Statistical learning for OCR error correction

Modern OCR engines incorporate some form of error correction, typically based on dictionaries. However, there are still residual errors that decrease performance of natural language processing algorithms applied to OCR text. In this paper, we present a statistical learning model for post-processing OCR errors, either in a fully automatic manner or followed by minimal user interaction to further reduce error rate. Our model employs web-scale corpora and integrates a rich set of linguistic features. Through an interdependent learning pipeline, our model produces and continuously refines the error detection and suggestion of candidate corrections. Evaluated on a historical biology book with complex error patterns, our model outperforms various baseline methods in the automatic mode and shows an even greater advantage when involving minimal user interaction. Quantitative analysis of each computational step further suggests that our proposed model is well-suited for handling volatile and complex OCR error patterns, which are beyond the capabilities of error correction incorporated in OCR engines.     

Evaluation of soccer players under the Moneyball concept

ABSTRACT

The recruitment of athletes with limited resources is a global problem in professional sports. In US Major League Baseball, the experience of the Oakland Athletics’ general manager in the last decade turned his “Moneyball” model into a synonym of quantitative analysis in the transfer market of baseball players. His strategy focused on hiring players with outstanding technical skills but relatively low market value. This study adapted this model to the framework of a multiple criteria decision aid (MCDA), by selecting undervalued players who have complementary abilities. The novelty here refers to the joint use of four algorithms explored by the composition of probabilistic preferences (CPP) (i.e., ranking, classification, dynamic evaluation and regularity analysis) and their application to soccer player performance evaluation. The new model analysed the recent transfer of a left-back soccer player to Europe. The results indicated 12 opportunities for better investment, among 32 left and right-back players considered. Two years later, the value of the same player was considerably lower. He played only five matches in the 2018–2019 season, without scoring or providing any assists. On the other hand, the players better classified by the CPP-MB model presented higher performances and market values.     

A biomechanical comparison of conventional classifications of bowling action-types in junior fast bowlers

ABSTRACT

Fast bowling is categorised into four action types: side-on, front-on, semi-open and mixed; however, little biomechanical comparison exists between action types in junior fast bowlers. This study investigated whether there are significant differences between action-type mechanics in junior fast bowlers. Three-dimensional kinematic and kinetic analyses were completed on 60 junior male fast bowlers bowling a five-over spell. Mixed-design factorial analyses of variance were used to test for differences between action-type groups across the phases of the bowling action. One kinetic difference was observed between groups, with a higher vertical ground reaction force loading rate during the front-foot contact phase in mixed and front-on compared to semi-open bowlers; no other significant group differences in joint loading occurred. Significant kinematic differences were observed between the front-on, semi-open and mixed action types during the front-foot contact phase for the elbow and trunk. Significant kinematic differences were also present for the ankle, T12-L1, elbow, trunk and pelvis during the back-foot phase. Overall, most differences in action types for junior fast bowlers occurred during the back-foot contact phase, particularly trunk rotation and T12-L1 joint angles/ranges of motion, where after similar movement patterns were utilized across groups during the front-foot contact phase.     

The Ultimate Survey: Asking One Question at a Time to Get Feedback from Library Users

ABSTRACT

Libraries use surveys and other instruments to gather feedback from their users. However, with competition from many other campus entities requesting input, how can survey fatigue be avoided? To gather this feedback, the Social Sciences Reference Desk utilized a short, fun and easy method. Each week for 10 weeks during the fall semester of 2017, the reference assistants wrote a survey question on a whiteboard and provided sticky notes for patrons to write their answers. The survey provided valuable information about patrons’ needs and expectations. It also fostered changes that were easily implemented to improve the delivery of library services.     

Engagement with news on Twitter: insights from Australia and Korea

ABSTRACT

This study examines the patterns of news engagement among news consumers with different political affiliation and cultural background. We use computational methods and data from Twitter in a cross-country comparison of engagement with six online news sources in Australia and South Korea. For our analysis, we used a subset of Twitter users who retweeted at least one political story during the period of collection, and for whom we were able to predict political affiliation using correspondence analysis and data on Twitter follower ties to politicians. We find that right-wing Australian retweeters are more intense in their news engagement, compared with their left-wing counterparts, whereas in South Korea it was the opposite. Australian right-wing political retweeters have more diverse information sources, while there was no difference in information diversity between the right and left in South Korea. We discuss how the political situation in South Korea at the time of data collection may have affected our analysis. We emphasise the methodological contributions of our research and its connection to on-going research into the behavioural foundations of ‘filter bubbles’.     

Has inclusion become a barrier to inclusion?

In this reflective piece the author focuses on the meaning that he made of inclusion in a book that was published almost twenty years ago. He then reflects on the meaning he makes of the current situation. He explores the problematic nature of the concept of inclusion, whether labels actually enable inclusion, research-informed implications of inclusion for those who do not experience SEND and the damaging impact of hyperaccountability and marketisation. He proposes that we may find future solutions by embracing complexities and educating for diversities.     

Meeting health information needs of people with HIV/AIDS: sources and means of collaboration

Background: Internet‐based applications, in particular those that allow communication, have great potential to meet information needs. Limited research has indicated that people with human immunodeficiency virus/acquired immune deficiency syndrome (HIV/AIDS; PHAs) use these technologies, but it has not yet been examined how resources are used collaboratively and in conjunction with offline sources. Objectives: The purpose of this study was to determine in what ways PHAs collaborate to meet treatment information needs and what role Internet‐based computer‐mediated communication (CMC) played in meeting this goal. Methods: This exploratory study was implemented using surveys and focus groups with 23 participants in Toronto, Canada. The purposive sample included men and women. Results: A variety of both off‐ and online resources were used to learn about HIV/AIDS treatment information, including web‐based and print. All participants were communicating with others, primarily in person, and most desired anecdotal treatment information. However, few reported using CMC to accomplish this goal. Harris and Dewdney's Principles of Information Seeking was used to frame the findings. Conclusions: Despite technical proficiency with CMC, few participants in this study reported use of this communication tool. Information professionals need to ensure access to HIV health information including those in remote areas who have fewer resources.     

The Big-fish–little-pond-effect Stands Up to Critical Scrutiny: Implications for Theory, Methodology, and Future Research

The big-fish–little-pond effect (BFLPE) predicts that equally able students have lower academic self-concepts (ASCs) when attending schools where the average ability levels of classmates is high, and higher ASCs when attending schools where the school-average ability is low. BFLPE findings are remarkably robust, generalizing over a wide variety of different individual student and contextual level characteristics, settings, countries, long-term follow-ups, and research designs. Because of the importance of ASC in predicting future achievement, coursework selection, and educational attainment, the results have important implications for the way in which schools are organized (e.g., tracking, ability grouping, academically selective schools, and gifted education programs). In response to Dai and Rinn (Educ. Psychol. Rev., 2008), we summarize the theoretical model underlying the BFLPE, minimal conditions for testing the BFLPE, support for its robust generalizability, its relation to social comparison theory, and recent research extending previous implications, demonstrating that the BFLPE stands up to scrutiny. Quotations (associated page numbers) to the Dai and Rinn (2008) article are based on a prepublication version of the article available to the authors of this article that may have changed during the final preparation for publication. The authors would also like to express thanks to David Dai and Anne Rinn for their encouragement and assistance to us in preparation of our article, whilst still acknowledging that they might not agree will all the views expressed here.     

索尼爱立信K750c挑战相机

这款200万像素的可拍照手机能否带来令人惊喜的照片。