Provides a thorough coverage and comparison of a wide array of time series models and methods: Exponential Smoothing, Holt Winters, ARMA and ARIMA, deep learning models including RNNs, LSTMs, GRUs, and ensemble models composed of combinations of these models. Introduces the factor table representation of ARMA and ARIMA models. This representation is not available in any other book at this level and is extremely useful in both practice and pedagogy. Uses real world examples that can be readily found via web links from sources such as the US Bureau of Statistics, Department of Transportation and the World Bank. There is an accompanying R package that is easy to use and requires little or no previous R experience. The package implements the wide variety of models and methods presented in the book and has tremendous pedagogical use.

This book introduces readers to the methods, types of data, and scale of analysis used in the context of health. The challenges of working with big data are explored throughout the book, while the benefits are also emphasized through the discoveries made possible by linking large datasets. Methods include thorough case studies from statistics, as well as the newest facets of data analytics: data visualization, modeling and simulation, and machine learning. The diversity of datasets is illustrated through chapters on networked data, image processing, and text, in addition to typical structured numerical datasets. While the methods, types of data, and scale have been individually covered elsewhere, by bringing them all together under one "umbrella" the book highlights synergies, while also helping scholars fluidly switch between tools as needed. New challenges and emerging frontiers are also discussed, helping scholars grasp how methods will need to change in response to the latest challenges in health.

Optimization techniques are at the core of data science, including data analysis and machine learning. An understanding of basic optimization techniques and their fundamental properties provides important grounding for students, researchers, and practitioners in these areas. This text covers the fundamentals of optimization algorithms in a compact, self-contained way, focusing on the techniques most relevant to data science. An introductory chapter demonstrates that many standard problems in data science can be formulated as optimization problems. Next, many fundamental methods in optimization are described and analyzed, including: gradient and accelerated gradient methods for unconstrained optimization of smooth (especially convex) functions; the stochastic gradient method, a workhorse algorithm in machine learning; the coordinate descent approach; several key algorithms for constrained optimization problems; algorithms for minimizing nonsmooth functions arising in data science; foundations of the analysis of nonsmooth functions and optimization duality; and the back-propagation approach, relevant to neural networks.

Human error is implicated in nearly all aviation accidents, yet most investigation and prevention programs are not designed around any theoretical framework of human error. Appropriate for all levels of expertise, the book provides the knowledge and tools required to conduct a human error analysis of accidents, regardless of operational setting (i.e. military, commercial, or general aviation). The book contains a complete description of the Human Factors Analysis and Classification System (HFACS), which incorporates James Reason's model of latent and active failures as a foundation. Widely disseminated among military and civilian organizations, HFACS encompasses all aspects of human error, including the conditions of operators and elements of supervisory and organizational failure. It attracts a very broad readership. Specifically, the book serves as the main textbook for a course in aviation accident investigation taught by one of the authors at the University of Illinois. This book will also be used in courses designed for military safety officers and flight surgeons in the U.S. Navy, Army and the Canadian Defense Force, who currently utilize the HFACS system during aviation accident investigations. Additionally, the book has been incorporated into the popular workshop on accident analysis and prevention provided by the authors at several professional conferences world-wide. The book is also targeted for students attending Embry-Riddle Aeronautical University which has satellite campuses throughout the world and offers a course in human factors accident investigation for many of its majors. In addition, the book will be incorporated into courses offered by Transportation Safety International and the Southern California Safety Institute. Finally, this book serves as an excellent reference guide for many safety professionals and investigators already in the field.

Statistical and machine learning methods have many applications in the environmental sciences, including prediction and data analysis in meteorology, hydrology and oceanography, pattern recognition for satellite images from remote sensing, management of agriculture and forests, assessment of climate change, and much more. With rapid advances in machine learning in the last decade, this book provides an urgently needed, comprehensive guide to machine learning and statistics for students and researchers interested in environmental data science. It includes intuitive explanations covering the relevant background mathematics, with examples drawn from the environmental sciences. A broad range of topics are covered, including correlation, regression, classification, clustering, neural networks, random forests, boosting, kernel methods, evolutionary algorithms, and deep learning, as well as the recent merging of machine learning and physics. End-of-chapter exercises allow readers to develop their problem-solving skills and online data sets allow readers to practise analysis of real data.

