Ping Li - US grants

A crucial aspect of human language learning is the learner's ability to generalize existing patterns to novel instances. The issue of generalization is a focal point of current debates on mechanisms of language acquisition. In the last ten years, many researchers have used the acquisition of the English past tense as an example to debate whether language acquisition should be viewed as a symbolic, rule-based learning process or as a connectionist, statistical learning process. However, most of this debate has revolved around a specific cluster of connectionist models, the back-propagation network as a model of language acquisition. The back-propagation algorithm, especially in the context of language acquisition, has several limitations now. In this study, we explore self-organizing neural networks, in particular, the self-organizing feature maps as models of language acquisition. In contrast to back-propagation, the self-organizing network uses unsupervised learning that requires no presence of a supervisor or an explicit teacher; learning is achieved entirely by the system's self-organization in response to the input environment. Moreover, multiple self-organizing networks can be connected via Hebbian learning, a biologically motivated co-occurrence learning mechanism.

Our project involves first the development of a connectionist model of language acquisition based on principles of self-organization and Hebbian learning. It further involves the modeling of the acquisition of the lexicon and morphology, with respect to the emergence of structured lexical representations and the relationship between generalization and representation. These studies should allow us to determine (1) whether and how structured lexical representations can emerge from self-organizing processes of learning, rather than being available innately; (2) the extent to which generalization is a function of the new representation; and (3) whether and how self-organizing processes lead to the recovery from generalization errors. We propose that self-organization and Hebbian learning provide the necessary computational and psycholinguistic mechanisms for lexical representation, morphological generalization, and recovery in language acquisition. Our project will integrate previous empirical and modeling results in a new light, and offer a new theoretical perspective as well as a methodological tool for the study of the acquisition of the lexicon and the morphology.

In the last fifty years, linguists and psycholinguists have produced a large body of work on language processing in adults and language learning in children. Most of our knowledge in this domain has come from studies of Indo-European languages, in particular, English. Such results are usually interpreted as universal properties of language, and generalized accordingly. In recent years, however, there has been a surge of interest in the study of non-Indo-European languages. This interest to a certain degree reflects a 'paradigm shift', a reconceptualization of the role of cross-linguistic variation, in place of an emphasis on linguistic universals. The psycholinguistic study of Chinese represents one very important step in this direction.

The phonological, orthographic, lexical, and grammatical structures of the Chinese language differ significantly from those of Indo-European languages on which major theories of linguistics and psycholinguistics are based. One the one hand, Chinese presents a major challenge to psycholinguists who attempt to understand the dynamics of language processing and language acquisition; on the other hand, it also presents new windows on cognitive processes and new opportunities for psycholinguistic analyses. In the last two decades, researchers have used a variety of theoretical and experimental paradigms to examine Chinese language processing and its acquisition. More recently, they have used Chinese as a crucial test case, applying neural and computational approaches to examine core problems in psycholinguistics. These new approaches have not only examined the psycholinguistic processes of Chinese, but also attempted to shed new light on language processing and language acquisition in general.

The specific aim of this project is to advance our understanding of the psycholinguistic processes and mechanisms in Chinese language processing and language acquisition by organizing a workshop as part of the Linguistic Society of America (LSA) Summer Linguistic Institute 2001 (to be held at the University of California, Santa Barbara, July 13-15, 2001). The workshop draws together eminent scholars who have done pioneering work in Chinese psycholinguistics. It provides an ideal forum for researchers to disseminate and integrate their exciting new ideas and share their aspirations for future directions.

Psychology - Cognitive (73)
A shared Eye-Tracking Laboratory (ELT) has been established at the University of Richmond to advance undergraduate coursework and research. This laboratory is modeled after a similar one at the University of Chicago, but has been adapted for primary use by undergraduate students. Although eye tracking is being increasingly used for research in psychology, engineering, human factors, and education, students at primarily undergraduate institutions rarely gain experience with this advanced technology. The goal of the proposed ELT is to give students a greater understanding of advanced research methodologies in psychology, greater preparation for advanced study in a variety of related fields, and a deeper understanding of mind, brain, and eye.

