Research And Realization Of Character Relationship Graph In Microblogs Based On Deep Learning

The two methods dominating RE research in the last decade are the feature-based method and the kernel-based method.These research extensively studies the leverage of linguistic analysis and knowledge resources to construct the feature representations,involving the combination of discrete properties such as lexicon,syntax,gazetteers.Although these approaches are able to exploit the symbolic(discrete)structures within relation mentions,they also suffer from the difficulty to generalize over the unseen words,motivating some very recentwork on employing the continuous representations of words(word embeddings)to do RE.Recently,convolutional and recurrent neural networks has provided very effective mechanisms to capture the hidden structures within sentences via continuous representations,thereby significantly advancing the performance of relation extraction.The advantage of convolutional neural networks is their capacity to generalize the consecutive n-grams in the sentences while recurrent neural networks are effective to encode long ranges of sentence context.This thesis proposes to combine the traditional feature-based method,the convolutional and recurrent neural networks to simultaneously benefit from their advantages.Our systematic evaluation of different network architectures and combination methods demonstrates the effectiveness of this approach and results in the state-of-the-art performance on the ACE 2005 and SemEval dataset.The main content of the thesis includes:(1)Set out the structure and principle of CNN and RNNs and perform a systematic exploration of various network architectures to seek the best RNN model for RE.(2)In the next step,this thesis extensively study different methods to assemble the CNNs and RNNs for RE.In order to further improve the RE performance of models above,we investigate the integration of these neural network models with the traditional loglinear model that relies on various linguistic features that yield the state-of-the-art performance on the ACE 2005 and SemEval dataset.(3)Finally,this thesis applies the hybrid model to the Sina Microblogs,and achive the extraction and visual display of character relationship extraction,experiments show that the hybrid model has outstand performance in informal texts and unseen domains...