invalid byte sequence for encoding "UTF8": 0xf480

My db is in UTF-8 encoding. When I tried to generate gist index for some
rows I got the strange error:
select to_tsvector('chinesecfg',content) from article where
article_id=23834922;
ERROR:  invalid byte sequence for encoding "UTF8": 0xf480
However, "select content from article where article_id=23834922" works
just fine and shows good content. Why the to_tsvector reported invalid
byte sequence since the 'content' already exists in db as UTF-8?