The ELT enhances the curriculum of advanced research methods courses in social psychology, cognitive psychology, cognitive science, adult development, and behavioral neuroscience. This lab is being used to demonstrate prior findings, and to conduct experiments that extend earlier work. In addition, students in these advanced methods courses are learning how to collect and analyze eye-tracking data in order to investigate their own research questions. Students who have been trained in eye tracking also have the opportunity to use the ELT for independent research projects under the direction of the principal investigator or co-principal investigators. Typically, 35 psychology majors conduct independent research at the University of Richmond Psychology Department each year.

An ordinary adult speaker has active control of tens of thousands of words in any given language. Unlike a dictionary that lists words alphabetically, the mental dictionary organizes words in the mind in complex ways according to their uses in language, for example, their grammatical and semantic functions. This research will address how such organization arises in childhood, settles in adulthood, and sometimes breaks down in disordered minds. The research will provide an alternative approach to current neural network models of language, because it aims at developing a cognitively and neuropsychologically plausible model that relies on self-organizing principles. Self-organization, a dynamic process of human learning, allows the learner to gather information about the "input space" (i.e., the limits, constraints, and possibilities of things) and to continuously organize the information in ways optimal for the task at hand. Building on Li's developmental lexicon model (DevLex), the new model will incorporate properties of self-organization, Hebbian learning, lexical co-occurrence learning, and dynamic growth. These computational properties should make the model well suited for the study of the human mental lexicon, its structure, representation, and processing in children, normal adults, second language learners, and brain-injured patients. The model will attempt to account for a wide variety of phenomena in language use. In particular, its design characteristics will permit the evaluation of important problems from a number of domains: (1) the development of structurally organized representation as a function of learning the linguistic input, and the impact of the organizational structure on linguistic generalization (child language acquisition); (2) the distinct versus integrated nature of bilingual lexicon, and crosslinguistic differences in bilingual lexical representation and acquisition (bilingual language processing); (3) the development of lexical ambiguity and grammatical ambiguity, and the processing of ambiguity in patients (lexical ambiguity processing); (4) the interaction between orthography, phonology, and semantics in reading acquisition, and the crosslinguistic differences in normal reading and developmental dyslexia (normal and impaired reading); and (5) the acquisition of category-specific representation, and the structure of lesioned semantic representations in patients (category-specific language impairment). Results from the modeling of these aspects will provide significant insights into theoretical and empirical issues in psycholinguistics and cognitive science. Understanding of normal and disordered processes in different languages will also have significant implications for language education.

[unreadable] DESCRIPTION (provided by applicant): Polyhydroxyalkanoates (PHAs) are a class of microbial polyesters produced by many bacteria as intracellular carbon and energy storage polymers. They display material properties ranging from thermoplastics to elastomers. An important characteristic of PHAs is their inherent biodegradability in various environments or biosystems. Thus, the desire to replace the conventional petrochemical-based plastics with biodegradable plastics in an environmentally friendly and economically competitive fashion has served an impetus for this project. The new engineered materials are expected find wide applications such as in heart valves and as scaffolds for tissue engineering. The long-term goal of this proposal is to identify and understand the machinery involved in PHA biosynthesis and its regulation. Specifically, in order to distinguish between the two elongation mechanisms proposed for the formation of polyhydroxybutyrates (PHBs), approaches involving chemical methods, in combination with the construction of mutant enzymes, to trap the polymerization intermediates will be employed. Crystal structures will greatly enhance the understanding of the elongation mechanism and help design mechanism-based inhibitors more rationally. However, to date, no crystals of PHB synthases have been obtained although some progress has been made. A series of compounds will be prepared and explored as covalent inhibitors of synthases in an effort to obtain their crystals along with mutagenetic methods. PHAs are of general interest as they possess properties that range from thermoplastics to elastomers and are biodegradable. Understanding the PHA biosynthesis is crucial to engineering new materials that are currently being examined for their applications such as in heart valves and as scaffolds for tissue engineering. The goal of this project is to identify and understand the machinary involved in PHA biosynthesis and its regulation. [unreadable] [unreadable] [unreadable]

The ability of children to acquire their native vocabulary with such ease is truly remarkable, and one that we are only beginning to understand scientifically. In this light it is even more remarkable how easily children raised in bilingual environments can learn two different vocabularies. How are their language systems developed and organized to cope with two different sets of mental dictionaries? Cognitive scientists have addressed this question by studying the language behaviors of bilinguals through their course of development and into adulthood, and in their social and cultural contexts. They have even begun to examine the neural bases of bilingualism, but one approach that has not received much attention is the use of computational models. Meteorologists, for instance, use computational models of weather systems like hurricanes and tornadoes, not just to predict their occurrence, but to more basically understand their dynamics and underlying mechanisms. Likewise, models of cognitive systems have a rich tradition of making progress in many areas of cognitive science, yet to date, bilingualism is not one of them.

