FAIRSEQ is an open-source sequence modeling toolkit that allows researchers and developers to train custom models for translation, summarization, language modeling, and other text generation tasks. The toolkit is based on PyTorch and supports distributed training across multiple GPUs and machines. We also support fast mixed-precision training and inference on modern GPUs. A demo video can be found here: https://www.youtube. com/watch?v=OtgDdWtHvto.

We explore story generation: creative systems that can build coherent and fluent passages of text about a topic. We collect a large dataset of 300K human-written stories paired with writing prompts from an online forum. Our dataset enables hierarchical story generation, where the model first generates a premise, and then transforms it into a passage of text. We gain further improvements with a novel form of model fusion that improves the relevance of the story to the prompt, and adding a new gated multi-scale self-attention mechanism to model long-range context. Experiments show large improvements over strong baselines on both automated and human evaluations. Human judges prefer stories generated by our approach to those from a strong non-hierarchical model by a factor of two to one.

BackgroundSafety activities have been initiated at many hospitals in Taiwan, but little is known about the safety culture at these hospitals. The aims of this study were to verify a safety culture survey instrument in Chinese and to assess hospital safety culture in Taiwan.MethodsThe Taiwan Patient Safety Culture Survey was conducted in 2008, using the adapted Safety Attitude Questionnaire in Chinese (SAQ-C). Hospitals and their healthcare workers participated in the survey on a voluntary basis. The psychometric properties of the five SAQ-C dimensions were examined, including teamwork climate, safety climate, job satisfaction, perception of management, and working conditions. Additional safety measures were asked to assess healthcare workers' attitudes toward their collaboration with nurses, physicians, and pharmacists, respectively, and perceptions of hospitals' encouragement of safety reporting, safety training, and delivery delays due to communication breakdowns in clinical areas. The associations between the respondents' attitudes to each SAQ-C dimension and safety measures were analyzed by generalized estimating equations, adjusting for the clustering effects at hospital levels.ResultsA total of 45,242 valid questionnaires were returned from 200 hospitals with a mean response rate of 69.4%. The Cronbach's alpha was 0.792 for teamwork climate, 0.816 for safety climate, 0.912 for job satisfaction, 0.874 for perception of management, and 0.785 for working conditions. Confirmatory factor analyses demonstrated a good model fit for each dimension and the entire construct. The percentage of hospital healthcare workers holding positive attitude was 48.9% for teamwork climate, 45.2% for perception of management, 42.1% for job satisfaction, 37.2% for safety climate, and 31.8% for working conditions. There were wide variations in the range of SAQ-C scores in each dimension among hospitals. Compared to those without positive attitudes, healthcare workers with positive attitudes to each SAQ dimension were more likely to perceive good collaboration with coworkers, and their hospitals were more likely to encourage safety reporting and to prioritize safety training programs (Wald chi-square test, p < 0.001 for all).ConclusionsAnalytical results verified the psychometric properties of the SAQ-C at Taiwanese hospitals. The safety culture at most hospitals has not fully developed and there is considerable room for improvement.

Current models for document summarization disregard user preferences such as the desired length, style, the entities that the user might be interested in, or how much of the document the user has already read. We present a neural summarization model with a simple but effective mechanism to enable users to specify these high level attributes in order to control the shape of the final summaries to better suit their needs. With user input, our system can produce high quality summaries that follow user preferences. Without user input, we set the control variables automatically -on the full text CNN-Dailymail dataset, we outperform state of the art abstractive systems (both in terms of F1-ROUGE1 40.38 vs. 39.53 F1-ROUGE and human evaluation). User Controllable SummarizationWe introduce our summarization model and describe the control variables users can modify. Convolutional Sequence-to-SequenceOur approach builds upon the convolutional model of Gehring et al. (2017). The encoder and decoder arXiv:1711.05217v2 [cs.CL]

Writers often rely on plans or sketches to write long stories, but most current language models generate word by word from left to right. We explore coarse-to-fine models for creating narrative texts of several hundred words, and introduce new models which decompose stories by abstracting over actions and entities. The model first generates the predicate-argument structure of the text, where different mentions of the same entity are marked with placeholder tokens. It then generates a surface realization of the predicate-argument structure, and finally replaces the entity placeholders with context-sensitive names and references. Human judges prefer the stories from our models to a wide range of previous approaches to hierarchical text generation. Extensive analysis shows that our methods can help improve the diversity and coherence of events and entities in generated stories.