Search

[Purpose/Significance] Deep learning for language modeling is one of the major methods and advanced technologies to enhance language intelligence of machines at present, which has become an indispensable important technical means for automatic processing and analysis of data resources, and intelligent mining of information and knowledge. However, there are still some difficulties in using deep learning for language modeling for technology development and application service in the library and information science (LIS) field. Therefore, this study systematically reviews and reveals the research progress, technical principles, and development methods of deep learning for language modeling, with the aim at providing reliable theoretical basis and feasible methodological paths for the deep understanding and application of deep learning for language modeling for librarians and fellow practitioners. [Method/Process] The data used in this study were collected from the WOS core database, CNKI literature database, arXiv preprint repository, GitHub open-source software hosting platform and the open resources on the Internet. Based on these data, this paper first systematically investigates the background, basic feature representation algorithms, and representative application development tools of deep learning for language modeling, reveals their dynamic evolution and technical principles, and analyzes the advantages and disadvantages and applicability of each algorithm model and development tool. Second, an in-depth analysis of the possible challenging problems faced by the development and application of deep learning for language modeling was performed, and two strategic approaches to expand their application capabilities were put forward. [Results/Conclusions] The important challenges faced by the application and development of deep learning for language modeling include numerous parameters and difficulties to adjust accuracy, relying on a large amount of accurate training data, difficulties in making changes, and the intellectual property and information security issues. In the future, we will start from two aspects of specific domains and feature engineering to expand and improve the application capabilities of deep learning for language modeling. Specifically, we focus on consideration of the collection and preparation of domain data, selection of model architecture, participation of domain experts, and optimization for specific tasks, in order to ensure that the data source of the model is more reliable and secure, and the application effect is more accurate and practical. Moreover, the strategic methods for feature engineering to expand the application capabilities of deep learning for language modeling include selecting appropriate features, feature pre-processing, feature selection, and feature dimensionality reduction. These strategies can help improve the performance and efficiency of deep learning for language models, making them more suitable for specific tasks or domains. To sum up, LIS institutions should leverage the deep learning for language modeling related technologies, guided by the needs of scientific research and social development, and based on advantages of existing literature data resources and knowledge services; they should carry out innovative professional or vertical domain intelligent knowledge management and application service, and develop technology and systems with independent intellectual property rights, which is their long-term sustainable development path.

Abstract Background Non‐alcoholic fatty liver disease (NAFLD) is a metabolic syndrome of the liver, and its incidence is increasing worldwide. Accumulating evidence suggests that bile acids are associated with NAFLD. Although many studies on bile acids and NAFLD have been published over the past 20 years, the authors of this study have not found a relevant bibliometric analysis in this field. Therefore, this study aimed to evaluate the trend of publications, summarize current research hotspots and predict future research directions through bibliometric analysis in this field. Method Articles related to bile acids and NAFLD published between 2002 and 2022 were obtained from the Science Citation Index‐Expanded of Web of Science Core Collection. Microsoft Excel, CiteSpace, VOSviewer and Bibliometric Online Analysis Platform were used to analyse the publication trends and research hotspots in this field. Results Among the articles published between 2002 and 2022, we retrieved 1284 articles related to bile acids and NAFLD, and finally included 568 articles. The USA was dominant until 2020, after which China surpassed the USA to become the dominant force. These two countries cooperate the most closely, and are also the most active in international cooperation. The University of California (UCL) was the most published institution, with a total of 31 publications. There were six authors who have published nine articles and ranked first. The keywords cluster labels show the 10 main clusters: #0fatty liver, #1obeticholic acid, #2oxidative stress, #37 alpha hydroxy 4 cholesten 3 one, #4deoxycholic acid, #5nonalcoholic fatty liver disease, #6mouse model, #7fibroblast growth factor 21, #8animal models, #9high‐fat diet. Keywords burst analysis revealed a higher intensity of study for the nuclear receptor, FXR, and metabolic syndrome. Conclusion Bile acids have become an important research direction in the field of NAFLD, and the intervention of gut microbiota in NAFLD by acting on bile acids may become a potential hotspot for future research. This study provides reference and guidance for future research, and will help scholars better explore the field and innovatively discover the mechanisms and treatments of NAFLD.