With support of the National Science Foundation, Dr. Li is developing a large-scale computational model of bilingual vocabulary development. The modeling efforts are based on self-organizing connectionist networks that have been used previously in a number of areas of language research, including vocabulary development in monolingual children. Such self-organizing models are well-suited to questions of learning and development, in that these questions are naturally cast in terms of balancing competitive and cooperative interactions among system components. In the case of bilingual vocabulary development, the components can be construed as individual words from the languages being learned. By modeling vocabulary development in this way, Dr. Li is investigating how language learning is a fundamentally incremental process whereby bilingual knowledge and skills learned later in life build upon earlier language learning. This point has important implications for bilingual education, and the debilitating effects that brain trauma or disease can have on bilingual language processing.

The ubiquitous phenomenon of massive data (including data streams) imposes considerable challenges in data visualization and exploratory data analysis. About 15 years ago, terabyte datasets were still considered `ridiculous.' However, modern datasets managed by Stanford Linear Acceleration Center (SLAC), NASA, NSA, etc. have reached the perabyte scale or larger. Corporations such as Amazon, Wal-Mart, Ebay, and search engine firms are also major generators and users of massive data. The general theme of data reduction and summarization has become an active and highly inter-disciplinary area of research. This project proposes to develop various approximation techniques, which generate a "fingerprint" or "sketch" of the massive data by transforming the original data. These `sketches' are reasonably small (hence easy to store) and can provide approximate answers which are usually good enough for practical purposes.

This proposal concerns the fundamental problems of processing/transforming massive (possibly dynamic) data. In particular, it focuses on (A) developing systematic fundamental tools for effective data reduction and efficient data summarization; (B) applying these tools to improve numerical analysis, visualization, and exploratory data analysis. Two lines of theoretically sound techniques for data reduction and summarization will be developed and further improved: (1) the method of stable random projections (SRP), effective in heavy-tailed data; (2) the method of Conditional Random Sampling (CRS), mainly for sparse data. Concrete applications of SRP and CRS will be investigated. Widely-used basic numerical algorithms can be rewritten by taking advantage of SRP or CRS. Popular methods/tools for exploratory data analysis will also benefit considerably from the development of data reduction techniques.

This PIRE project, a collaboration between three U.S. and seven foreign institutions in Europe and Asia, will investigate the cognitive and neural consequences of bilingualism to understand the ways in which multiple languages are learned and used. Recent behavioral and neuroscience evidence suggests that there is more extensive processing interaction between the two languages of a bilingual than previously thought, and this is true even when bilinguals are using only one language. Bilingual science therefore provides a tool for revealing fundamental principles about the mind and the brain otherwise obscured in research focused on monolinguals. The next stage of research on bilingualism calls for national and international collaborations to unify our understanding of the nature of the bilingual mind and brain, the process of bilingual language development, and the consequences of bilingualism for cognition. International collaboration is essential for accessibility to widely differing bilingual populations of several spoken, written, and signed languages. This award enables an international network of collaborators with common research goals and methods to exploit unique and complementary opportunities to investigate properties of human languages. Leveraging the diverse perspectives inherent in interdisciplinary and cross-cultural research will facilitate the establishment of a world-class research context for investigating bilingualism science, enable generalization of research findings, and exploit bilingualism as a tool for investigating the representation and processing of language in the mind and brain.

This PIRE project will bring together the complementary international expertise of collaborators studying bilinguals who communicate in a variety of languages (e.g., Spanish, Catalan, Welsh, and Chinese). A unique feature of this project is the partnership of U.S. and Dutch scientists exploring the consequences of bimodal bilingualism in deaf people. The NSF-funded VL2 Science of Learning Center at Gallaudet University, a world leader in education for deaf students and research on topics related to deaf people, focuses on issues of visual language processing recognizing deaf readers as bilinguals using a signed language for communication yet reading a written language. Researchers in The Netherlands also study sign language and gesture, deaf literacy development, and speech-sign translation but using different signed and written languages. The convergence of these projects provides a unique opportunity for cross-linguistic collaboration and training that would not be possible in the U.S. alone.

