removal of paragraphs
     run_p_test = False
-    if options.focus == "recall":
+    if options.focus == "recall" and tree.find('.//p') is not None:
         tcopy = deepcopy(tree)
-        if tree.find('.//p') is not None:
-            run_p_test = True
+        run_p_test = True
 
     # delete targeted elements
     for expression in cleaning_list:
         for element in tree.getiterator(expression):
             delete_element(element)
-            if run_p_test and tree.find('.//p') is None:
-                tree = tcopy
+    if run_p_test and tree.find('.//p') is None:
+        tree = tcopy
 
     return prune_html(tree)
 

trafilatura/main_extractor.py

@@ -548,9 +548,7 @@ def _extract(tree, options):
             continue
         # no paragraphs containing text, or not enough
         ptest = subtree.xpath('//p//text()')
-        if options.focus == "recall":
-            factor = 5
-        elif options.focus == "precision":
+        if options.focus == "precision":
             factor = 1
         else:
             factor = 3