Throughout the physical and social sciences, researchers face the challenge of fitting statistical distributions to their data. Although the study of statistical modelling has made great strides in recent years, the number and variety of distributions to choose from-all with their own formulas, tables, diagrams, and general properties-continue to create problems. For a specific application, which of the dozens of distributions should one use? What if none of them fit well?

Fitting Statistical Distributions helps answer those questions. Focusing on techniques used successfully across many fields, the authors present all of the relevant results related to the Generalized Lambda Distribution (GLD), the Generalized Bootstrap (GB), and Monte Carlo simulation (MC). They provide the tables, algorithms, and computer programs needed for fitting continuous probability distributions to data in a wide variety of circumstances-covering bivariate as well as univariate distributions, and including situations where moments do not exist.

Regardless of your specific field-physical science, social science, or statistics, practitioner or theorist-Fitting Statistical Distributions is required reading. It includes wide-ranging applications illustrating the methods in practice and offers proofs of key results for those involved in theoretical development. Without it, you may be using obsolete methods, wasting time, and risking incorrect results.

The construction industry is vital to any national economy; it is also one of the industries most susceptible to workplace incidents. The unacceptably high rates of incidents in construction have huge socio-economic consequences for the victims, their families and friends, co-workers, employers and society at large. Construction safety researchers have introduced numerous strategies, models and tools through scientific inquiries involving primary data collection and analyses. While these efforts are commendable, there is a huge potential to create new knowledge and predictive models to improve construction safety by utilising already existing data about workplace incidents. In this new book, Imriyas Kamardeen argues that more sophisticated approaches need to be deployed to enable improved analyses of incident data sets and the extraction of more valuable insights, patterns and knowledge to prevent work injuries and illnesses. The book aims to apply data mining and analytic techniques to past workplace incident data to discover patterns that facilitate the development of innovative models and strategies, thereby improving work health, safety and well-being in construction, and curtailing the high rate of incidents. It is essential reading for researchers and professionals in construction, health and safety and anyone interested in data analytics.

This is the first comprehensive overview of the 'science of science,' an emerging interdisciplinary field that relies on big data to unveil the reproducible patterns that govern individual scientific careers and the workings of science. It explores the roots of scientific impact, the role of productivity and creativity, when and what kind of collaborations are effective, the impact of failure and success in a scientific career, and what metrics can tell us about the fundamental workings of science. The book relies on data to draw actionable insights, which can be applied by individuals to further their career or decision makers to enhance the role of science in society. With anecdotes and detailed, easy-to-follow explanations of the research, this book is accessible to all scientists and graduate students, policymakers, and administrators with an interest in the wider scientific enterprise.

The global data market is estimated to be worth $64 billion dollars, making it a more valuable resource than oil. But data is useless without the analysis, interpretation and innovations of data scientists. With Confident Data Science, learn the essential skills and build your confidence in this sector through key insights and practical tools for success. In this book, you will discover all of the skills you need to understand this discipline, from primers on the key analytic and visualization tools to tips for pitching to and working with clients. Adam Ross Nelson draws upon his expertise as a data science consultant and, as someone who made moved into the industry late in his career, to provide an overview of data science, including its key concepts, its history and the knowledge required to become a successful data scientist. Whether you are considering a career in this industry or simply looking to expand your knowledge, Confident Data Science is the essential guide to the world of data science. About the Confident series... From coding and data science to cloud and cyber security, the Confident books are perfect for building your technical knowledge and enhancing your professional career.

