Search

One of the tasks faced by young children is the segmentation of a continuous stream of speech into discrete linguistic units. Early in development, syllables emerge as perceptual primitives, and the wholesale storage of syllable chunks is one possible strategy for bootstrapping the segmentation process. Here, we investigate what types of chunks children store. Our method involves selecting syllabified utterances from corpora of child-directed speech, which we vary according to (a) their length in syllables, (b) the mutual predictability of their syllables, and (c) their frequency. We then use the number of utterances within which words are contained to predict the time course of word learning, arguing that utterances which perform well at this task are also more likely to be stored, by young children, as undersegmented chunks. Our results show that short utterances are best-suited for predicting when children acquire the words contained within them, although the effect is rather small. Beyond this, we also find that short utterances are the most likely to correspond to words. Together, the two findings suggest that children may not store many complete utterances as undersegmented chunks, with most of the units that children store as hypothesized words corresponding to actual words. However, dovetailing with an item-based account of language-acquisition, when children do store undersegmented chunks, these are likely to be short sequences-not frequent or internally predictable multi-word chunks. We end by discussing implications for work on formulaic multi-word sequences.

Shallow syntactic annotation in the corpus of Wrocław University of TechnologyIn this paper we present shallow syntactic annotation of The Wrocław University of Technology Corpus. We discuss some theoretical and practical considerations related to shallow parsing of Polish, then we present our annotation guidelines. The proposed annotation scheme includes chunking – four chunk types are defined with reference to the notion of accommodation and syntactic connotation, as well as annotation of four inter-chunk predicate-argument relations. Until now almost 18k chunk and 4k relation instances have been annotated. We believe that both the corpus and the annotation guideliness will prove their applicability in construction of automatic shallow parsers.

International audience; There is growing evidence that, faced with a complex environment, participants subdivide the incoming information into small perceptual units, called chunks. Although statistical properties have been identified as playing a key role in chunking, we wanted to determine whether perceptual (repetitions) and positional (initial units) features might provide immediate guidance for the parsing of information into chunks. Children aged 5 and 8 years were exposed to sequences of 3, 4, or 5 colours. Sequence learning was assessed either through an explicit generation test (Experiment 1) or through a recognition test (Experiment 2). Experiment 1 showed that perceptual and positional saliencies benefited learning and that sensitivity to repetitions was age dependent and permitted the formation of longer chunks (trigrams) in the oldest children. Experiment 2 suggested that children became sensitive to perceptual and positional saliencies regardless of age and that the both types of saliencies supported the formation of longer chunks in the oldest children. The discussion focuses on the multiple factors intervening in sequence learning and their differential effects as a function of the instructions used at test to assess sequence learning.

We report on TExSIS, a flexible bilingual terminology extraction system that uses a sophisticated chunk-based alignment method for the generation of candidate terms, after which the specificity of the candidate terms is determined by combining several statistical filters. Although the set-up of the architecture is largely language-independent, we present terminology extraction results for four different languages and three language pairs. Gold standard data sets were created for French-Italian, French-English and French-Dutch, which allowed us not only to evaluate precision, which is common practice, but also recall. We compared the TExSIS approach, which takes a multilingual perspective from the start, with the more commonly used approach of first identifying term candidates monolingually and then aligning the source and target terms. A comparison of our system with the LUIZ approach described by Vintar (2010) reveals that TExSIS outperforms LUIZ both for monolingual and bilingual terminology extraction. Our results also clearly show that the precision of the alignment is crucial for the success of the terminology extraction. Furthermore, based on the observation that the precision scores for bilingual terminology extraction outperform those of the monolingual systems, we conclude that multilingual evidence helps to determine unithood in less related languages.

The purposes of the present study were twofold. First, we sought to establish whether tonal symmetry produces processing fluency. Second, we sought to explore whether symmetry and chunk strength express themselves differently in fluency, as an indication of different mechanisms being involved for sub- and supra-finite state processing. Across two experiments, participants were asked to listen to and memorize artificial poetry showing a mirror symmetry (an inversion, i.e., a type of cross serial dependency); after this training phase, people completed a four-choice RT task in which they were presented with new artificial poetry. Participants were required to identify the stimulus displayed. We found that symmetry sped up responding to the second half of strings, indicating a fluency effect. Furthermore, there was a dissociation between fluency effects arising from symmetry vs. chunk strength, with stronger fluency effects for symmetry rather than chunks in the second half of strings. Taken together, we conjecture a divide between finite state and supra-finite state mechanisms in learning grammatical sequences.

