基于双向GRU-CRF中文包装产品实体识别.PDFVIP

DOI 锛?0.14135/ki.1006-3080.20180407001 鍩轰簬鍙屽悜GRU-CRF 鐨勪腑鏂囧寘瑁呬骇鍝佸疄浣撹瘑鍒 1飥 鏉庝竴鏂 锛 寮犳娆 (鍗庝笢鐞嗗伐澶у淇℃伅绉戝涓庡伐绋嬪闄? 涓婃捣200237 锛 鎽樿锛氫负浜嗗疄鐜板寘瑁呰涓氱殑淇℃伅鑷姩鎶藉彇锛岄渶瑕佸鏂囨湰涓殑鍖呰浜у搧杩涜鍛藉悕瀹炰綋璇嗗埆宸ヤ綔銆傝璁′簡涓€ 绉嶅熀浜庡弻鍚慓RU-CRF 鐨勪腑鏂囧寘瑁呬骇鍝佸疄浣撹瘑鍒柟娉曘€備互棰勮缁冪殑棰嗗煙璇嶅悜閲忎负杈撳叆锛岄€氳繃鍙屽悜GRU 缃戠粶 瀵逛笂涓嬫枃璇箟淇℃伅杩涜寤烘ā锛屽苟浣跨敤杈撳嚭绔殑CRF 灞傚鏈€浣虫爣绛惧簭鍒楄繘琛岄娴嬨€傚湪鍖呰浜у搧鏂囨湰鏁版嵁闆嗕笂 灏嗚妯″瀷涓庝紶缁熺殑搴忓垪鏍囨敞妯″瀷浠ュ強寰幆绁炵粡缃戠粶妯″瀷杩涜浜嗗姣旓紝瀹為獙缁撴灉琛ㄦ槑锛屾湰鏂囨ā鍨嬪叿鏈夎緝灏戜汉宸 鐗瑰緛骞查鍜屾洿楂樺噯纭巼鍜屽彫鍥炵巼绛変紭鐐广€ 鍏抽敭璇岤 鍛藉悕瀹炰綋璇嗗埆;鍙屽悜GRU 缃戠粶;CRF 妯″瀷;鍖呰浜у搧;娣卞害瀛︿範 涓浘鍒嗙被鍙耳 TP391 鏂囩尞鏍囧織鐮侊細A Chinese Packaging Product Entity Recognition Based on Bidirectional GRU-CRF 1 LI Yi-bin , ZHANG Huan-huan (School of Information Science and Engineering, East China University of Science and Technology, Shanghai 200237锛孋hina) Abstract: With a prevailing trend of packaging industry, there is a diversity in the naming conventions of packaging products. So named entity recognition (NER) of these products becomes a necessity for packaging information extraction. Statistically speaking, Chinese product names are characterized by a complex composition and a long length, which makes them more complex and difficult to recognize in textual corpora. Based on thorough research of current algorithms of NER, we introduce a Chinese packaging products NER method using bidirectional GRU- CRF model. GRU which is the abbreviation of gated recurrent unit, is an improved structure of hidden layer nodes in recurrent neural network (RNN). In this model, a bidirectional GRU network is used to store and