The X-ray spectral state transitions of Galactic black hole transients (GBHTs) are often linked to the changes in the mass accretion rate. A narrow distribution of transition luminosity in terms of the Eddington ratio has been found in previous studies of GBHTs based on RXTE data (Maccarone, 2003, A&A, 409, 697; Vahdat Motlagh et al. 2019, MNRAS, 485, 2744) and this Eddington ratio at the transition is often used in recent studies with instruments such as Swift/XRT and NICER/XTI, covering soft energies below 1 to 10 keV. However, the X-ray states characterized by spectral parameters may have different definitions depending on the energy ranges adopted in the spectral analysis, leaving the question of whether the distribution of transition luminosity obtained with RXTE remains the same when we use the instruments covering softer energy bands. In this work, we investigated the X-ray state evolutions and the variations of luminosities of eight outbursts of seven GBHTs. We found that the bolometric luminosity of the power-law component was tightly constrained to |$\sim\! 1.3\%$| Eddington luminosity at index transition when the photon index starts to decrease towards the hard state, which is consistent with the previous RXTE results (Vahdat Motlagh et al. 2019, MNRAS, 485, 2744; Kalemci et al. 2013, ApJ, 779, 95). Moreover, the tightest clustering was found to be the power-law luminosity right after the start of disk recession, with a mean logarithmic Eddington ratio of −1.84 ± 0.28. In addition, our results suggest that the disk recession starts after the bolometric disk luminosity drops below 1% Eddington luminosity. [ABSTRACT FROM AUTHOR]

We report on the evolution of X-ray spectral and timing properties of the X-ray binary MAXI J1727 − 203 based on NICER/XTI and MAXI/GSC observations. Over the course of the outburst, a transition from the intermediate state to the high/soft state, and then back to the low/hard state was observed. During the high/soft state, the innermost radius estimated from the multi-colour disc model remained constant at |$\sim 145.0\ (\frac{D}{10\ \mathrm{kpc}}) \ {(\frac{\cos i}{\cos 0^{\circ }})}^{-1/2}$| km, where D is the source distance and i is the inclination. Assuming that the binary system contain a Schwarzschild black hole and has an inclination angle of 0°–84°, and considering the typical Eddington ratio at the transition back to the low/hard state, the black hole mass was estimated to be M ≥ 11.5 M⊙ for a distance of D ≥ 5.9 kpc. We also attempted to constrain the black hole mass and distance with a different method by combining the results from optical and near-infrared photometric observations. We modelled the near-infrared to X-ray spectral energy distributions obtained in the outburst period with an irradiated disk model, and estimated the lower limit of the black hole mass for a given distance, assuming both accretion disk and companion star fill their Roche lobe. The lower limit was, however, found to be much higher than the constraint obtained from the X-ray data. We discuss several possible causes of this inconsistency. It is difficult to fully resolve the conflict by a single cause and hence a combination of causes is required. [ABSTRACT FROM AUTHOR]

We developed a high-speed image reduction pipeline using Graphics Processing Units (GPUs) as hardware accelerators. Astronomers desire to detect the emission measure counterpart of gravitational-wave sources as soon as possible and to share in the systematic follow-up observation. Therefore, high-speed image processing is important. We developed a new image-reduction pipeline for our robotic telescope system, which uses a GPU via the Python package CuPy for high-speed image processing. As a result, the new pipeline has increased in processing speed by more than 40 times compared with the current one, while maintaining the same functions. [ABSTRACT FROM AUTHOR]

Seismic signal is always contaminated by coherent and incoherent noise. Conventional incoherent noise suppression methods use the statistical difference between the signal and incoherent noise to distinguish between them. These methods are effective for random noise but are limited or even invalid for erratic noise because the latter far from satisfies the statistical assumption behind the conventional techniques. The presented study proposes a novel mathematical morphology based technique for attenuating both erratic and random noise. In this technique, 2-D mathematical morphological filtering (MMF) is implemented in each frequency slice of the 3-D seismic data cube (f – xy domain) to detect the morphological scale difference between the signal and noise. The proposed technique is compared with state-of-the-art low-rank/sparse representation based techniques by using both synthetic and real data sets. The various examples demonstrate that the proposed f – xy domain 2-D MMF method can preserve more useful signal and suppress more incoherent noise (erratic and random noise) than the low-rank and sparse representations based methods. [ABSTRACT FROM AUTHOR]

Crude glycerol is a primary by-product in the biodiesel industry. Microbial fermentation on crude glycerol for producing value-added products provides opportunities to utilize a large quantity of this by-product. This study investigates the potential of using the crude glycerol to produce vancomycin (glycopeptide antibiotics) through fermentation of Amycolatopsis orientalis XMU-VS01. The results show that crude glycerol was the most effective carbon source for mycelium growth and vancomycin production, with 40-60 g/L glycerol concentration as optimal range. Among other culture medium components, potato protein (nitrogen source) and the phosphate concentration had significant effects ( p<0.05) for vancomycin production. A Box-Behnken design and response surface methodology were employed to formulate the optimal medium. Their optimal values were determined as 52.73 g/L of glycerol, 17.36 g/L of potato protein, and 0.1 g/L of dipotassium phosphate. A highest vancomycin yield of 7.61 g/L with biomass concentration of 15.8 g/L was obtained after 120 h flask fermentation. The yield of vancomycin was 3.5 times higher than with basic medium. The results suggest that biodiesel-derived crude glycerol is a promising feedstock for production of vancomycin from A. orientalis culture. [ABSTRACT FROM AUTHOR]