Gamma-band activity (GBA) bursts have been viewed as a signature for ignitions in localized Hebbian cell assemblies and are thought to indicate active memory. Using whole-head magnetoencephalography, we recorded focal bursts of GBA during chess playing. Unlike highly skilled chess grandmasters, with amateur chess players focal gamma bursts prevailed in deeper structures in the region of the medial temporal lobes. This observation is consistent with the interpretation of memory formation in amateur chess players. For the frontal and parietal cortex the relationship was reversed, with more frequent gamma bursts found in chess grandmasters, indicating that the retrieval of chunks from expert memory is based on the recruitment of these neocortical areas. The results suggest the possibility that time-dependent reorganization during the formation of expert memory can be studied in humans and support the theory that the medial temporal lobe and hippocampal formation play a transitional role during the creation of expert memory in the neocortex.

Across many domains, experts make decisions based on the spatial relationships of objects within an environment. Firefighters, for example, need to evaluate the fire in front of them, radiologists the medical scan, and chess players the position on the chess board before making a decision. In order to be effective, experts need to assess these spatial relationships quickly and despite possible uncertainty. This ability is facilitated by cognitive processes. According to the so-called chunking and template theories, this expertise is made possible by pattern recognition (Gobet and Simon, 1996; Gobet, 1997). Based on their experience and practice, experts store in their long-term memory meaningful configurations of elements (i.e., patterns—chunks or templates) in their domain of expertise. These patterns are associated with typical decisions and strategies. Thus, when an expert faces a situation in their domain of expertise, they are able to rapidly recognize patterns, which, in turn, prompt them to consider decisions or strategies that have been effective in previous situations. This allows experts to make superior decisions to novices who are limited to a real-time search through a large number of possible options (Feltovich et al., 2006; Gobet and Charness, 2006; Connors et al., 2011). Leone et al.'s (2014) elegant study adds to existing evidence for this account. In particular, Leone et al. demonstrate that experts' representation of space in their domain of expertise is qualitatively different to that of novices. Leone et al. focused on chess. Chess has the advantage of being a constrained task environment with relatively high ecological validity to other domains. Chess also has the advantage of a precise measure of expertise in the form of a numerical rating that is assigned to each player on the basis of their performance against their opponents. For these reasons, a large amount of expertise research has examined chess (Gobet and Charness, 2006). For Leone et al.'s study, chess has the additional advantage of having large amounts of data available: Leone et al. examined well over 175,000 games drawn from an internet chess server. Given experts' pattern recognition ability, Leone et al. predicted that experts should make moves that were best for the position at hand and so not be constrained by the mere physical distance between objects. In contrast, novices would be more likely to make moves based on spatial proximity to previous moves due to their relative lack of pattern recognition and their need to conserve cognitive resources. Leone et al. made four specific hypotheses. First, the authors predicted that novices should make moves that are closer to their or their opponent's previous move. To test this, they developed a novel methodology in which they calculated the distance between a player's move and their previous move, and the distance between a player's move and their opponent's previous move. Second, the authors predicted that novice players would be more likely to move the same piece multiple times. To test this, they examined the frequency with which players moved the same piece more than once on consecutive turns. Third, the authors predicted that novices would be more likely to simplify the position by exchanging pieces (capturing an opponent's piece when the opponent can recapture their own piece), thereby reducing the cognitive load of dealing with large numbers of pieces. To test this, they examined the rate at which pieces were removed from the board according to the number of moves in the game. Finally, the authors predicted that novices would be less likely to keep with general strategic principles (e.g., knights are more effective when centralized) than experts. To test this, they examined the frequency with which pieces were placed in suitable parts of the board (e.g., knights in central regions). Leone et al. found evidence for all four hypotheses. First, novices were more likely to move pieces that were closer to their or their opponent's previous move than experts. Second, novices were more likely to move the same piece on consecutive turns than experts. Third, novices were more likely to simplify positions than experts. Finally, novices were less likely to keep to general strategic principles than experts. Although the relative size of the effects were small, Leone et al. demonstrated the robustness of the effects across very large datasets and across games of different durations. Similar differences between experts and novices were evident regardless of whether players had a total of 3, 5, or 15 min each. Although the authors did not directly assess the quality of the players' decisions, their work provides evidence that experts base their decision on what is most meaningful in a position, rather than being limited by spatial proximity. As Leone et al. note, these findings are consistent with chunking and template models of expertise as the differences between experts and novices exist at very short time limits, when searching through various options is not possible. The findings are also likely to generalize to a wide range of other domains of expertise. In addition to revealing these effects, the work is very important because it develops novel methodologies for assessing spatial relationships in chess positions and across large datasets. Indeed, Leone et al.'s use of big data to test specific hypotheses about cognition is particularly innovative. In future, these methods might provide further insight into how experts interpret and represent space.