1 Reliability: Past, Present, Future.- 2 Reliability Analysis as a Tool for Expressing and Communicating Uncertainty.- 3 Modeling a Process of Non-Ideal Repair.- 4 Some Models and Mathematical Results for Reliability of Systems of Components.- 5 Algorithms of Stochastic Activity and Problems of Reliability.- 6 Some Shifted Stochastic Orders.- 7 Characterization of Distributions in Reliability.- 8 Asymptotic Analysis of Reliability for Switching Systems in Light and Heavy Traffic Conditions.- 9 Nonlinearly Perturbed Markov Chains and Large Deviations for Lifetime Functionals.- 10 Evolutionary Systems in an Asymptotic Split Phase Space.- 11 An Asymptotic Approach to Multistate Systems Reliability Evaluation.- 12 Computer Intensive Methods Based on Resampling in Analysis of Reliability and Survival Data.- 13 Statistical Analysis of Damage Processes.- 14 Data Analysis Based on Warranty Database.- 15 Failure Models Indexed by Time and Usage.- 16 A New Multiple Proof Loads Approach For Estimating Correlations.- 17 Conditional and Partial Correlation For Graphical Uncertainty Models.- 18 Semiparametric Methods of Time Scale Selection.- 19 Censored and Truncated Lifetime Data.- 20 Tests for a Family of Survival Models Based on Extremes.- 21 Software Reliability Models - Past, Present and Future.- 22 Dynamic Analysis of Failures in Repairable Systems and Software.- 23 Precedence Test and Maximal Precedence Test.- 24 Hierarchical Bayesian Inference in Related Reliability Experiments.- 25 Tests for Equality of Intensities of Failures of a Repairable System Under Two Competing Risks.- 26 Semiparametric Estimation in Accelerated Life Testing.- 27 A Theoretical Framework for Accelerated Testing.- 28 Unbiased Estimation in Reliability and Similar Problems.- 29 Prediction Under Association.- 30 Uniform Limit Laws for Kernel Density Estimators on Possibly Unbounded Intervals.- 31 A Weak Convergence Result Relevant in Recurrent and Renewal Models.

THE NEW YORK TIMES BESTSELLER

AN ECONOMIST BOOK OF THE YEAR 2017

Insightful, surprising and with ground-breaking revelations about our society, Everybody Lies exposes the secrets embedded in our internet searches, with a foreword by bestselling author Steven Pinker

Everybody lies, to friends, lovers, doctors, pollsters - and to themselves. In Internet searches, however, people confess their secrets - about sexless marriages, mental health problems, even racist views. Seth Stephens-Davidowitz, an economist and former Google data scientist, shows that this could just be the most important dataset ever collected.

This huge database of secrets - unprecedented in human history - offers astonishing, even revolutionary, insights into humankind. Anxiety, for instance, does not increase after a terrorist attack. Crime levels drop when a violent film is released. And racist searches are no higher in Republican areas than in Democrat ones.

Stephens-Davidowitz reveals information we can use to change our culture, and the questions we're afraid to ask that might be essential to our health - both emotional and physical. Insightful, funny, and always surprising, Everybody Lies exposes the biases and secrets embedded deeply within us, at a time when things are harder to predict than ever.

How data-informed decision making can make colleges and universities more effective institutions. The continuing importance of data analytics is not lost on higher education leaders, who face a multitude of challenges, including increasing operating costs, dwindling state support, limits to tuition increases, and increased competition from the for-profit sector. To navigate these challenges, savvy leaders must leverage data to make sound decisions. In Big Data on Campus, leading data analytics experts and higher ed leaders show the role that analytics can play in the better administration of colleges and universities. Aimed at senior administrative leaders, practitioners of institutional research, technology professionals, and graduate students in higher education, the book opens with a conceptual discussion of the roles that data analytics can play in higher education administration. Subsequent chapters address recent developments in technology, the rapid accumulation of data assets, organizational maturity in building analytical capabilities, and methodological advancements in developing predictive and prescriptive analytics. Each chapter includes a literature review of the research and application of analytics developments in their respective functional areas, a discussion of industry trends, examples of the application of data analytics in their decision process, and other related issues that readers may wish to consider in their own organizational environment to find opportunities for building robust data analytics capabilities. Using a series of focused discussions and case studies, Big Data on Campus helps readers understand how analytics can support major organizational functions in higher education, including admission decisions, retention and enrollment management, student life and engagement, academic and career advising, student learning and assessment, and academic program planning. The final section of the book addresses major issues and human factors involved in using analytics to support decision making; the ethical, cultural, and managerial implications of its use; the role of university leaders in promoting analytics in decision making; and the need for a strong campus community to embrace the analytics revolution. Contributors: Rana Glasgal, J. Michael Gower, Tom Gutman, Brian P. Hinote, Braden J. Hosch, Aditya Johri, Christine M. Keller, Carrie Klein, Jaime Lester, Carrie Hancock Marcinkevage, Gail B. Marsh, Susan M. Menditto, Jillian N. Morn, Valentina Nestor, Cathy O'Bryan, Huzefa Rangwala, Timothy Renick, Charles Tegen, Rachit Thariani, Chris Tompkins, Lindsay K. Wayt, Karen L. Webber, Henry Y. Zheng, Ying Zhou