Enthusiasm for bilingualism research naturally draws an unusually diverse group of students, scientists, and research participants. This PIRE project will be committed to harnessing that excitement to create opportunities for broadening participation in science by research participants from a broad spectrum of ages and linguistic abilities, and by students and researchers from groups under-represented in the sciences. This PIRE project will provide training and research opportunities to students and scientists not possible without the international collaboration, such as conducting research abroad, participating in virtual international colloquia, developing and sustaining international collaborations, and training by industrial partners with specific expertise in speech, literacy, and neuroimaging. The project also provides institutional opportunities for research with diverse populations, enriching undergraduate, graduate, and post-doctoral training, and increasing opportunities for early career faculty to develop research programs globally engaged and solidly grounded in cross-disciplinary collaborations.

The nature of the science of bilingualism is inherently interdisciplinary and cross-cultural and this project provides opportunities for the participating U.S. institutions to strengthen international offices and activities, develop survey tools to evaluate student's international experiences, and provide energy and synergy for integration and for strengthening links across disciplinary units. This project will strengthen the U.S.'s scientific capital through international training not otherwise available in the U.S. U.S. institutions will benefit from attracting international visiting researchers and students to enrich the internationalizing initiatives and cultures on their campuses. The U.S. population is also increasingly bilingual with ever-diversifying demographic and cultural characteristics so research results are expected to reach well beyond academia.

U.S. project partners include The Pennsylvania State University, Gallaudet University (D.C.), and Haskins Laboratories at Yale University (CT). International partners include ESRC Centre for Research on Bilingualism in Theory and Practice, Bangor University (Bangor, UK), the Max Planck Institute for Human Cognitive and Brain Sciences (Leipzig, Germany), Universidad de Granada (Granada, Spain), Universitat Pompeu Fabra (Barcelona, Spain), Radboud University Nijmegen (Nijmegen, The Netherlands), Beijing Normal University (Beijing, China), and the University of Hong Kong (China).

This project was jointly funded by NSF's Office of International Science and Engineering and the Division of Behavioral and Cognitive Sciences.

The era of public-use micro-datasets as a cornerstone of empirical research in the social sciences is coming to an end. While it still is feasible to create such data without breaching confidentiality, scholars are pursuing research programs that mandate inherently identifiable data, such as geospatial relations, exact genome data, networks of all sorts, and linked administrative records. These researchers acquire authorized restricted access to the confidential identifiable data and perform their analyses in secure environments. The researcher is allowed to publish results that have been filtered through a statistical disclosure limitation protocol. Scientific scrutiny is hampered because the researcher cannot effectively implement a data-management plan that permits sharing these restricted-access data with other scholars. The data-custody problem is impeding the "acquire, archive, and curate" model that dominated social science data preservation in the era of public-use micro-data. This project will bridge the transition to restricted-access data and offer the scholar, the scientific community, and the custodial agency a feasible path to long-term data preservation. The Comprehensive Census Bureau Metadata Repository (CCBMR) will be a Data Documentation Initiative-based curation system designed and implemented in a manner that permits synchronization between the public and confidential versions of the repository. The scholarly community will use the CCBMR as it would use a conventional metadata repository, deprived only of the values of certain confidential information, but not their metadata. The authorized user, working on the secure Census Bureau network, will use the CCBMR with full information in authorized domains. There is no duplication of effort, and the project will implement fully automatic disclosure avoidance review of the metadata where feasible. The preservation function operates indefinitely on the original scientific inputs as long as the researchers cooperate and the agency continues to fund the preservation component of the CCBMR. Doctoral students will be taught how to develop research programs using restricted-access Census Bureau data and the repository tools developed in this project in combination with previously developed tools. The same tools will be used to develop computational statistics algorithms based on boosting to improve the integration, editing, and imputation models that assemble the micro-data used for the Census Bureau's longitudinally linked employer-employee database.

