diff options
| -rw-r--r-- | python/sandcrawler/ingest_html.py | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/python/sandcrawler/ingest_html.py b/python/sandcrawler/ingest_html.py index 7e6e5e3..9c72dd5 100644 --- a/python/sandcrawler/ingest_html.py +++ b/python/sandcrawler/ingest_html.py @@ -21,7 +21,7 @@ TRAFILATURA_AGENT = f"trafilatura/{trafilatura.__version__}" def html_extract_body_teixml(doc: bytes) -> dict: try: tei_xml = trafilatura.extract(doc, - tei_output=True, + output_format='xmltei', include_comments=False, include_formatting=True, ) |