English as a lingua franca (ELF) refers to ‘any use of English among speakers of different first languages for whom English is the communicative medium of choice, and often the only option’ (Seidlhofer, 2011, p. 7*). ELF research started relatively recently. It was only discussed occasionally in the last century. Landmark changes were the publications of Jenkins (2000*) and Seidlhofer (2001*). These works inspired more research into ELF, as witnessed by a dramatically increased interest in ELF since then, resulting in a large number of journal articles, monographs, edited books (e.g. Mauranen & Ranta, 2009*) and large corpora (e.g. the Vienna-Oxford International Corpus of English, the Corpus of English as a Lingua Franca in Academic Settings, and the Asian Corpus of English). In addition, ELF researchers have launched the annual conference series (International Conference of English as a Lingua Franca), the Journal of English as a Lingua Franca, and the De Gruyter book series Developments in English as a Lingua Franca. These publications move from an initial understanding of ELF as a ‘variety’ or ‘varieties’ to a later conceptualisation of ELF as a dynamic, fluid and variable phenomenon. ELF has become a major focus of discussions and activities among both applied linguists and English language teaching professionals (Jenkins, Cogo, & Dewey, 2011).

Twitter and other social media platforms are often used for sharing interest in products. The identification of purchase decision stages, such as in the AIDA model (Awareness, Interest, Desire, and Action), can enable more personalized e-commerce services and a finer-grained targeting of advertisements than predicting purchase intent only. In this paper, we propose and analyze neural models for identifying the purchase stage of single tweets in a user’s tweet sequence. In particular, we identify three challenges of purchase stage identification: imbalanced label distribution with a high number of non-purchase-stage instances, limited amount of training data, and domain adaptation with no or only little target domain data. Our experiments reveal that the imbalanced label distribution is the main challenge for our models. We address it with ranking loss and perform detailed investigations of the performance of our models on the different output classes. In order to improve the generalization of the models and augment the limited amount of training data, we examine the use of sentiment analysis as a complementary, secondary task in a multitask framework. For applying our models to tweets from another product domain, we consider two scenarios: for the first scenario without any labeled data in the target product domain, we show that learning domain-invariant representations with adversarial training is most promising, while for the second scenario with a small number of labeled target examples, fine-tuning the source model weights performs best. Finally, we conduct several analyses, including extracting attention weights and representative phrases for the different purchase stages. The results suggest that the model is learning features indicative of purchase stages and that the confusion errors are sensible.

Evaluation of Cr, Mn, Fe, Zn and Se in humans is challenged by the potentially high within-individual variability of these elements in biological specimens, which are poorly characterised. This study aimed to evaluate their within-day, between-day and between-month variability in spot samples, first-morning voids and 24-h collections. A total of 529 spot urine samples (including eighty-eight first-morning voids and 24-h collections) were collected from eleven Chinese adult men on days 0, 1, 2, 3, 4, 30, 60 and 90 and analysed for these five elements using inductively coupled plasma-MS. Intraclass correlation coefficients (ICC) were utilised to characterise the reproducibility, and their sensitivity and specificity were analysed to assess how well a single measurement classified individuals’ 3-month average exposures. Serial measurements of Zn in spot samples exhibited fair to good reproducibility (creatinine-adjusted ICC = 0·47) over five consecutive days, which became poor when the samples were gathered months apart (creatinine-adjusted ICC = 0·33). The reproducibility of Cr, Mn, Fe and Se in spot samples was poor over periods ranging from days to months (creatinine-adjusted ICC = 0·01–0·12). Two spot samples were sufficient for classifying 60 % of the men who truly had the highest (top 33 %) 3-month average Zn concentrations; for Cr, Mn, Fe and Se, however, at least three specimens were required to achieve similar sensitivities. In conclusion, urinary Cr, Mn, Fe, Zn and Se concentrations showed a strong within-individual variability, and a single measurement is not enough to efficiently characterise individuals’ long-term exposures.

This study proposes two multimodal frameworks to classify pathological voice samples by combining acoustic signals and medical records. In the first framework, acoustic signals are transformed into static supervectors via Gaussian mixture models; then, a deep neural network (DNN) combines the supervectors with the medical record and classifies the voice signals. In the second framework, both acoustic features and medical data are processed through first-stage DNNs individually; then, a second-stage DNN combines the outputs of the first-stage DNNs and performs classification. Voice samples were recorded in a specific voice clinic of a tertiary teaching hospital, including three common categories of vocal diseases, i.e. glottic neoplasm, phonotraumatic lesions, and vocal paralysis. Experimental results demonstrated that the proposed framework yields significant accuracy and unweighted average recall (UAR) improvements of 2.02–10.32% and 2.48–17.31%, respectively, compared with systems that use only acoustic signals or medical records. The proposed algorithm also provides higher accuracy and UAR than traditional feature-based and model-based combination methods.