In recent years, popular media have inundated audiences with sensationalised headlines recounting data breaches, new forms of surveillance and other dangers of our digital age. Despite their regularity, such accounts treat each case as unprecedented and unique. This book proposes a radical rethinking of the history, present and future of our relations with the digital, spatial technologies that increasingly mediate our everyday lives. From smartphones to surveillance cameras, to navigational satellites, these new technologies offer visions of integrated, smooth and efficient societies, even as they directly conflict with the ways users experience them. Recognising the potential for both control and liberation, the authors argue against both acquiescence to and rejection of these technologies. Through intentional use of the very systems that monitor them, activists from Charlottesville to Hong Kong are subverting, resisting and repurposing geographic technologies. Using examples as varied as writings on the first telephones to the experiences of a feminist collective for migrant women in Spain, the authors present a revolution of everyday technologies. In the face of the seemingly inevitable dominance of corporate interests, these technologies allow us to create new spaces of affinity, and a new politics of change.

Just Plain Data Analysis teaches students statistical literacy skills that they can use to evaluate and construct arguments about public affairs issues grounded in numerical evidence. The book addresses skills that are often not taught in introductory social science research methods courses and that are often covered sketchily in the research methods textbooks: where to find commonly used measures of political and social conditions; how to assess the reliability and validity of specific indicators; how to present data efficiently in charts and tables; how to avoid common misinterpretations and misrepresentations of data; and how to evaluate causal arguments based on numerical data. With a new chapter on statistical fallacies and updates throughout the text, the new edition teaches students how to find, interpret, and present commonly used social indicators in an even clearer and more practical way.

As a result of the COVID-19 pandemic, medical statistics and public health data have become staples of newsfeeds worldwide, with infection rates, deaths, case fatality and the mysterious R figure featuring regularly. However, we don't all have the statistical background needed to translate this information into knowledge. In this lively account, Stephen Senn explains these statistical phenomena and demonstrates how statistics is essential to making rational decisions about medical care. The second edition has been thoroughly updated to cover developments of the last two decades and includes a new chapter on medical statistical challenges of COVID-19, along with additional material on infectious disease modelling and representation of women in clinical trials. Senn entertains with anecdotes, puzzles and paradoxes, while tackling big themes including: clinical trials and the development of medicines, life tables, vaccines and their risks or lack of them, smoking and lung cancer, and even the power of prayer.

Achieve successful digital transformation with this authoritative guide designed specifically for established organizations. At a time where even the most recognized business models are under threat, organizations risk devastation if they do not transition successfully to the new digital reality. Yet what works for digital natives does not always work for established organizations. Recognized as one of the world's top global executives leading innovative transformation, Neetan Chopra's deep experience of steering organizations through digital disruption drives the practical approach of Accelerated Digital Transformation. Having designed transformation journeys, overcome setbacks and driven outcomes within multiple leading companies, Neetan Chopra tackles key factors for established organizations including inertia, impetus, outcomes, digital capabilities and culture. The book is underpinned by a tried and tested framework that will guide readers step by step through the entire digital transformation journey. This will be an essential resource for leaders, managers and practitioners leading and executing digital transformation.

This is a book about the scientific process and how you apply it to data in ecology. You will learn how to plan for data collection, how to assemble data, how to analyze data and finally how to present the results. The book uses Microsoft Excel and the powerful Open Source R program to carry out data handling as well as producing graphs. Statistical approaches covered include: data exploration; tests for difference - t-test and U-test; correlation - Spearman's rank test and Pearson product-moment; association including Chi-squared tests and goodness of fit; multivariate testing using analysis of variance (ANOVA) and Kruskal-Wallis test; and multiple regression. Key skills taught in this book include: how to plan ecological projects; how to record and assemble your data; how to use R and Excel for data analysis and graphs; how to carry out a wide range of statistical analyses including analysis of variance and regression; how to create professional looking graphs; and how to present your results. New in this edition: a completely revised chapter on graphics including graph types and their uses, Excel Chart Tools, R graphics commands and producing different chart types in Excel and in R; an expanded range of support material online, including; example data, exercises and additional notes & explanations; a new chapter on basic community statistics, biodiversity and similarity; chapter summaries and end-of-chapter exercises. Praise for the first edition: This book is a superb way in for all those looking at how to design investigations and collect data to support their findings. - Sue Townsend, Biodiversity Learning Manager, Field Studies Council [M]akes it easy for the reader to synthesise R and Excel and there is extra help and sample data available on the free companion webpage if needed. I recommended this text to the university library as well as to colleagues at my student workshops on R. Although I initially bought this book when I wanted to discover R I actually also learned new techniques for data manipulation and management in Excel - Mark Edwards, EcoBlogging A must for anyone getting to grips with data analysis using R and excel. - Amazon 5-star review It has been very easy to follow and will be perfect for anyone. - Amazon 5-star review A solid introduction to working with Excel and R. The writing is clear and informative, the book provides plenty of examples and figures so that each string of code in R or step in Excel is understood by the reader. - Goodreads, 4-star review

