A New Feature Selection Scheme for Emotion Recognition from Text

This paper presents a new scheme for term selection in the field of emotion recognition from text. The proposed framework is based on utilizing moderately frequent terms during term selection. More specifically, all terms are evaluated by considering their relevance scores, based on the idea that moderately frequent terms may carry valuable information for discrimination as well. The proposed feature selection scheme performs better than conventional filter-based feature selection measures Chi-Square and Gini-Text in numerous cases. The bag-of-words approach is used to construct the vectors for document representation where each selected term is assigned the weight 1 if it exists or assigned the weight 0 if it does not exist in the document. The proposed scheme includes the terms that are not selected by Chi-Square and Gini-Text. Experiments conducted on a benchmark dataset show that moderately frequent terms boost the representation power of the term subsets as noticeable improvements are observed in terms of Accuracies.

Construct Validity of Multiple Mini Interviews – Investigating the Role of Stations, Skills, and Raters using Bayesian G-theory

One popular procedure in the medical student selection process are multiple mini-interviews (MMIs), which are designed to assess social skills (e.g., empathy) by means of brief interview and role-play stations. However, it remains unclear whether MMIs reliably measure desired social skills or rather general performance differences that do not depend on specific social skills. Here, we provide a detailed investigation into the construct validity of MMIs, including the identification and quantification of performance facets (social skill-specific performance, station-specific performance, general performance) and their relations with other selection measures. We used data from three MMI samples (N = 376 applicants, 144 raters) that included six interview and role-play stations and multiple assessed social skills. Bayesian generalizability analyses show that, the largest amount of reliable MMI variance was accounted for by station-specific and general performance differences between applicants. Furthermore, there were low or no correlations with other selection measures. Our findings suggest that MMI ratings are less social skill-specific than originally conceptualized and are due more to general performance differences (across and within-stations). Future research should focus on the development of skill-specific MMI stations and on behavioral analyses on the extents to which performance differences are based on desirable skills versus undesired aspects.

Decision tree classification: Ranking journals using IGIDI

Selection of an attribute for placement of the decision tree at an appropriate position (e.g. root of the tree) is an important decision. Many attribute selection measures such as Information Gain, Gini Index and Entropy have been developed for this purpose. The suitability of an attribute generally depends on the diversity of its values, relevance and dependency. Different attribute selection measures have different criteria for measuring the suitability of an attribute. Diversity Index is a classical statistical measure for determining the diversity of values, and according to our knowledge, it has never been used as an attribute selection method. In this article, we propose a novel attribute selection method for decision tree classification. In the proposed scheme, the average of Information Gain, Gini Index and Diversity Index are taken into account for assigning a weight to the attributes. The attribute with the highest average value is selected for the classification. We have empirically tested our proposed algorithm for classification of different data sets of scientific journals and conferences. We have developed a web-based application named JC-Rank that makes use of our proposed algorithm. We have also compared the results of our proposed technique with some existing decision tree classification algorithms.

A comparison of parameter selection measures for sensor learning from financial news events

The popularization of web platforms promoted a significant increase in the publication of financial news and reports in digital media. In this sense, a multidisciplinary research area called “learning to sense” (or sensor learning) has received attention recently. Unlike traditional machine learning methods, in sensor learning there is an interest in obtaining a time series that indicates the activity of a particular topic over time. A sensor is represented by a set of parameters learned from a historical news events dataset. The sensor generates time series as news events are processed and these time series are used in decision support systems. This paper presents an overview of sensor learning for financial news. We compared six parameter selection measures for sensor learning, with the differential of considering an unsupervised scenario. The general idea is to use the concept of k-recurrent events, i.e, news events that are similar and occur together in different periods of up-trends and down-trends of a financial time series. Thus, if a specific event (extracted from news) occurred at least k times in the past always associated with up-trends, then such news is labeled as positive news. Analogously, it can be labeled as negative. The experimental results from real data provided evidence that the approach investigated in this work is a promising alternative for sensor learning from financial news events, especially in contexts where there are no domain experts or external information to label a training set.

Evaluating the validity of the selection measures used for the UK’s foundation medical training programme: a national cohort study

ObjectivesCurrently relative performance at medical school (educational performance measure (EPM) decile), additional educational achievements and the score on a situational judgement test (SJT) are used to rank applicants to the UK Foundation Years postgraduate medical training programme. We sought to evaluate whether these three measures were predictive of subsequent successful completion of the programme, and thus were valid selection criteria.MethodsData were obtained from the UK Medical Education Database (UKMED) on 14 131 UK applicants to the foundation programme starting in 2013 and 2014. These data included training outcomes in the form of Annual Reviews of Competency Progression (ARCPs), which indicated whether the programme was successfully completed. The relationship between applicants’ performance on the three selection measures to the odds of successful programme completion were modelled.ResultsOn univariable analyses, all three measures were associated with the odds of successful completion of the programme. Converting the SJT score to deciles to compare the effect sizes suggested that one decile increase in the EPM increased the odds of completing the programme by approximately 15%, whereas the equivalent value was 8% for the SJT scores. On multivariable analyses (with all three measures included in the model), these effects were only independently and statistically significant for EPM decile (OR 1.14, 95% CI 1.10 to 1.18, p<0.001) and SJT z-score decile (OR 1.05, 95% CI 1.01 to 1.09, p=0.02).ConclusionsThe EPM decile and SJT scores may be effective selection measures for the foundation programme. However, educational achievements does not add value to the other two measures when predicting programme completion. Thus, its usefulness in this context is less clear. Moreover, our findings suggest that the weighting for the EPM decile score, relative to SJT performance, should be increased.