Narrow Results

Sina Weibo, the most popular Chinese social platform with hundreds and millions of user-contributed images and texts, is growing rapidly. However, the noise between the image and text, as well as their incomplete correspondence, makes accurate image retrieval and ranking difficult. In this paper, we propose a deep learning framework using visual features, text content and popularity of Weibo to calculate the similarity between the image and the text based on training the model to maximize the likelihood of the target description sentence given the training image. In addition, the retrieval results are reranked using the popularity of the image. The comparison experiment of the large-scale Sina Weibo dataset proves the validity of the proposed method.

The research object of this paper is rumors, mainly focusing on the dissemination characteristics of rumors on Weibo platforms and conducting rumor attribute detection. Aiming at the problems of low accuracy and poor timeliness of current rumor detection methods, a network rumor detection method using attention mechanism and Gated Recurrent Unit (GRU) neural network is proposed. First, pre-processing operations such as data noise cleaning, Chinese word segmentation and stop word removal are performed on the constructed Weibo corpus data, and word vectors are obtained by using the Continuous Bag-of-Words (CBOW) model in the Word2vec neural language model. Then, the Bidirectional GRU neural network (BiGRU) is used to obtain information, combined with the structured attention mechanism to build an GRU rumor detection model. The introduction of attention mechanism makes the network model tend to grasp the text Semantic information. Finally, the Adam algorithm is used to optimize the proposed model, and the loss function of the model is constructed and minimized in combination with the binary cross-entropy loss function. The results show that the detection precision rate, recall rate, F1 value and accuracy rate of the proposed method are the largest, reaching 95.28%, 88.78%, 88.69% and 94.89%, respectively, and the detection performance is better than the other three comparison algorithms.