Improving document clustering in a learned concept space

Most document clustering algorithms operate in a high dimensional bag-of-words space. The inherent presence of noise in such representation obviously degrades the performance of most of these approaches. In this paper we investigate an unsupervised dimensionality reduction technique for document clustering. This technique is based upon the assumption that terms co-occurring in the same context with the same frequencies are semantically related. On the basis of this assumption we first find term clusters using a classification version of the EM algorithm. Documents are then represented in the space of these term clusters and a multinomial mixture model (MM) is used to build document clusters. We empirically show on four document collections, Reuters-21578, Reuters RCV2-French, 20Newsgroups and WebKB, that this new text representation noticeably increases the performance of the MM model. By relating the proposed approach to the Probabilistic Latent Semantic Analysis (PLSA) model we further propose an extension of the latter in which an extra latent variable allows the model to co-cluster documents and terms simultaneously. We show on these four datasets that the proposed extended version of the PLSA model produces statistically significant improvements with respect to two clustering measures over all variants of the original PLSA and the MM models.     

From organelle to protein gel: a 6-wk laboratory project on flagellar proteins

Research suggests that undergraduate students learn more from lab experiences that involve longer-term projects. We have developed a one-semester laboratory sequence aimed at sophomore-level undergraduates. In designing this curriculum, we focused on several educational objectives: 1) giving students a feel for the scientific research process, 2) introducing them to commonly used lab techniques, and 3) building skills in both data analysis and scientific writing. Over the course of the semester, students carry out two project-based lab experiences and write two substantial lab reports modeled on primary literature. Student assessment data indicate that this lab curriculum achieved these objectives. This article describes the first of these projects, which uses the biflagellate alga Chlamydomonas reinhardtii to introduce students to the study of flagellar motility, protein synthesis, microtubule polymerization, organelle assembly, and protein isolation and characterization.     

Effects of an Educational Experience Incorporating an Inventory of Factors Potentially Influencing Student Acceptance of Biological Evolution

This investigation provides an extensive review of scientific, religious, and otherwise non-scientific factors that may influence student acceptance of biological evolution. We also measure the extent to which students' levels of acceptance changed following an educational experience designed to address an inclusive inventory of factors identified as potentially affecting student acceptance of evolution (n = 81, pre-test/post-test) n = 37, one-year longitudinal). Acceptance of evolution was measured using the Measure of Acceptance of the Theory of Evolution (MATE) instrument among participants enrolled in a secondary-level academic programme during the summer prior to their final year of high school and as they transitioned to the post-secondary level. Student acceptance of evolution was measured to be significantly higher than initial levels both immediately following and over one year after the educational experience. Results reported herein carry implications for future quantitative and qualitative research as well as for cross-disciplinary instruction plans related to evolutionary science and non-scientific factors which may influence student understanding of evolution.     

Assessing and tracking students’ problem solving performances in anchored learning environments

The purpose of this randomized experiment was to compare the performance of high-, average-, and low-achieving middle school students who were assessed with parallel versions of a computer-based test (CBT) or a paper-pencil test (PPT). Tests delivered in interactive, immersive environments like the CBT may have the advantage of providing teachers with diagnostic tools that can lead to instruction tailored to the needs of students at different achievement levels. To test the feasibility of CBT, students were randomly assigned to the CBT or PPT test conditions to measure what they had learned from an instructional method called enhanced anchored math instruction. Both assessment methods showed that students benefited from instruction and differentiated students by achievement status. The navigation maps generated from the CBT revealed that the low-achieving students were able to navigate the test, spent about the same amount of time solving the subproblems as the more advanced students, and made use of the learning scaffolds.

Submaximal exercise and maturation in 12-year-olds

The aim of this study was to examine the maturation responses of young people to submaximal treadmill exercise. Body mass was controlled using both the conventional ratio standard and allometric modelling. Ninety-seven boys and 97 girls with a mean age of 12.2 years completed a discontinuous, incremental exercise test to voluntary exhaustion. We measured peak oxygen uptake (VO2peak) and VO2 when running at 8, 9 and 10 km x h(-1). Sexual maturation was assessed visually using Tanner's indices of pubic hair. Peak VO2 was significantly higher in boys (P<0.001); this was still the case when the influence of body mass was covaried out. During submaximal exercise, no significant differences in absolute VO2 were observed between the sexes (P>0.05); however, values of VO2, expressed both in ratio with body mass and adjusted for body mass using allometry, were significantly greater in boys than in girls (P<0.001). For absolute VO2, significant main effects (P<0.05) were seen for maturity at each exercise stage. With the influence of body mass controlled using either the ratio standard or allometry, no significant main effects (P>0.05) for maturity were observed. Our results indicate that boys are less economical than girls while running at 8-10 km x h(-1) and that, independently of body mass, maturation does not influence the VO2 response to submaximal exercise.     

Differential reactions to assertive and communicative acts of toddler boys and girls

34 children were observed in infant play groups. 2 sets of infant behaviors were coded: assertive acts and attempts to communicate with adults. No sex differences were observed at 13 to 14 months in any of these behaviors. However, adults attended to girls' assertive behaviors far less of the time than to boys' assertive behaviors. They attended more to girls' less intense communication attempts and to boys' more intense attempts. When 29 of the same children were observed in toddler play groups no more than 11 months later, there were sex differences in behavior. Boys were more assertive; girls talked to teachers more. Teachers no longer differentiated their responses to boys and girls. Peers reacted more to boys' assertive behavior than to girls.' We hypothesized that caregivers may use stereotypes to guide their reactions to infants because infant behavior is ambiguous. For the toddlers, behavior had become more defined, and caregivers reacted to the behaviors. By using the sex stereotype to guide their reactions to younger children, the caregivers may have perpetuated the stereotype.     

Confirmatory factor analysis of word recognition and process measures in the Colorado Reading Project

Measures of word recognition (REC) and two component skills, phonological coding (PHON) and orthographic coding (ORTH), were subjected to multivariate behavioral genetic analysis. Data were obtained from a sample of identical and fraternal twin pairs wherein at least one member of each pair was reading disabled (RD), and from a sample of twins wherein both members of each pair read in the normal range. Confirmatory factor analysis was used to fit the genetic, common environmental, and specific environmental covariance components for REC, PHON, and ORTH within the RD and normal simples. The resulting heritability estimates for REC, PHON, and ORTH were 0.59, 0.41, and 0.05 in the RD sample, and 0.35, 0.52, and 0.20 in the normal sample. After dropping the nonsignificant common environment parameters from the models, the genetic correlations between REC and PHON and between REC and ORTH were respectively 0.81 and 0.45 in the RD sample, and 0.68 and 0.45 in the normal sample. Differences between the genetic correlations were significant in the RD sample (p<0.005), marginally significant in the normal sample (p<0.10), and highly significant in the combined sample (p<0.001), indicating that genetic influences on individual differences in REC are more strongly related to genetic variance in PHON than in ORTH. These results are consistent with previous demonstrations of substantial genetic covariance between the disabled group's deficits in REC and PHON, but not between REC and ORTH (Olson et al., 1989; Olson and Rack, 1990).     

What next? Three knowledge objectives for communication researchers

This essay identifies as knowledge objectives the need to (1) understand the interrelatedness of communicative contexts, (2) pinpoint more precisely the serial nature of communication, and (3) establish the relationship between laboratory and naturalist investigations