With　the　development　of　the　Internet　age,　collected　data　have　become　an　important　source　of　knowledge.　The　field　of　unstructured　text　contains　many　named　entities,　but　includes　very　little　detailed　information　about　those　entities.　However,　the　Baidu　encyclopedia　website　is　a　type　of　semistructured　data　that　in　many　cases　includes　a　detailed　introduction　of　entities.　By　combining　the　advantages　of　these　two　kinds　of　data,　we　can　enrich　the　knowledge　base　of　a　knowledge　graph.　This　paper　aims　to　extract　semistructured　data　consisting　of　named　entities　starting　from　raw　text　data.　On　one　hand,　this　paper　extracts　named　entities　with　the　help　of　the　Harbin　Institute　of　Technology　model,　parses　semistructured　content　about　the　named　entities　using　the　Octopus　tool,　constructs　a　local　ontology,　and　merges　the　ontology　using　Python＇s　built-in　difflib.　SequenceMatcher　tool　and　the　Deckard　similarity　algorithm.　On　the　other　hand,　we　create　an　XPath-based　wrapper　to　extract　the　attributes　and　attribute　values　of　named　entities　from　semistructured　data.　The　experimental　results　show　that　this　approach　can　extract　information　related　to　named　entities　from　the　Baidu　encyclopedia　automatically　to　supplement　the　knowledge　base　of　a　water　domain　knowledge　graph.　This　article　can　also　serve　as　a　reference　for　constructing　domain　knowledge　graphs　in　other　fields.　©　2019　IEEE.