Recurrent Visual Relationship Recognition with Triplet Unit for Diversity

The task of visual relationship recognition (VRR) is to recognize multiple objects and their relationships in an image. A fundamental difficulty of this task is class–number scalability, since the number of possible relationships we need to consider causes combinatorial explosion. Another difficulty of this task is modeling how to avoid outputting semantically redundant relationships. To overcome these challenges, this paper proposes a novel architecture with a recurrent neural network (RNN) and triplet unit (TU). The RNN allows our model to be optimized for outputting a sequence of relationships. By optimizing our model to a semantically diverse relationship sequence, we increase the variety in output relationships. At each step of the RNN, our TU enables the model to classify a relationship while achieving class–number scalability by decomposing a relationship into a subject–predicate–object (SPO) triplet. We evaluate our model on various datasets and compare the results to a baseline. These experimental results show our model’s superior recall and precision with fewer predictions compared to the baseline, even as it produces greater variety in relationships.

Self-testing produces superior recall of both familiar and unfamiliar muscle information

Dozens of studies have found learning strategies based on the “testing effect” promote greater recall than those that rely solely on reading; however, the advantages of testing are often only observed after a delay (e.g., 2–7 days later). In contrast, our research, which has focused on kinesiology students learning kinesiology information that is generally familiar to them, has consistently demonstrated that testing-based strategies produce greater recall both immediately and after a delay. In an attempt to understand the discrepancies in the literature, the purpose of the present study was to determine if the time-related advantages of a testing-based learning strategy vary with one's familiarity with the to-be-learned information. Participants used both read-only and testing-based strategies to repeatedly study three different sets of information: 1) previously studied human muscle information (familiar information), 2) a mix of previously studied and previously unstudied human muscle information (mixed information), and 3) previously unstudied muscle information that is unique to sharks (unfamiliar information). Learning was evaluated via free recall assessments administered immediately after studying and again after a 1-wk delay and a 3-wk delay. Across those three assessments, the read-only strategy resulted in mean scores of 29.26 ± 1.43, 15.17 ± 1.29, and 5.33 ± 0.77 for the familiar, mixed, and unfamiliar information, respectively, whereas the testing-based strategy produced scores of 34.57 ± 1.58, 16.90 ± 1.31, and 8.33 ± 0.95, respectively. The results indicate that the testing-based strategy produced greater recall immediately and up through the 3-wk delay regardless of the participants' level of familiarity with the muscle information.

Mentaal Visualiseren Als Leerstrategie

This article reports an experiment in which mental imagery was used as a mnemonic strategy to enhance learners' retention of figurative idioms. Language students in tertiary education were provided with on-line exercises on 120 English idioms. Under the experimental condition, participants were presented with multiple-choice exercises in which they were asked to hypothesise about the etymological origin of the given idioms. This task was meant to elicit mental imagery. Under the control condition, participants were presented with traditional multiple-choice exercises in which they were asked to identify the correct figurative meaning of the idioms. Retention was measured one week later by means of a gap-fill exercise in which the participants were asked to produce the keywords of the idioms in context. The results of the experiment suggest (i) that mental imagery can be a powerful mnemonic strategy, and (ii) that this strategy generates superior recall, especially with regard to etymologically rather transparent figurative idioms, even though processing these may require relatively little cognitive effort.

Role of Distinctive Processing during Retrieval

Encoding tasks that engage both distinctive and organizational processing produce superior recall relative to tasks that engage only one type of processing (e.g., Hunt & Einstein, 1981). In 1993, Hunt and McDaniel proposed that organizational processing entails the relationships that specify the encoding episode, while distinctive processing allows the discrimination of particular items that share those relationships. The experiments presented here replicate previous findings from cued recall tests supporting the proposed roles for distinctive and organizational processing. Unlike previous research that generally has been limited to the use of encoding manipulations to investigate the benefits of these processes, the present experiments contrasted performance on tests of cued recall and category production. The results showed that the usefulness of distinctive processing is limited to testing situations that require production of specific items. In contrast, organization is shown to be beneficial for both cued recall and category production. These findings are consistent with the hypothesis that organization and distinctiveness serve different functions in recall and support the use of Tenpenny and Shoben's 1992 component process analysis in the investigation of retrieval.

Declarative Knowledge in Sport: A By-Product of Experience or a Characteristic of Expertise?

This research examined whether skilled sports performers’ enhanced declarative knowledge base is a by-product of experience or a characteristic of expertise. Experienced high-skill (n = 12) and low-skill (n = 12) soccer players and physically disabled spectators (n = 12) were tested on soccer recall, recognition, and anticipation ability. MANCOVA showed that high-skill players demonstrated superior anticipatory performance compared with low-skill players, who in turn were better than physically disabled spectators. ANOVA showed that high-skill players demonstrated superior recall performance on structured trials only. Also, low-skill players were significantly better than physically disabled spectators on the structured trials. MANCOVA showed that high-skill players were better at recognizing structured and unstructured trials. No differences were found between low-skill players and physically disabled spectators. It appears that high-skill players possess a larger and more elaborate declarative knowledge base. Thus, declarative knowledge is a constituent of skill rather than a by-product of experience.

Developmental Aspects of Phonemic and Visual Similarity Effects: Further Evidence in Italian Children

Phonemic and visual similarity effects were investigated in Italian children of different ages. In Experiment 1, two groups of children (mean age 5;1 and 10;3 years) were asked to recall either pictures of common objects with phonemically similar or dissimilar names, or the spoken names of the pictures. Although a similarity effect was present in older children for both words and drawings, in younger children only a tendency in the expected direction occurred. The lack of a phonemic similarity effect for spoken words was attributed to the presence of a ceiling effect. In addition, results showed a significant superior recall for words in younger children and for drawings in older ones. An additional group of 5-year-old children was tested, increasing the list length to four items. Results indicated a significant similarity effect for words but not for drawings, together with a superior recall for words. These findings, in agreement with previous results, suggest that phonological memory traces contribute to performance of younger children only when material to be recalled is in the auditory modality, whereas in older children phonological coding is independent in the input modality. In Experiment 2 the performance of 5-and 10-year-old children was compared for immediate recall of two different sets of visually similar and dissimilar drawings. Results showed a significant effect of visual similarity in younger children only, for both sets of drawings, extending previously obtained results (e.g. Hitch, Halliday, Schaaftal, & Scrhaagen, 1988) to different materials and to Italian subjects. In Experiment 3, the visual similarity effect was investigated with a delayed recall procedure in a 5-year-old group. Four delay intervals (0, 5, 10, 15 seconds) and two activities during delay (articulatory suppression and a tapping task) were considered. Results obtained indicated that the visual similarity effect is present at all delay intervals for both activities during delay; and are discussed in terms of alternative interpretations of the visual similarity effect.

Effect of Verbal Interpretation on Recall and Recognition of Ambiguous Pictorial Stimuli

66 subjects characterized as either good or poor visualizers were presented ambiguous pictorial stimuli accompanied by either a verbal interpretation or no interpretation. Both recall and recognition measures for the stimuli were obtained. Good visualizers showed superior recall and recognition compared with poor visualizers in the absence of verbal interpretation. Good and poor visualizers did not differ on either measure of memory when an interpretation accompanied the stimuli. Good visualizers performed equally well either with or without an interpretation, while poor visualizers performed significantly better with an interpretation.