The proliferation of spam in China has a negative impact on internet users’ experiences online. Existing methods for detecting spam are primarily based on machine learning. However, it has been discovered that these methods are susceptible to adversarial textual spam that has frequently been imperceptibly modified by spammers. Spammers continually modify their strategies to circumvent spam detection systems. Text with Chinese homophonic substitution may be easily understood by users according to its context. Currently, spammers widely use homophonic substitution to break down spam identification systems on the internet. To address these issues, we propose a Bidirectional Gated Recurrent Unit (BiGRU)–Text Convolutional Neural Network (TextCNN) hybrid model with joint embedding for detecting Chinese spam. Our model effectively uses phonetic information and combines the advantages of parameter sharing from TextCNN with long-term memory from BiGRU. The experimental results on real-world datasets show that our model resists homophone noise to some extent and outperforms mainstream deep learning models. We also demonstrate the generality of joint textual and phonetic embedding, which is applicable to other deep learning networks in Chinese spam detection tasks.

Content-based image retrieval (CBIR) is the problem of searching for items in an image database that are similar to the query image. Most of the existing image retrieval methods are trained based on metric learning loss functions (e.g. contrastive loss or triplet loss), however, which require the use of hard sample mining strategies (HMS) to better train the model. The HMS implies that picking out hard positive or negative samples increases the complexity of model training and requires a large amount of additional training time. To address this issue, lessons from recent work are leveraged on representation learning and a model called GS is proposed that combines the state-of-the-art Generalized-Mean (GeM) pooling and the smoothed average precision (AP). The entire network can be learned end-to-end by approximating the non-differentiable AP function to a differentiable onewithout mining hard samples, only image-level annotations. A model named GSA is also presented which achieves excellent retrieval performance jointly trained by two various loss functions. Experimental results validate the effectiveness of the proposed approach and demonstrate the competitive performance on a common standard image retrieval dataset (Revisited Oxford and Paris).