Because the Confidential Information Protection and Statistical Efficiency Act of 2002 formalized the obligation of every statistical agency in the United States to take long-term custody of the confidential micro-data used for its work, all federal statistical agencies face the same problem as the Census Bureau. The CCBMR, the education based on this repository, and the collaborative computational statistics model all can be generalized to meet the restricted-access research requirements of other statistical agencies. These tools allow statistical agencies to harness the efforts of researchers who want to understand the structure and complexity of the confidential data they intend to analyze in order to propose and implement reproducible scientific results. Future generations of scientists will be able to build on those efforts because the long-term data preservation in the CCBMR will operate on the original scientific inputs, not inputs that have been subjected to statistical disclosure limitation prior to entering the repository. This curation will result in a viable system for enforcing data management plans on projects, ensuring that results can be tested and replicated by future scientists. This activity is supported by the NSF-Census Research Network funding opportunity.

Languages differ in how they carve up the world by their labeling of objects and events. For example, the Chinese word closest to the English word "sofa" includes padded, upholstered seats for one person, while the Chinese word closest to the English word "chair" is limited to unpadded seating made of hard materials, such as wood. How do people learning two languages handle such differences? Do they develop separate ways of connecting words to the world in each language or do they learn a single and unique way that does not fully match monolinguals in either language? In this project, the investigators attempt to understand, within the broader context of language interaction in the bilingual mind, how the pattern of word use in one's first language (L1) can influence that in the second language (L2), how L1 knowledge itself can change as L2 knowledge increases, and how the fluctuating experience and knowledge of one language can create the conditions for language interactions to occur. The project will use both behavioral studies and computational modeling to explore the unique and joint contributions of a set of cognitive variables (age of exposure to each language, proficiency in each, and the type of exposure to each) to bilingual lexical knowledge.

As globalization advances, more peopleare becoming bilingual or multilingual. The study of language interaction in individuals has implications for understanding the bilingual person's verbal communication, social integration, and consequent career opportunities and may yield information useful for designing learning interventions to improve language proficiency. The proposed work will integrate research and education across the two collaborative sites (Pennsylvania State University and Lehigh University). The research also involves international collaborations between scientists in the US, Europe, and China. The cross-disciplinary nature of the project should attract students from psychology, linguistics, and cognitive and computational sciences, providing opportunities particularly to students from bilingual and bi-cultural backgrounds.

Many emerging applications of data mining call for techniques that can deal with data instances with millions, if not billions of dimensions. Hence, there is a need for effective approaches to dealing with extremely high dimensional data sets.

This project focuses on a class of novel theoretically well-founded hashing algorithms that allow high dimensional data to be encoded in a form that can be efficiently processed by standard machine learning algorithms. Specifically, it explores: One-permutation hashing, to dramatically reduce the computational and energy cost of hashing; Sparsity-preserving hashing, to take advantage of data sparsity for efficient data storage and improved generalization; Application of the new hashing techniques with standard algorithms for learning "linear" separators in high dimensional spaces. The success of this EAGER project could lay the foundations of a longer-term research agenda by the PI and other investigators focused on developing effective methods for building predictive models from extremely high dimensional data using "standard" machine learning algorithms.

Broader Impacts: Effective approaches to building predictive models from extremely high dimensional data can impact many areas of science that rely on machine learning as the primary methodology for knowledge acquisition from data. The PI's education and outreach efforts aim to broaden the participation of women and underrepresented groups. The publications, software, and datasets resulting from the project will be freely disseminated to the larger scientific community.

Numerous applications involve massive, high-dimensional datasets. For example, the search industry routinely deals with billions of web pages, where each page is often represented as a binary vector in 2^64 dimensions. In computer vision, images are often represented as non-binary vectors in millions of dimensions. Algorithms which are capable of efficiently compressing, retrieving, and mining these datasets are of high practical importance. Mathematically rigorous and computationally efficient hashing methods will be developed to dramatically reduce ultra-high-dimensional datasets. These algorithms will be integrated with a variety of learning techniques including classification, clustering, near-neighbor search, matrix factorizations, etc.

The project builds on and extends minwise hashing, and b-bit minwise hashing which are standard hashing techniques in search applications. The project aims to (i) rigorously analyze b-bit minwise hashing and develop, analyze, and apply significantly more efficient (and more accurate) to problems in search and learning; (ii) develop a unified framework of probabilistic hashing which essentially consists of one permutation followed by (at most) one random projection; (iii) develop a unified theory of summary statistics under a variety of engineering constraints (storage space, computational speed, indexing capability, adaptation to streaming, etc.).

