The spread of fake news on online media is very dangerous and can lead to casualties, effects on psychology, character assassination, elections for political parties, and state chaos. Fake news that concerning Covid-19 massively spread during the pandemic. Detecting misinformation on the Internet is an essential and challenging task since humans face difficulty detecting fake news. We applied BERT and GPT2 as pre-trained using the BiGRU-Att-CapsuleNet model and BiGRU-CRF features augmentation to solve Fake News detection in Constraint @ AAAI2021 - COVID19 Fake News Detection in English Dataset. This research proved that our hybrid model with augmentation got better accuracy compared to our baseline model. It also showed that BERT gave a better result than GPT2 in all models; the highest accuracy we achieved for BERT is 0.9196, and GPT2 is 0.8986.

Protein remote homology detection is one of the central problems in bioinformatics. Although some computational methods have been proposed, the problem is still far from being solved. In this paper, an ensemble classifier for protein remote homology detection, called SVM-Ensemble, was proposed with a weighted voting strategy. SVM-Ensemble combined three basic classifiers based on different feature spaces, including Kmer, ACC, and SC-PseAAC. These features consider the characteristics of proteins from various perspectives, incorporating both the sequence composition and the sequence-order information along the protein sequences. Experimental results on a widely used benchmark dataset showed that the proposed SVM-Ensemble can obviously improve the predictive performance for the protein remote homology detection. Moreover, it achieved the best performance and outperformed other state-of-the-art methods.

Presently, sentence-level researches are very significant in fields like natural language processing, information retrieval, machine translation etc. In this paper we present a practical task on sentence classification. The main purpose of this work is to classify the abstract sentences of scientific papers in the corpus built by ourselves into four categories- the background, the goal, the method and the result- which differ from each other in common usage, so that we can do further researches such as frequent pattern mining, information extraction and making a corpus for writing assistant system of scientific paper with these results. The main method of the classification is the Support Vector Machine, which is acknowledged among the best machine learning methods in the common text classification tasks. A semi-supervised method, Transductive Support Vector Machine, is also introduced into this four-class classification task to improve the accuracy. The experiments are conducted upon the corpus made by ourselves that consists of abstract sentences of scientific papers. The accuracy of the classifier finally reaches 75.86% with the semi-supervised method.

In human-to-human conversations, the context generally provides several backgrounds and strategic points for the following response. Therefore, many response generation approaches have explored the methodologies to incorporate the context into the Encoder-Decoder architecture, which is the fundamental of neural generation tasks, to generate context-aware responses that are remarkably relevant and cohesive to the given context. However, most approaches pay less attention to semantic interactions implicitly existing within contextual utterances, which are of great importance to capture semantic clues of the given dialog context, indeed. This paper proposes a dynamic working memory mechanism to model long-term semantic hints hidden in the conversation context, by performing semantic interactions between utterances and updating context representation dynamically. Then, the outputs of the dynamic working memory are employed to provide helpful clues for the Encoder-Decoder architecture to generate responses to the given dialog. We have evaluated the proposed approach on Twitter Customer Service Corpus and OpenSubtitles Corpus, with several automatic evaluation metrics and the human evaluation, and the empirical results show the effectiveness of the proposed method.