The sum of its parts: reducing sparsity in click estimation with query segments

The critical task of predicting clicks on search advertisements is typically addressed by learning from historical click data. When enough history is observed for a given query-ad pair, future clicks can be accurately modeled. However, based on the empirical distribution of queries, sufficient historical information is unavailable for many query-ad pairs. The sparsity of data for new and rare queries makes it difficult to accurately estimate clicks for a significant portion of typical search engine traffic. In this paper we provide analysis to motivate modeling approaches that can reduce the sparsity of the large space of user search queries. We then propose methods to improve click and relevance models for sponsored search by mining click behavior for partial user queries. We aggregate click history for individual query words, as well as for phrases extracted with a CRF model. The new models show significant improvement in clicks and revenue compared to state-of-the-art baselines trained on several months of query logs. Results are reported on live traffic of a commercial search engine, in addition to results from offline evaluation.     

Muscle metabolism and performance improvement after two training programmes of sprint running differing in rest interval duration

Repeated-sprint training often involves short sprints separated by inadequate recovery intervals. The effects of interval duration on metabolic and performance parameters are unclear. We compared the effects of two training programmes, differing in rest interval duration, on muscle (vastus lateralis) metabolism and sprint performance. Sixteen men trained three times a week for 8 weeks, each training session comprising 2-3 sets of two 80-m sprints. Sprints were separated by 10 s (n = 8) or 1 min (n = 8). Both training programmes improved performance in the 100-, 200-, and 300-m sprints, but the improvement was greater in the 10-s group during the final 100 m of the 200- and 300-m runs. Independent of interval duration, training mitigated the drop of muscle ATP after two 80-m sprints. The drop in phosphocreatine and the increases in glucose-6-phosphate and fructose-6-phosphate after two 80-m sprints were greater in the 10-s group. In conclusion, training with a limited number of repeated short sprints (≤10 s) may be more effective in improving speed maintenance in 200- and 300-m runs when performed with a 1:1 rather than a 1:6 exercise-to-rest ratio. This may be due to a greater activation of glycolysis caused, in part, by the limited resynthesis of phosphocreatine during the very short rest interval.     

Validation of MET estimates and step measurement using the ActivPAL physical activity logger

The purpose of this study was to determine the validity of the metabolic equivalent (MET) equation and step rate function of the ActivPAL? physical activity logger in a group of females. Using a standard treadmill protocol, 62 females aged 15-25 years walked on a treadmill at speeds between 3.2 and 7.0 km · h(-1) while wearing an ActivPAL. Oxygen consumption was measured using expired gas analysis at each speed and METs for each speed were estimated based on each participant's own resting metabolic rate. A sub-set of 18 participants also wore an Actigraph. Results showed that the in-built equation in the ActivPAL significantly underestimated (P < 0.001) METs under treadmill conditions at higher intensities. The ActivPAL equation is based on step rate yet the relationship between counts and measured METs (r = 0.76; P < 0.001) is stronger than that between steps and measured METs (r = 0.59; P < 0.001). Both the ActivPAL and Actigraph step functions showed no significant difference (P > 0.05) to video recorded step rate except at the slowest walking speed where the Actigraph significantly underestimated steps (P < 0.05). The development of a new equation based on the counts-METs relationship that includes a variety of speeds and activities would be useful. The ActivPAL step function performs better than the Actigraph at the slowest walking speed under treadmill conditions.     

Agreement in activity energy expenditure assessed by accelerometer and self-report in adolescents: variation by sex, age, and weight status

The agreement between self-reported and objective estimates of activity energy expenditure was evaluated in adolescents by age, sex, and weight status. Altogether, 403 participants (217 females, 186 males) aged 13-16 years completed a 3-day physical activity diary and wore a GT1M accelerometer on the same days. Partial correlations (controlling for body mass) were used to determine associations between estimated activity energy expenditure (kcal · min(-1)) from the diary and accelerometry. Differences in the magnitude of the correlations were examined using Fisher's r to z transformations. Bland-Altman procedures were used to determine concordance between the self-reported and objective estimates. Partial correlations between assessments of activity energy expenditure (kcal · min(-1)) did not differ significantly by age (13-14 years: r = 0.41; 15-16 years: r = 0.42) or weight status (normal weight: r = 0.42; overweight: r = 0.39). The magnitude of the association was significantly affected by sex (Δr = 0.11; P < 0.05). The agreement was significantly higher in males than in females. The relationship between activity energy expenditure assessed by the objective method and the 3-day diary was moderate (controlling for weight, correlations ranged between 0.33 and 0.44). However, the 3-day diary revealed less agreement in specific group analyses; it markedly underestimated activity energy expenditure in overweight/obese and older adolescents. The assessment of activity energy expenditure is complex and may require a combination of methods.     