Hashing algorithms developed under this framework are expected to be substantially much more efficient and more accurate than existing popular algorithms such as random projections and minwise hashing. This general framework allows the design algorithms to accommodate many different data types (sparse or dense data, binary or real-valued data, static or streaming data), many different engineering needs (computing inner products or lp distances, kernel learning or linear learning), and different storage requirements. Anticipated results of the proposed research include rigorous and computationally efficient hashing algorithms for dealing with ultra-high-dimensional datasets, the integration of the resulting hashing algorithms into with a variety of learning techniques for classification, clustering, near-neighbor search, singular value decompositions, matrix factorization, etc; and rigorous experimental evaluation of the resulting methods on big (e.g., TeraByte or potentially PetaByte) data of the order of up to 2^64 dimensions.

Broader Impacts: Effective approaches to building predictive models from extremely high dimensional data can impact many areas of science that rely on machine learning as the primary methodology for knowledge acquisition from data. The PI's education and outreach efforts aim to broaden the participation of women and underrepresented groups. The publications, software, and datasets resulting from the project will be freely disseminated to the larger scientific community.

Both folk wisdom and educational practices point to the benefits of study-abroad experiences for the learning of a new language. But why is language learning so much more effective when conducted in the target language environment, as compared with learning in a classroom? The proposed catalytic research project addresses this question with a neurocognitive approach by comparing two groups of learners: American students who are immersed in the second language environments (study-abroad students in Milan, Italy), and American students studying Italian in a classroom setting (in State College, Pennsylvania). This initial comparison will provide the basis for uncovering the role of learning context (immersion or no immersion). The investigators will use functional magnetic resonance imaging (fMRI) to examine the effect of learning context on how second language material is represented and processed, as compared with that of native language and language-ambiguous materials (e.g., words that could occur in both languages, such as homographs). Furthermore, the investigators will examine the impact of the learner?s individual differences in specific cognitive capacities on the successfulness of second language learning, and how such differences interact with the context of learning. These capacities, we hypothesize, include different levels of inhibitory control and working memory abilities, because the learners always need to inhibit the native language while speaking the second language and to keep track of the language being spoken. We also hypothesize that the immersion experience provides a context for more effective inhibition of their native language, thereby promoting direct mapping of new words to existing concepts for learners, especially for those with weaker control abilities. Such interactions between cognitive capacities and learning context are hypothesized to show in differential neural networks underlying bilingual performance in several key brain regions including the left prefrontal, anterior cingulate, and middle temporal cortical areas.

As our world becomes increasingly globalized, there is need for more effective cross-cultural communications via the use of multiple languages. It is thus important to understand the cognitive and neural bases of what makes second language learning successful. The proposed catalytic work provides an ideal forum for new, previously unexplored, international collaborations in the context of bilingual communities (Milan, Italy and State College, USA). It will lead to new theories and data in a rapidly developing field, the cognitive neuroscience of bilingualism that crosses the boundaries of psychology, linguistics, and neuroscience. The project also reflects our attempt to understand the bilingual mind and brain in a socially relevant context, such as the continuing social pressures faced by immigrants struggling with their second language. Increased knowledge in this domain could also have pedagogical implications for more effective foreign language teaching, for example, by providing classroom training that targets the direct connections between words and concepts (rather than second language learning through one?s native language). This project will provide further catalyst for research leading to large-scale collaborations between the investigators? institutions in the USA and Italy for longitudinal studies of second language learning in children and adults.

With the advent of Internet, numerous applications in the context of network traffic, search, and databases are faced with very large, inherently high-dimensional, or naturally streaming datasets. To effectively tackle these extremely large-scale practical problems (e.g., building statistical models from massive data, real-time network traffic monitoring and anomaly detection), methods based on statistics and probability have become increasingly popular. This proposal aims at developing theoretical, well-grounded statistical methods for massive data based on random projections, including data stream algorithms, quantized projection algorithms, and sparse projection algorithms.

Massive data are often generated as high-rate streams. Network traffic is a typical example. Effective measurements (and updates) of network traffic in real-time using small storage space are crucial for detecting anomaly events, for example the DDoS (Distributed Denial of Service) attacks. For many applications such as databases and machine learning, appropriate quantization of random projections will substantially improve the accuracies (in terms of variance per bit) and provide efficient indexing and dimension reductions to facilitate efficient search and learning. The proposed research will tackle a series of mathematically challenging problems in the development of random projections. A wide range of statistical learning and numerical linear algebra algorithms will be re-engineered to take advantage of the state-the-art projection methods.