This paper presents a complete two-step phase-shifting (TSPS) spectral phase interferometry for direct electric-field reconstruction (SPIDER) to improve the reconstruction of ultrafast optical fields. Here, complete TSPS acts as a balanced detection that can not only remove the effect of the dc term of the interferogram, but also reduce measurement noises, and thereby improve the capability of SPIDER to measure the pulses with narrow spectra or complex spectral structures. Some prisms are chosen to replace some environment-sensitive optical components, especially reflective optics to improve operating stability and improve signal-to-noise ratio further. Our experiments show that the available shear can be decreased to 1.5% of the spectral width, which is only about $1/3$ compared with traditional SPIDER.

In this work, carbon nanotubes (CNTs)-templated binuclear metallophthalocyanines (MTAPcCF3)2C (M = Mn, Fe, Co, Ni, Cu, Zn) assemblies (MTAPcCF3)2C–COOH–CNTs are designed and obtained. Whereafter, the structure and morphology of target products are analyzed by many means such as infrared, X-ray diffraction, X-ray photoelectron spectroscopy, and scanning electron microscopy. The electrocatalytic performances of lithium-thionyl chloride battery catalyzed by (MTAPcCF3)2C–COOH–CNTs were carried out. The result shows that all catalysts can improve the battery performance including the discharge time and the initial voltage. The catalytic performance of (MTAPcCF3)2C–COOH–CNTs is ordered following the central metal: Mn > Fe > Ni > Co > Cu > Zn. The cell capacity catalyzed by optimal catalyst (MnTAPcCF3)2C–COOH–CNTs can expand to 28.08 mAˑh, with increase by 142.07%, and the (MnTAPcCF3)2C–COOH–CNTs can extend the discharge time to 551.6 s. Besides, the reaction mechanism is presented on the basis of cyclic voltammetry measurements.

BiCuTeO is a potential thermoelectric material owing to its low thermal conductivity and high carrier concentration. However, the thermoelectric performance of BiCuTeO is still below average and has much scope for improvement. In this study, we manipulated the nominal oxygen content in BiCuTeO and synthesized BiCuTeOx (x = 0.94–1.06) bulks by a solid-state reaction and pelletized them by a cold-press method. The power factor was enhanced by varying the nominal oxygen deficiency due to the increased Seebeck coefficient. The thermal conductivity was also reduced due to the decrease in lattice thermal conductivity owing to the small grain size generated by the optimal nominal oxygen content. Consequently, the ZT value was enhanced by ∼11% at 523 K for stoichiometric BiCuTeO0.94 compared to BiCuTeO. Thus, optimal oxygen manipulation in BiCuTeO can enhance the thermoelectric performance. This study can be applied to developing oxides with high thermoelectric performances.

In this study, we investigate a new simple scheme using a planar undulator (PU) together with a properly dispersed electron beam ( $e$  beam) with a large energy spread ( ${\sim}1\%$ ) to enhance the free-electron laser (FEL) gain. For a dispersed $e$ beam in a PU, the resonant condition is satisfied for the center electrons, while the frequency detuning increases for the off-center electrons, inhibiting the growth of the radiation. The PU can act as a filter for selecting the electrons near the beam center to achieve the radiation. Although only the center electrons contribute, the radiation can be enhanced significantly owing to the high-peak current of the beam. Theoretical analysis and simulation results indicate that this method can be used for the improvement of the radiation performance, which has great significance for short-wavelength FEL applications.

Written word recognition in Chinese links the perception of individual characters with whole words. With experience in reading, a high-quality word representation can provide top-down influence on the perception of its constituent characters, thus producing a word superiority effect (WSE). In experiments using the Reicher–Wheeler paradigm, we examined the WSE in two-character words for native Chinese readers (Experiment 1) and low-proficiency adult Chinese learners with Thai (Experiment 2a) and Indonesian (Experiment 2b) as native language backgrounds. For native Chinese readers, the WSE was smaller for high-frequency than low-frequency characters, reflecting rapid access to more frequently experienced characters and a consequent reduction of top-down word-level effects. Learners of Chinese, however, showed a strong WSE for both low-frequency and high-frequency characters, reflecting less well-established character representations combined with word-level knowledge sufficient to support character recognition. The results suggest that native Chinese readers develop strong representations at both the character and the word level, while low-proficiency Chinese learners are more dependent on the word level. We discuss the possibility that a word-level emphasis Chinese foreign language instruction is one reason for this pattern.