Accelerated materials language processing enabled by GPT

Materials language processing (MLP) is one of the key facilitators of materials science research, as it enables the extraction of structured information from massive materials science literature. Prior works suggested high-performance MLP models for text classification, named entity recognition (NER), and extractive question answering (QA), which require complex model architecture, exhaustive fine-tuning and a large number of human-labelled datasets. In this study, we develop generative pretrained transformer (GPT)-enabled pipelines where the complex architectures of prior MLP models are replaced with strategic designs of prompt engineering. First, we develop a GPT-enabled document classification method for screening relevant documents, achieving comparable accuracy and reliability compared to prior models, with only small dataset. Secondly, for NER task, we design an entity-centric prompts, and learning few-shot of them improved the performance on most of entities in three open datasets. Finally, we develop an GPT-enabled extractive QA model, which provides improved performance and shows the possibility of automatically correcting annotations. While our findings confirm the potential of GPT-enabled MLP models as well as their value in terms of reliability and practicability, our scientific methods and systematic approach are applicable to any materials science domain to accelerate the information extraction of scientific literature.

Rate-Splitting Multiple Access for 6G Networks: Ten Promising Scenarios and Applications

In the upcoming 6G era, multiple access (MA) will play an essential role in achieving high throughput performances required in a wide range of wireless applications. Since MA and interference management are closely related issues, the conventional MA techniques are limited in that they cannot provide near-optimal performance in universal interference regimes. Recently, rate-splitting multiple access (RSMA) has been gaining much attention. RSMA splits an individual message into two parts: a common part, decodable by every user, and a private part, decodable only by the intended user. Each user first decodes the common message and then decodes its private message by applying successive interference cancellation (SIC). By doing so, RSMA not only embraces the existing MA techniques as special cases but also provides significant performance gains by efficiently mitigating inter-user interference in a broad range of interference regimes. In this article, we first present the theoretical foundation of RSMA. Subsequently, we put forth four key benefits of RSMA: spectral efficiency, robustness, scalability, and flexibility. Upon this, we describe how RSMA can enable ten promising scenarios and applications along with future research directions to pave the way for 6G.

Phase-Shift Design and Channel Modeling for Focused Beams in IRS-Assisted FSO Systems

Interest in free-space optics (FSO) is rapidly growing as a potential solution for the backhaul of next-generation mobile or low-orbit satellite communications. Various techniques have been suggested for employing an intelligent reflecting surface (IRS) in FSO systems, such as anomalous reflection, power amplification, and beam splitting. It is possible to deliver more power to the receiver (Rx) by collimating or focusing the reflected beam at the Rx lens. In this study, we propose a phase-shift design of an IRS for beam focusing. In addition, we propose a new pointing error model and an outage performance analysis applicable when the beam width is comparable to or less than the aperture size of the Rx. The analytical results are validated by Monte Carlo simulations. This study provides essential preliminary results for future researches that assume a focused beam in FSO systems.