These days, many industries such as search are in urgent demand for statistical algorithms which can effectively handle massive data. It is expected that algorithms to be developed in this proposal will be integrated with parallel platforms, to solve truly large-scale real-world problems. Research results will be disseminated to practitioners through publications, conference presentations, industry visits and collaborations, tutorials, and open-source distributions. Many of the proposed research problems involve statistical analysis and may continue to help attract statisticians/mathematicians to work on area of big data. The proposed research activities will engage both undergraduate and graduate students in statistics and engineering, through innovative curriculum and research training.

Being able to speak more than one language is key to success in a wide range of professional and academic fields. As conventional wisdom has assumed that "younger is better" for second language learning, an increasing number of children learn a second language (L2) at school, sometimes as early as Kindergarten. However, we know remarkably little about the cognitive and neural mechanisms underlying the initial stages of lexical and syntactic learning in a second language classroom setting at school, or how the processes subserving the acquisition and use of L2 knowledge change with increasing age. The currently available neurocognitive evidence is mostly based on adult second language learners, but because children are still developing their language and literacy skills in their first language, children may differ in principled ways from adults in how they integrate novel second language lexical and syntactic knowledge into their first language system. With support from the National Science Foundation, Dr. Janet van Hell and colleagues Dr. Ping Li and Dr. Darren Tanner, will use behavioral and electrophysiological measures to longitudinally study cognitive and neural mechanisms associated with the initial stages of lexical and syntactic processing in novice classroom second language learners at three ages: 5-6 years (Kindergarten), 11-12 years (6th grade), and young adulthood. The research project will also lead to an understanding of how individual differences among learners (i.e., variations in first and second language proficiency, working memory, executive control functions, and attitude/motivation) impact the rate and nature of early-stage second language learning and lexical and syntactic processing. More generally, by studying novice second language learners at three different ages the research will provide insights into the neural plasticity of language learning.

Educators, parents, business professionals, and policymakers increasingly acknowledge the importance of teaching foreign languages in US elementary, middle, and high schools. This project will contribute to much needed knowledge of how second language lexical and syntactic learning occurs at different ages, and will provide further input for more effective classroom instruction and instructional methods. Because the research team trains many female students, often from an ethnically diverse student population, the project contributes to enhancing diversity in the field of neuroscience and the planned outreach activities in the school will promote involvement by underrepresented populations in scientific inquiry.

The overall goal of this project from researchers at Pennsylvania State University is to understand the neurocognitive mechanisms underlying reading comprehension of expository scientific texts by school-aged children, adult first language readers, and adult second language readers. The proposed research integrates knowledge from several largely separate research traditions that are related to reading comprehension: (1) cognitive psychological and educational research in adult first language reading comprehension, (2) cognitive psychological and educational research in child first language reading comprehension, (3) neuroimaging research in text comprehension, and (4) graph-theoretical modeling of knowledge representation. Findings from this project will have significant implications for STEM education. It was funded by the Integrated Strategies for Understanding Neural and Cognitive Systems program, which included support from the EHR Core Research (ECR) program and the Behavioral and Cognitive Sciences division of SBE. The ECR program emphasizes fundamental STEM education research that generates foundational knowledge in the field. Investments are made in critical areas that are essential, broad and enduring: STEM learning and STEM learning environments, broadening participation in STEM, and STEM workforce development.

The research team will study the behavioral and neural patterns during the reading of science text, in an attempt to unravel the brain's text reading network underlying first and second languages, and the neurocognitive differences between good versus poor readers. It combines methods from functional magnetic resonance imaging, cognitive study of learner abilities, and advanced data-analytic techniques in cognitive modeling and brain networks. The study of brain networks through the connectivity that exists in the functional and structural pathways of the learning brain holds the promise of providing new insights into the neural bases of individual differences, neuroplasticity, and language learning and representation. Data analytics will be applied to probe into the dynamic changes in connectivity patterns. This approach will allow the study not only of learning-induced or experience-dependent neural changes, but also what brain networks characterize individual differences in learning and representation (including intrinsic neural patterns captured by resting-state functional connectivity). Observed neural changes and patterns will allow the researchers to predict who might be more successful learners.