Theories of expertise based on the acquisition of chunk and templates suggest a differential geometric organization of perception between experts and novices. It is implied that expert representation is less anchored by spatial (Euclidean) proximity and may instead be dictated by the intrinsic relation in the structure and grammar of the specific domain of expertise. Here we set out to examine this hypothesis. We used the domain of chess which has been widely used as a tool to study human expertise. We reasoned that the movement of an opponent piece to a specific square constitutes an external cue and the reaction of the player to this "perturbation" should reveal his internal representation of proximity. We hypothesized that novice players will tend to respond by moving a piece in closer squares than experts. Similarly, but now in terms of object representations, we hypothesized weak players will more likely focus on a specific piece and hence produce sequence of actions repeating movements of the same piece. We capitalized on a large corpus of data obtained from internet chess servers. Results showed that, relative to experts, weaker players tend to (1) produce consecutive moves in proximal board locations, (2) move more often the same piece and (3) reduce the number of remaining pieces more rapidly, most likely to decrease cognitive load and mental effort. These three principles might reflect the effect of expertise on human actions in complex setups. © 2014 Leone, Fernandez Slezak, Cecchi and Sigman. Fil: Leone, Maria Juliana. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Departamento de Física. Laboratorio de Neurociencia Integrativa; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Ciudad Universitaria. Instituto de Física de Buenos Aires. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Instituto de Física de Buenos Aires; Argentina Fil: Fernandez Slezak, Diego. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Departamento de Computación; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Ciudad Universitaria. Instituto de Física de Buenos Aires. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Instituto de Física de Buenos Aires; Argentina Fil: Cecchi, Guillermo Alberto. Thomas J. Watson Research Center; Estados Unidos Fil: Sigman, Mariano. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Ciudad Universitaria. Instituto de Física de Buenos Aires. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Instituto de Física de Buenos Aires; Argentina. Universidad de Buenos Aires. Facultad de Ciencias Exactas y Naturales. Departamento de Física. Laboratorio de Neurociencia Integrativa; Argentina. Universidad Torcuato Di Tella; Argentina

In this paper we explore a multicarrier system in the downlink that applies transmit antenna diversity for multiple users with a chunk based decision. Chunks are blocks of multiple subcarriers that allow the base station to allocate resources for multiple users experiencing a similar channel. The decision of similar channels is described only by the K factor which defines the ratio of the constant part versus the Rayleigh fading part. Higher K factors, like 10 and above, of the channel ask for different strategies as the performance degrades significantly in case of transmit imported frequency diversity. This allows to create diversity for users in Rayleigh fading channels and apply coherent addition of the received signals for users in Ricean fading channels.

International Conference of the Spanish Association of Applied Linguistics (AESLA): Language Industries and Social Change (32nd. 2014. Seville, Spain), This paper provides a window on the role of creating a self-study language course for wine-growers from a beginners' level, with the aim of gaining insights into the use of lexical chunks in language for specific purposes (LSP) teaching and learning. It focuses on describing the practical aspects, such as material design or chunk parsing, of a specific project named VinoLingua. But it also aims at going beyond our individual practice to set out the sources and grounds of the VinoLingua scientific framework and to examine the ideas about language and learning on which our approach is based: the importance of working with real language corpora and the use of standardised expressions and off-the-peg chunks of varying degrees of complexity for LSP teaching and learning. It may, therefore, serve as a model for any basic language courses designed for other professional groups., This research has been conducted within the frame of the VinoLingua Project (LLP/LDV/MP-Vinolingua-50315-LLP-1-2009-1-AT-LEONARDO-LMP), which has been funded with support from the European Commission.

Current cluster computing frameworks suffer from load imbalance and limited parallelism due to skewed data distributions, processing times, and machine speeds. We observe that the underlying cause for these issues in current systems is that they partition work statically. Hurricane is a high-performance large-scale data analytics system that successfully tames skew in novel ways. Hurricane performs adaptive work partitioning based on load observed by nodes at runtime. Overloaded nodes can spawn clones of their tasks at any point during their execution, with each clone processing a subset of the original data. This allows the system to adapt to load imbalance and dynamically adjust task parallelism to gracefully handle skew. We support this design by spreading data across all nodes and allowing nodes to retrieve data in a decentralized way. The result is that Hurricane automatically balances load across tasks, ensuring fast completion times. We evaluate Hurricane’s performance on typical analytics workloads and show that it significantly outperforms state- of-the-art systems for both uniform and skewed datasets, because it ensures good CPU and storage utilization in all cases.