Prosocial and antisocial behavior in sport: the role of coaching style, autonomous vs. controlled motivation, and moral disengagement

The purpose of this study was to examine whether the relationships between contextual factors (i.e., autonomy-supportive vs. controlling coaching style) and person factors (i.e., autonomous vs. controlled motivation) outlined in self-determination theory (SDT) were related to prosocial and antisocial behaviors in sport. We also investigated moral disengagement as a mediator of these relationships. Athletes' (n = 292, M = 19.53 years) responses largely supported our SDT-derived hypotheses. Results indicated that an autonomy-supportive coaching style was associated with prosocial behavior toward teammates; this relationship was mediated by autonomous motivation. Controlled motivation was associated with antisocial behavior toward teammates and antisocial behavior toward opponents, and these two relationships were mediated by moral disengagement. The results provide support for research investigating the effect of autonomy-supportive coaching interventions on athletes' prosocial and antisocial behavior.     

Are governments complying with transparency? Findings from 15 years of evaluation

How well do governments comply with their own transparency statutes? Alarmingly, answers to this question are in short supply because of patchy comparative data. This study begins to address this gap by analyzing an exhaustive compilation of 265 transparency compliance evaluations (i.e. audits) authored by NGOs, academics, and government oversight authorities across Latin America between 2003 and 2018. Aggregating data on patterns of evaluation and public sector compliance with transparency statutes, we find modest increases in compliance over time, strikingly low compliance with passive transparency compliance (i.e. governmental responses to citizen requests) at the local versus national levels of government, and significantly higher compliance scores when government oversight agencies act as evaluators. Our data also reveal worrying gaps in evaluation efforts. Evaluators tend to focus far more on active transparency (i.e. website-based disclosure) than passive transparency, more on national level governments than subnational governments, and preponderantly on the executive as opposed to legislative or judicial branches of government. Textual analyses show that education is the most evaluated policy theme, with financial policies and social services close behind. In presenting the first large-scale cross-national assessment of transparency compliance, the current study brings to light trends in the effectiveness of transparency regimes and patterns of transparency evaluation across Latin America.     

Are Elementary Preservice Teachers Floating or Sinking in Their Understanding of Buoyancy?

International Journal of Science and Mathematics Education - This research study investigated conceptions about buoyancy held by 55 elementary preservice teachers enrolled in a science methods...     

Using Drawing Technology to Assess Students’ Visualizations of Chemical Reaction Processes

In this study, we investigated how students used a drawing tool to visualize their ideas of chemical reaction processes. We interviewed 30 students using thinking-aloud and retrospective methods and provided them with a drawing tool. We identified four types of connections the students made as they used the tool: drawing on existing knowledge, incorporating dynamic aspects of chemical processes, linking a visualization to the associated chemical phenomenon, and connecting between the visualization and chemistry concepts. We also compared students who were able to create dynamic visualizations with those who only created static visualizations. The results indicated a relationship between students constructing a dynamic view of chemical reaction processes and their understanding of chemical reactions. This study provides insights into the use of visualizations to support instruction and assessment to facilitate students’ integrated understanding of chemical reactions.     

Note-taking within MetaTutor: interactions between an intelligent tutoring system and prior knowledge on note-taking and learning

Hypermedia learning environments (HLE) unevenly present new challenges and opportunities to learning processes and outcomes depending on learner characteristics and instructional supports. In this experimental study, we examined how one such HLE—MetaTutor, an intelligent, multi-agent tutoring system designed to scaffold cognitive and metacognitive self-regulated learning (SRL) processes—interacts with learner’s prior domain knowledge to affect their note-taking activities and subsequent learning outcomes. Sixty (N = 60) college students studied with MetaTutor for 120 min and took notes on hypermedia content of the human circulatory system. Log-files and screen recordings of learner-system interactions were used to analyze notes for several quantitative and qualitative variables. Results show that most note-taking was a verbatim copy of instructional content, which negatively related to the post-test measure of learning. There was an interaction between prior knowledge and pedagogical agent scaffolding, such that low prior knowledge students took a greater quantity of notes compared to their high prior knowledge counterparts, but only in the absence of MetaTutor SRL scaffolding; when agent SRL scaffolding was present, the note-taking activities of low prior knowledge students were statistically equivalent to the number of notes taken by their high prior knowledge counterparts. Theoretical and instructional design implications are discussed.     

Suicidal Behavior and Help Seeking Among Diverse College Students

Suicidal and help‐seeking behaviors of students of color remain a significant problem on college campuses. Self‐reported suicidal experiences and help‐seeking behavior of diverse students are examined on the basis of results from a national survey of college student mental health. The results suggest significant differences in the expression of suicidal thoughts and behavior across racial and ethnic groups and different experiences in their referral for, and utilization of, professional help.