? DESCRIPTION (provided by applicant) Polyhydroxyalkanoates (PHAs) are polyoxoesters produced by a wide range of bacteria under nutrient-limited growth conditions except for carbon. Due to their excellent biocompatibility, biodegradability, and versatility, PHAs have been developed for various biomedical applications in medical devices, drug delivery, and tissue engineering. The FDA approved the first medical use of PHAs in 2009 as an absorbable suture under the trade name TephaFLEX. However, the high cost of PHA production has been an impediment to their further development and downstream commercialization. Our goal is to identify and understand the complete PHA biosynthetic machinery so that PHAs with defined properties can be produced economically. To facilitate this, the present proposal will focus on the PHA synthase (PhaC) and phasin protein (PhaP), which are key to both PHA production and the properties of the material produced. The specific aims are: (1) to characterize the mechanism of PhaC in PHA production and control of molecular weight (MW). We will investigate chain elongation of class I synthases that are much more challenging than the class III enzymes using multiple approaches involving enzymology, molecular biology, and synthetic chemistry. Efforts will also be made to look for the additional factors that are proposed to participate in the control of PHA MWs using genetically modified organisms. Protein-protein interactions will be identified through pull-down assays for strong interactions and by incorporating photoactive unnatural amino acids for weak interactions. The MW control by PhaC itself will also be studied in vitro through a synthetic analog or in vivo through identifying the residues involved in the chain termination/re-initiation processes; (2) to obtain structural information on PHA synthases through X-ray crystallography. In collaboration with Dr. Geisbrecht who is an accomplished crystallographer on the same campus, synthases from different bacterial sources will be purified and screened for crystallization in the absence and presence of ligands. Our preliminary results of co-crystallization with a nonhydrolyzable CoA analog have provided a clear path toward an initial PhaC structure. The availability of this X-ray structure will provide us with valuable insight on substrate recognition and enzyme mechanism as well as enabling our long- term goal of protein engineering; (3) to characterize roles of PhaP in PHA production and granule formation. The relationship of PhaC and PhaP will be characterized in vitro and in vivo using various binding assays and with Escherichia coli supplemented with a PHA biosynthetic pathway. Granule formation will be monitored in vivo for the first time through a combination of fluorescence microscopy and click-chemistry. Elucidating the roles and relationships of PhaC, PhaP, additional factors and granule (PHA) formation at the molecular level is of great importance to complete our understanding of PHA production. Ultimately, this will allow PHAs with defined properties to be economically produced for medical applications. Our results will also shed light on the widespread reactions of template-independent polymerizations where the mechanism remains enigmatic.

With this award, the Chemistry of Life Processes Program in the Chemistry Division is funding Dr. Ping Li from Kansas State University and Dr. Likai Song from Florida State University to investigate the molecular mechanism of A- and B-class dye-decolorizing peroxidases (DyPs) and their structure-property-reactivity-function (SPRF) relationship. The results of this investigation can be used in the development of biocatalysts for wastewater treatment and biofuel production. The proposed studies fill the critical knowledge gap regarding DyPs and guide protein engineering that could lead to enzymes with enhanced oxidative activity. This pursuit allows the graduate students and postdoctoral trainees to acquire expertise in multidisciplinary areas spanning from Chemistry to Biology, thus preparing them to become successful next-generation scientists. Contributions are also made to an outreach program that encourages students from groups underrepresented in STEM careers to major in science.

This research project seeks to elucidate the molecular mechanism and SPRF relationship of DyPs, a new family of H2O2-dependent heme peroxidases. The studies focus on TcDyP from Thermomonospora curvata and ElDyP from Enterobacter lignolyticus as the representatives for the A- and B-class enzymes from this family, respectively. The residues situated on the enzyme surface that are responsible for substrate oxidation in TcDyP, the residue(s) that control one- and two-electron reduction reactions that involve ElDyP compound I, and the dynamics of the loop surrounding the heme pocket in DyPs are being studied using structural biology, enzymology, and electron paramagnetic resonance (EPR) investigation methods. Organic synthesis and protein engineering are being utilized to define and modulate the SPRF relationship. Information from this study not only fills the fundamental knowledge gap on DyPs but also paves the way for DyPs' applications in dye removal and lignin degradation.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.