As reported in https://github.com/pelias/pelias/issues/847, we can improve fuzzy-matching by applying an apostrophe tokenfilter. https://www.elastic.co/guide/en/elasticsearch/reference/current/analysis-apostrophe-tokenfilter.html