Enhancing title and abstract screening for systematic reviews with GPT-3.5 turbo

The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data. 15 pages, 5 figures

Abstract screening is one important aspect of conducting a high‐quality and comprehensive systematic review and meta‐analysis. Abstract screening allows the review team to conduct the tedious but vital first step to synthesize the extant literature: winnowing down the overwhelming amalgamation of citations discovered through research databases to the citations that should be “full‐text” screened and eventually included in the review. Although it is a critical process, few guidelines have been put forth since the publications of seminal systematic review textbooks. The purpose of this paper, therefore, is to provide a practical set of best practice guidelines to help future review teams and managers. Each of the 10 proposed guidelines is explained using real‐world examples or illustrations from applications. We also delineate recent experiences where a team of abstract screeners double‐screened 14 923 abstracts in 89 days.

In the management of insomnia, physicians and patients are seeking alternative therapeutics to sleeping pills, in addition to sleep hygiene and cognitive behavioural therapy. Bright light therapy (LT) has proven its efficacy in circadian and mood disorders. We conducted a systematic literature review and meta‐analysis according to Cochrane and PRISMA guidelines and using the databases Medline, Cochrane, and Web of Science, with a special focus on light therapy and insomnia. Twenty‐two studies with a total of 685 participants were included, five of which with a high level of proof. Meta‐analysis was performed with 13 of them: light therapy for insomnia compared with control conditions significantly improved wake after sleep onset (WASO: SMD = −0.61 [−1.11, −0.11]; p = 0.017; weighted difference of 11.2 min ±11.5 based on actigraphy, and SMD = −1.09 [−1.43, −0.74] ( p < 0.001) weighted difference of −36.4 min ±15.05) based on sleep diary, but no other sleep measures such as sleep latency, total sleep time (TST), or sleep efficiency. Qualitative analysis of the review showed some improvement mainly in subjective measures. Morning light exposure advanced sleep–wake rhythms and evening exposure led to a delay. No worsening was observed in objective nor subjective measures, except for TST in one study with evening exposure. A light dose–response may exist but the studies’ heterogeneity and publication bias limit the interpretation. To conclude, light therapy shows some effectiveness for sleep maintenance in insomnia disorders, but further research is needed to refine the light parameters to be chosen according to the type of insomnia, in the hope of developing personalised therapeutics.