Scoring Spoken Second Language Assessment

Achieving scores that adequately reflect the test-takers’ proficiency level, as evidenced in spoken assessment tasks, has been the subject of a large body of research in second language assessment. In this chapter, the author outlines the work that has been undertaken in relation to the scoring of spoken assessments by human raters and automated scoring. The chapter focuses on research on rater effects, rater training and feedback, rater characteristics, interlocutor/interviewer effects, rating scales, and score resolution techniques. The section on automated scoring discusses research on the underlying construct and what limits this puts on the types of tasks that can be used in the assessment. The chapter concludes by setting out some future directions for the scoring of spoken responses.

Immune challenges increase network centrality in a queenless ant

Social animals display a wide range of behavioural defences against infectious diseases, some of which increase social contacts with infectious individuals (e.g. mutual grooming), while others decrease them (e.g. social exclusion). These defences often rely on the detection of infectious individuals, but this can be achieved in several ways that are difficult to differentiate. Here, we combine non-pathogenic immune challenges with automated tracking in colonies of the clonal raider ant to ask whether ants can detect the immune status of their social partners and to quantify their behavioural responses to this perceived infection risk. We first show that a key behavioural response elicited by live pathogens (allogrooming) can be qualitatively recapitulated by immune challenges alone. Automated scoring of interactions between all colony members reveals that this behavioural response increases the network centrality of immune-challenged individuals through a general increase in physical contacts. These results show that ants can detect the immune status of their nest-mates and respond with a general ‘caring’ strategy, rather than avoidance, towards social partners that are perceived to be infectious. Finally, we find no evidence that changes in cuticular hydrocarbon profiles drive these behavioural effects.

An Artificial Intelligence-Enabled Pipeline for Medical Domain: Malaysian Breast Cancer Survivorship Cohort as a Case Study

Automated artificial intelligence (AI) systems enable the integration of different types of data from various sources for clinical decision-making. The aim of this study is to propose a pipeline to develop a fully automated clinician-friendly AI-enabled database platform for breast cancer survival prediction. A case study of breast cancer survival cohort from the University Malaya Medical Centre was used to develop and evaluate the pipeline. A relational database and a fully automated system were developed by integrating the database with analytical modules (machine learning, automated scoring for quality of life, and interactive visualization). The developed pipeline, iSurvive has helped in enhancing data management as well as to visualize important prognostic variables and survival rates. The embedded automated scoring module demonstrated quality of life of patients whereas the interactive visualizations could be used by clinicians to facilitate communication with patients. The pipeline proposed in this study is a one-stop center to manage data, to automate analytics using machine learning, to automate scoring and to produce explainable interactive visuals to enhance clinician-patient communication along the survivorship period to modify behaviours that relate to prognosis. The pipeline proposed can be modelled on any disease not limited to breast cancer.

Video captioning with stacked attention and semantic hard pull

Video captioning, i.e., the task of generating captions from video sequences creates a bridge between the Natural Language Processing and Computer Vision domains of computer science. The task of generating a semantically accurate description of a video is quite complex. Considering the complexity, of the problem, the results obtained in recent research works are praiseworthy. However, there is plenty of scope for further investigation. This paper addresses this scope and proposes a novel solution. Most video captioning models comprise two sequential/recurrent layers—one as a video-to-context encoder and the other as a context-to-caption decoder. This paper proposes a novel architecture, namely Semantically Sensible Video Captioning (SSVC) which modifies the context generation mechanism by using two novel approaches—“stacked attention” and “spatial hard pull”. As there are no exclusive metrics for evaluating video captioning models, we emphasize both quantitative and qualitative analysis of our model. Hence, we have used the BLEU scoring metric for quantitative analysis and have proposed a human evaluation metric for qualitative analysis, namely the Semantic Sensibility (SS) scoring metric. SS Score overcomes the shortcomings of common automated scoring metrics. This paper reports that the use of the aforementioned novelties improves the performance of state-of-the-art architectures.