Social media has put mass communication in the hands of normal people on an unprecedented scale, and has also given social scientists the tools necessary to listen to the voices of everyday people around the world. This book gives social scientists the skills necessary to leverage that opportunity, and transform social media's vast stream of information into social science data. The book combines the big data techniques of computer science with social science methodology. Intended as a text for advanced undergraduates, graduate students, and researchers in the social sciences, this book provides a methodological pathway for scholars who want to make use of this new and evolving source of data. It provides a framework for building one's own data collection and analysis infrastructure, a toolkit of content analysis, geographic analysis, and network analysis, and meditations on the ethical implications of social media data.

THE NEW BOOK FROM THE BESTSELLING AUTHOR OF EVERYBODY LIES 'Don't Trust Your Gut is a tour de force - an intoxicating blend of analysis, humor, and humanity' DANIEL H. PINK 'Seth Stephens-Davidowitz is an expert on data-driven thinking, and this engaging book is full of surprising, useful insights for using the information at your fingertips to make better decisions' ADAM GRANT Big decisions are hard. We might consult friends and family, read advice online or turn to self-help books for guidance, but in the end we usually just do what feels right. But what if our gut is wrong? As economist and former Google data scientist Seth Stephens-Davidowitz argues, our gut is actually not that reliable - and data can prove this. In Don't Trust Your Gut, he unearths the startling conclusions that the right data can teach us about who we are and what will make our lives better. Over the past decade, scholars have mined enormous datasets to find remarkable new approaches to life's biggest self-help puzzles, from the boring careers that produce the most wealth, to old-school, data-backed relationship advice. While we often think we know how to better ourselves, the numbers, it turns out, disagree. Telling fascinating stories through the latest big data research, Stephens-Davidowitz reveals just how wrong we really are when it comes to improving our lives, and offers a new way of tackling our most consequential choices.

The Spark distributed data processing platform provides an easy-to-implement tool for ingesting, streaming, and processing data from any source. In Spark in Action, Second Edition, you'll learn to take advantage of Spark's core features and incredible processing speed, with applications including real-time computation, delayed evaluation, and machine learning. Unlike many Spark books written for data scientists, Spark in Action, Second Edition is designed for data engineers and software engineers who want to master data processing using Spark without having to learn a complex new ecosystem of languages and tools. You'll instead learn to apply your existing Java and SQL skills to take on practical, real-world challenges. Key Features * Lots of examples based in the Spark Java APIs using real-life dataset and scenarios * Examples based on Spark v2.3 Ingestion through files, databases, and streaming * Building custom ingestion process * Querying distributed datasets with Spark SQL For beginning to intermediate developers and data engineers comfortable programming in Java. No experience with functional programming, Scala, Spark, Hadoop, or big data is required. About the technology Spark is a powerful general-purpose analytics engine that can handle massive amounts of data distributed across clusters with thousands of servers. Optimized to run in memory, this impressive framework can process data up to 100x faster than most Hadoop-based systems. Author Bio An experienced consultant and entrepreneur passionate about all things data, Jean-Georges Perrin was the first IBM Champion in France, an honor he's now held for ten consecutive years. Jean-Georges has managed many teams of software and data engineers.

Categorical data-comprising counts of individuals, objects, or entities in different categories-emerge frequently from many areas of study, including medicine, sociology, geology, and education. They provide important statistical information that can lead to real-life conclusions and the discovery of fresh knowledge. Therefore, the ability to manipulate, understand, and interpret categorical data becomes of interest-if not essential-to professionals and students in a broad range of disciplines.

