Automatic retrieval of visual continuity errors in movies

Author(s):  
Lyndsey Pickup ◽  
Andrew Zisserman
2016 ◽  
Vol 23 (1) ◽  
pp. 44-55 ◽  
Author(s):  
Eleni Demetriou ◽  
Roee Holtzer

AbstractObjectives: Mild cognitive impairments (MCI) is a transitional state in aging associated with increased risk of incident dementia. The current study investigated whether MCI status moderated the effect of time on word generation during verbal fluency tasks. Specifically, the objective was to determine whether MCI status had differential effects on initial automatic or latter more effortful retrieval processes of fluency tasks. Methods: Participants were community residing older adults enrolled in a longitudinal cohort study. Of the 408 participants, 353 were normal (age=76.06±6.61; %female=57.8) and 55 were diagnosed with MCI (age=78.62±7.00; %female=52.7). Phonemic and category fluency were each administered for 60 s, but performance was recorded at three consecutive 20-s intervals (0–20 s [T1], 21–40 s [T2], 41–60 s [T3]. Separate linear mixed effects models for each fluency task were used to determine the effects of group, time, and their interaction on word generation. Results: In both fluency tasks, word generation declined as a function of time. Individuals with MCI generated fewer words compared to controls during the first 20 s of phonemic (beta=−1.56; p<.001; d=0.28) and category fluency (beta=−1.85; p<.001; d=0.37). Group by time interactions revealed that individuals with MCI demonstrated attenuated declines in word generation from the first to the second and third time intervals of both phonemic ([T1 vs. T2] beta=2.17, p=.001; d=0.41; [T1 vs. T3]beta=2.28, p=.001; d=0.45) and category ([T1 vs. T2] beta= 2.22, p=.002; d=0.50; [T1 vs. T3]beta=3.16, p<.001; d=0.71) fluency. Conclusions: Early automatic retrieval processes in verbal fluency tasks are compromised in MCI. (JINS, 2017, 23, 44–55)


2010 ◽  
Vol 16 (4) ◽  
pp. 469-491 ◽  
Author(s):  
YVES PEIRSMAN ◽  
DIRK GEERAERTS ◽  
DIRK SPEELMAN

AbstractLanguages are not uniform. Speakers of different language varieties use certain words differently – more or less frequently, or with different meanings. We argue that distributional semantics is the ideal framework for the investigation of such lexical variation. We address two research questions and present our analysis of the lexical variation between Belgian Dutch and Netherlandic Dutch. The first question involves a classic application of distributional models: the automatic retrieval of synonyms. We use corpora of two different language varieties to identify the Netherlandic Dutch synonyms for a set of typically Belgian words. Second, we address the problem of automatically identifying words that are typical of a given lect, either because of their high frequency or because of their divergent meaning. Overall, we show that distributional models are able to identify more lectal markers than traditional keyword methods. Distributional models also have a bias towards a different type of variation. In summary, our results demonstrate how distributional semantics can help research in variational linguistics, with possible future applications in lexicography or terminology extraction.


2001 ◽  
Vol 37 (3) ◽  
pp. 188 ◽  
Author(s):  
Jeong-Hun Shin ◽  
Hyun Deok Kim ◽  
Yunhee Cho ◽  
C.S. Shim ◽  
Chang-Hee Lee

2015 ◽  
Vol 10 (2) ◽  
pp. 91-98
Author(s):  
Mikhail Gorodilov ◽  
Boris Dolgovesov ◽  
Ivan Khramtsov ◽  
Aleksandr Radostev

This article is devoted to solving some of the issues of distributed multimedia data display on large screens. In particular, we consider the problem of synchronization of the process of parallel rendering and output of video fragments to the appropriate modules of multiscreen display systems. Proposed synchronization algorithm and its implementation using graphics accelerators, provides visual continuity of dynamic scenes when displayed on a multiscreen. The problems of managing distributed input data stream for multiscreen display are considered.


Author(s):  
Burasakorn Yoosooka ◽  
Vilas Wuwongse

This paper proposes a new approach to automatic retrieval and composition of Learning Objects (LOs) in an Adaptive Educational Hypermedia System (AEHS) using multidimensional learner characteristics to enhance learning effectiveness. The approach focuses on adaptive techniques in four components of AEHS: Learning Paths, LO Retrieval, LO Sequencing, and Examination Difficulty Levels. This approach has been designed to enable the adaptation of rules to become generic. Hence, the application to various domains is possible. The approach dynamically selects, sequences, and composes LOs into an individual learning package based on the use of domain ontology, learner profiles, and LO metadata. The Sharable Content Object Reference Model is employed to represent LO metadata and learning packages in order to support LO sharing. The IMS Learner Information Package Specification is used to represent learner profiles. A preliminary evaluation of the developed system indicates the system’s effectiveness in terms of learners’ satisfaction.


2012 ◽  
Vol 2 (4) ◽  
pp. 12-30
Author(s):  
Fethi Fkih ◽  
Mohamed Nazih Omri

Collocation is defined as a sequence of lexical tokens which habitually co-occur. This type of information is widely used in various applications such as Information Retrieval, document indexing, machine translation, lexicography, etc. Therefore, many techniques are developed for the automatic retrieval of collocations from textual documents. These techniques use statistical measures based on a joint frequency calculation to quantify the connection strength between the tokens of a candidate collocation. The discrimination between relevant and irrelevant collocations is performed using a priori fixed threshold. Generally, the discrimination threshold estimation is performed manually by a domain expert. This supervised estimation is considered as an additional cost which reduces system performance. In this paper, the authors propose a new technique for the threshold automatic learning to retrieve information from web text document. This technique is mainly based on the usual performance evaluation measures (such as ROC and Precision-Recall curves). The results show the ability to automatically estimate a statistical threshold independently of the treated corpus.


Sign in / Sign up

Export Citation Format

Share Document