Amharic Named Entity Recognition Using Neural Word Embedding As A Feature

In this paper, Amharic Named Entity Recognition problem is addressed by employingrna semi-supervised learning approach based on neural networks. The proposedrnapproach aims at automating manual feature design and avoiding dependency on otherrnnatural language processing tasks for classi cation features. In this work potential featurerninformation represented as word vectors are generated using neural network fromrnunlabeled Amharic text les. These generated features are used as features for AmharicrnNamed entity classi cation.rnSVM, J48, random tree, IBk(Instance based learning with parameter k), attributernselected and OneR(one rule) classi ers are tested with word vector features. AdditionallyrnBLSTM(bi-directional long short term memory), LSTM(long short term memory)rnand MLP(multi layer perceptron) deep neural networks are also tested to investigaternthe impact of proposed approach.rnFrom the experiments the highest F-score achieved was 95.5% using the SVM classi er.rnRelative to state-of-the-art approaches (SVM and J48) an average F-score improvementrnof 3.95% was achieved. The results showed that, automatically learned wordrnfeatures can substitute manually designed features for Amharic named entity recognition.rnAlso these features has given better performance while reducing the e ort inrnmanual feature design.

Get Full Work

Report copyright infringement or plagiarism