Although t-tests, linear regression, and analysis of variance are useful, valid methods for analysis of measurement data, categorical data requires a different methodology and techniques typically not encountered in introductory statistics courses. Developed from long experience in teaching categorical analysis to a multidisciplinary mix of undergraduate and graduate students, A Course in Categorical Data Analysis presents the easiest, most straightforward ways of extracting real-life conclusions from contingency tables. The author uses a Fisherian approach to categorical data analysis and incorporates numerous examples and real data sets. Although he offers S-PLUS routines through the Internet, readers do not need full knowledge of a statistical software package.

In this unique text, the author chooses methods and an approach that nurtures intuitive thinking. He trains his readers to focus not on finding a model that fits the data, but on using different models that may lead to meaningful conclusions. The book offers some simple, innovative techniques not highighted in other texts that help make the book accessible to a broad, interdisciplinary audience. A Course in Categorical Data Analysis enables readers to quickly use its offering of tools for drawing scientific, medical, or real-life conclusions from categorical data sets.

A comprehensive guide to data analysis techniques for physical scientists, providing a valuable resource for advanced undergraduate and graduate students, as well as seasoned researchers. The book begins with an extensive discussion of the foundational concepts and methods of probability and statistics under both the frequentist and Bayesian interpretations of probability. It next presents basic concepts and techniques used for measurements of particle production cross-sections, correlation functions, and particle identification. Much attention is devoted to notions of statistical and systematic errors, beginning with intuitive discussions and progressively introducing the more formal concepts of confidence intervals, credible range, and hypothesis testing. The book also includes an in-depth discussion of the methods used to unfold or correct data for instrumental effects associated with measurement and process noise as well as particle and event losses, before ending with a presentation of elementary Monte Carlo techniques.

Organizational surveys are widely recognized as a powerful tool for measuring and improving employee commitment. If poorly designed and administered, however, they can create disappointment and cynicism. There are many excellent books on sampling methodology and statistical analysis, but little has been written so far for those responsible for designing and implementing surveys in organizations. Now Allan H Church and Janine Waclawski have drawn on their extensive experience in this field to develop a seven-step model covering the entire process, from initiation to final evaluation. They explain in detail how to devise and administer different types of organizational surveys, leading the reader systematically through the various stages involved. Their text is supported throughout by examples, specimen documentation, work sheets and case studies from a variety of organizational settings. They pay particular attention to the political and human sensitivities concerned and show how to surmount the many potential barriers to a successful outcome. Designing and Using Organizational Surveys is a highly practical guide to one of the most effective methods available for organizational diagnosis and change.

High-dimensional probability offers insight into the behavior of random vectors, random matrices, random subspaces, and objects used to quantify uncertainty in high dimensions. Drawing on ideas from probability, analysis, and geometry, it lends itself to applications in mathematics, statistics, theoretical computer science, signal processing, optimization, and more. It is the first to integrate theory, key tools, and modern applications of high-dimensional probability. Concentration inequalities form the core, and it covers both classical results such as Hoeffding's and Chernoff's inequalities and modern developments such as the matrix Bernstein's inequality. It then introduces the powerful methods based on stochastic processes, including such tools as Slepian's, Sudakov's, and Dudley's inequalities, as well as generic chaining and bounds based on VC dimension. A broad range of illustrations is embedded throughout, including classical and modern results for covariance estimation, clustering, networks, semidefinite programming, coding, dimension reduction, matrix completion, machine learning, compressed sensing, and sparse regression.

Corpora are ubiquitous in linguistic research, yet to date, there has been no consensus on how to conceptualize corpus representativeness and collect corpus samples. This pioneering book bridges this gap by introducing a conceptual and methodological framework for corpus design and representativeness. Written by experts in the field, it shows how corpora can be designed and built in a way that is both optimally suited to specific research agendas, and adequately representative of the types of language use in question. It considers questions such as 'what types of texts should be included in the corpus?', and 'how many texts are required?' - highlighting that the degree of representativeness rests on the dual pillars of domain considerations and distribution considerations. The authors introduce, explain, and illustrate all aspects of this corpus representativeness framework in a step-by-step fashion, using examples and activities to help readers develop practical skills in corpus design and evaluation.