基于word2vec与LVQ的作业查重及评分系统.PDFVIP

鍩轰簬word2vec 涓嶭VQ 鐨勪綔涓氭煡閲嶅強璇勫垎绯荤粺 1 1 寮犲楹 锛屽瓟绻佸 (瑗垮崡澶у 璁＄畻鏈哄闄紝閲嶅簡甯 400715)1 鎽 瑕侊細涓轰簡鍑忓皯鏁欏笀鍦ㄦ壒鏀逛綔涓氫笂鑺辫垂鐨勫ぇ閲忔椂闂达紝璁捐鍑轰簡涓€涓熀浜巜ord2vec涓嶭VQ绁炵粡缃戠粶鐨勫鐢熶綔 涓氭煡閲嶅強璇勫垎绯荤粺銆傚厛杩愮敤LCS绠楁硶瀵规枃绔犺繘琛屾煡閲嶏紝鍐嶈鑰佸笀瀵?0%鐨勫鐢熶綔涓氳繘琛岃瘎鍒嗭紝浠庢瘡浠藉凡璇勫垎 浣滀笟涓娇鐢═F- IDF绠楁硶鎻愬彇鍏抽敭瀛楋紝閫氳繃鐢╳ord2vec鑾峰彇鍏抽敭瀛楃殑璇嶅悜閲忥紝鍐嶆妸姣忕瘒鐨勫涓瘝鍚戦噺鍜岃瘎鍒 浼犲叆LVQ绁炵粡缃戠粶杩涜璁粌寰楀嚭妯″瀷锛岀劧鍚庡鍏朵綑80%鐨勬枃绔犻噰鐢ㄥ悓鏍风殑鏂规硶鎻愬彇鍏抽敭瀛楋紝鎶婅繖浜涜瘝鍚戦噺浼 鍏VQ绁炵粡缃戠粶寰楀埌鍒嗘暟銆傚苟鍦ㄦ垜闄㈠鐢熶綔涓氫笂杩涜浜嗗疄楠岋紝鍦ㄧ畝绛旈銆佽嚜鎴戣瘎浠枫€佹€荤粨鍒嗘瀽绫绘枃鏈笂寰楀埌 浜嗙浉褰撲笉閿欑殑瀹為獙缁撴灉锛屼絾鍦ㄤ綔鏂囪繖鏍锋湁鍙ュ瓙缁撴瀯鐨勬枃鏈笂鏁堟灉涓€鑸€ 鍏抽敭璇嶏細LVQ绁炵粡缃戠粶锛泈ord2vec 锛涙煡閲嶈瘎鍒嗙郴缁燂紱鑷劧璇█澶勭悊; 涓浘鍒嗙被鍙凤細TP183 鏂囩尞鏍囧織鐮侊細A DOI: Design and Implementation of the Checking and Grading System for Students Homework ZHANG Jia-lin1 KONG Fan-chen1 (School of Computer and Information Science 锛孲outhwest University 锛孋hongqing 400715锛孋hina)1 Abstract: In order to reduce the amount of time spent on correcting homework, we design a Checking and Grading System based on word2vec and LVQ neural network. The LCS algorithm is used to check students鈥 homework at first, after that, we let the teacher to grade 20% of the students鈥 homework. By using the TF-IDF algorithm to extract keywords from the scored homework, we apply the word2vec to get the key words鈥 vectors, then the words鈥 vectors and their corresponding scores are used to train the LVQ neural network. As for the 80% homework remained, we apply the trained LVQ network to evaluate their scores by using the same step. Experiments were carried out on the students homework in our college, in the text of short answer, self-evaluation, summarize and analyze text, the results are very good, however, in the text which has sentence structure, such as composition, this model doesn鈥檛 work well. Key words: Learning Vector Quantization Neural Network ; Word