tei_to_fulltext.py

tei_to_fulltext.py

NS = {'tei': 'http://www.tei-c.org/ns/1.0'}

def tei_to_fulltext(tei):
    from lxml import etree
    from six import text_type
    parser = etree.XMLParser(encoding='UTF-8', recover=True)
    tei = tei if not isinstance(tei, text_type) else tei.encode('utf-8')
    root = etree.fromstring(tei, parser)
    return get_fulltext(root)


def get_fulltext(root):
    """TODO: move it to invenio_grobid`"""
    return ' '.join(root.xpath(
        '/tei:TEI/tei:text//text()',
        namespaces=NS))