Citation:
This thesis investigates the problem of natural language understanding tasks such as machine reading comprehension and natural language inference, which are core problems in artificial intelligence research. Moreover, they have wide industrial applications. This thesis not only advanced the state-of-the-art across a series of well-established international benchmarks but also proposed a myriad of universal, foundational components to the field of deep learning research.

Citation:
A sponsored content policy enables a content provider to pay a mobile network operator, and thereby mobile users can access contents from the content provider through network services from the network operator with lower charge. As a result, the content provider and users are both actively engaged into the sponsored content ecosystem. This thesis aims to resolve the problem of how to provide proper sponsorship in the situation of complex tripartite interactions among the three entities, namely, the network operator, the content provider, and users. Xiong's work addresses a few urgent challenging problems regarding the impacts of network effects, information asymmetry, and edge caching in the real implementation of sponsored content. Extensive numerical simulations were conducted to illustrate some important properties of the equilibrium solutions, and confirmed the effectiveness of the proposed game models and sponsorship schemes in sponsored content.