In this paper, we propose a novel graph convolutional network (GCN)-based recommendation method using both knowledge graph (KG) and review texts to solve the cold start problem and provide explainability. In our model, GCN-based collaborative filtering (CF) is parallelly performed on the user-item interaction graph and the KG to utilize the items’ additional information in the recommendation process. Also, we use a pretrained language model to generate embeddings from the user reviews and utilize them in the GCN embedding propagation process to reflect the users’ subjective sentiment and opinion. After the recommendation is performed, we generate the paths from the target user to recommended items by using the KG and embeddings from review texts. To prove the superiority of the proposed method, we conduct the experiment by comparing recommendation performance with baseline models. In the experiment, the proposed method outperformed the other models.

There are many attentions for the fact-checking to prevent the hallucination, malfunction of circulating content on the web such as deception, counterfeit, fake news regardless of whether they are intended or not. For fact-checking, ConFcheKG (Context-based Fact-checking using Knowledge Graph) is proposed based on the Knowledge Graph (KG). In the content including multiple entities, the KG is adopted to check coherence between the entities to determine whether there are semantic conflicts among them. The coherence is based on the temporal, spatial, and logical arrangement of the contents based on the associated relationships among the entities using KG. To do this, it checks the existence of intersected KGs and conflicted KGs through mutual comparison of KGs step by step. According to the verifying process, if is the constructed KT with coherent entities of the content, then the reliability of the content’s coherence is high because it has a high probability of not being false, hallucination, fake-news, and so on. Otherwise, the probability of being false increases. To check the fact, ConFcheKG can be applied based on semantic coherence of the content, to reduce the hallucination, to determine the fake news, to detect deceptions, to prevent counterfeits, to generate the well-composed prompt for the generative AI, and so on.