The focus of these two volumes is doing English corpus linguistics using corpora and corpus analysis tools that are available online. It was not too long ago that a decision to use language corpora involved a substantial commitment, of both time and perhaps cost, to acquire the resources, tools and skills necessary to gain access to the potential insights a corpus query might bring. This admittedly daunting situation might have discouraged a non-corpus linguist from entertaining a casual glance into corpus data to investigate an intuition or follow a passing curiosity as to how an unusual phrase or feature is used in general language or some specific text type. Most corpus software has developed to provide a considerable array of functions and sophisticated searches and to allow users a broad range of configuration options. Because of this the initial learning curve can be somewhat steep. In contrast, most computer users have little trouble using a search engine to find their way around information on the web. One of the ideas behind web-based corpus tools is that they might lower the bar for carrying out corpus analysis. To achieve this some tradeoffs in terms of functionality might result. But as Anderson & Corbett suggest in one of the titles under review: ‘For the relative beginner. . . having a limited number of possibilities can make getting to grips with corpus analysis rather less daunting’ (p. 9). Hoffmann, Evert, Smith, Lee & Prytz make a similar claim in their preface that: ‘The uninitiated corpus-user, with very little introduction and practice, will already find BNCweb a revealing window into the possibilities of corpus linguistics' (p. xiv). The two volumes under review thereby suggest that it is indeed possible for the bar of entry to be lowered. I will now discuss each in turn.
1 Corpus linguistics with BNCweb – a practical guide
The British National Corpus (BNC) is a 100-million-word collection of written and spoken British English compiled in the early 1990s. It continues to be used extensively as a large and varied sample of contemporary English even though it is twenty years old and does not include instances of texts in now dominant electronic genres (blogs, web pages, etc.) and only a limited number of email texts from an early mailing list (see pp. 17–18). Between them the authors of this volume have considerable expertise concerning the compilation, annotation, classification and analysis of the BNC, as well as the development of computational tools for its analysis – specifically the BNCweb tool. From the title the reader might reasonably expect to find at least three components in the volume: (1) material concerning corpus linguistic theory and method, (2) details about the British National Corpus itself and (3) a user guide and tutorials in the use of the BNCweb tool. These elements are all covered, but not given equal attention. Nine of the thirteen chapters relate to the use of BNCweb, one (chapter 3) to a description of the BNC and three (chapters 1, 2 and 5) serve to provide a more general introduction to corpus linguistics. This balance is appropriate and supports what must be the primary aim of the book as a reference tool and, to some extent, a user guide to BNCweb. Having said this, the authors have managed to weave discussion of corpus methodology and the interpretation and implications of the data for questions in language and linguistics into the more practical ‘how to’ chapters.
The first chapter introduces the types of questions about language in use that corpus analysis is best placed to address (What does goalless mean?, When is it appropriate to use shall instead of will? Do men or women talk more about cars?) and demonstrates how looking at a corpus, and the BNC in particular, provides both quantitative and qualitative insight. Chapter 2 is a short introduction to corpus linguistics with a particular focus on the question of representativeness and the value of corpus data – namely quantitative facts and qualitative descriptions about language use – in contributing to linguistic theorizing. This section contains a nice example of using the spoken section of the BNC to examine where filled pauses (er and erm) occur before, after and in the middle of complex prepositions, such as in terms of. They discuss the implications of the finding that very few pauses are found mid-position for how such items are stored and theories of linguistic competence. From this they argue that corpus linguistics is more than simply a descriptive method but can contribute to linguistic theory proper. This position concurs with recent work from cognitive and usage-based linguistic perspectives that reject the competence–performance distinction and value frequency data from corpora (see Gries Reference Gries2008a, Reference Gries, Robinson and Ellisb; Ellis Reference Ellis, Robinson and Ellis2008; Ellis & O'Donnell forthcoming). The third chapter is an excellent overview of the structure, components and linguistic nature of the BNC. It could serve as a useful concise introduction to the corpus outside use of the BNCweb. I will skip to chapter 5, which is the next chapter to deal with corpus linguistic method and theory. It is particularly worthy of note as a clear exposition of some of the key quantitative elements of corpus-linguistic methodology and builds on the discussion in chapter 2. It is one of the clearest explanations of the importance of considering relative frequency, statistical significance and issues related to the formulation of corpus searches (particularly the concepts of precision and recall) that I have come across in a corpus linguistics textbook.
Chapter 4 provides a step-by-step introduction to the basic functionality of BNCweb: creating a search query, paging through results and examining context. Chapter 6 builds upon this chapter through a detailed coverage of the ‘Simple Query Syntax’ supported by BNCweb, which, through a combination of wildcards and proximity, can actually allow the retrieval of quite complex structures such as the dative alternation. Chapter 7 focuses on the analysis of concordance lines, how query results can be broken down in various ways according to metatextual categories (register/genre and, for spoken texts, speaker factors such as age, gender and social class), and how sorting a concordance can reveal patterns and frequent co-occurrence (collocates). A nice example is given demonstrating the difference in the use of the words pretty and handsome as pre-modifiers. A look at the corpus evidence confirms the intuitive notion that one tends to be used to describe males and the other females but also reveals other usage patterns that are not so immediately accessible. In this chapter and throughout the authors are careful to balance the complex array of descriptive statistics available through searches and the cross-tabulation of features that the BNCweb tool makes possible, with the caution that such data alone should not be seen as providing an explanatory account. Chapter 9 develops the previous one through an in-depth examination of the concept of collocation, and how to retrieve collocates using BNCweb. It provides an excellent explanation and comparison of the various measures of collocational strength, and is again worthy of note for its clarity.
Although there are many advantages to online corpora and online tools for corpus analysis, there have been a number of areas of functionality that are better supported in traditional ‘desktop’ corpus tools, such as the ability to generate and manipulate frequency and ‘keyword’ lists and the ability to filter and classify concordance lines. Chapters 9, 10 and 11 are devoted to these topics and to illustrating how the BNCweb tool offers considerable flexibility for user-defined corpora (i.e. selected subsets of the BNC; chapter 10), the annotation of concordance lines (chapter 9) and the generation of frequency lists and keyword comparisons (chapter 11). Chapter 9 demonstrates how it is possible for a user to create a set of classification labels – an example is given classifying instances of you know into discoursal, lexical and other categories – and then apply these to lines in the concordance. The tool will then generate distributional statistics on the basis of this categorization. For more advanced classification the concordance lines can be exported to a spreadsheet, annotated and then re-imported (in part) into BNCweb. The BNCweb tool itself is built on top of the Corpus Query Processor (CQP), which is part of The IMS Open Corpus Workbench (cwb.sourceforge.net). Chapter 12 explores a range of more complex queries that are possible using that platform. Hardie (in preparation) is working to provide an online corpus platform, CQPweb, that mirrors BNCweb in terms of functionality, but will allow any corpus (in the appropriate format) to be uploaded and used. So skills learned using BNCweb will be transferable. The final chapter (13) is aimed at more technical users and those wishing to administer or install their own version of BNCweb.
2 Exploring English with online corpora
The orientation for this volume is the study of spoken and written English from the perspectives of lexis, grammar, discourse, pragmatics and sociolinguistics. These levels of language are viewed through corpus linguistic analysis using online corpora. The authors are members of the SCOTS (Scottish Corpus of Texts and Speech) project and that corpus and its online tool (www.scottishcorpus.ac.uk) are used for many of the examples and tasks throughout the book. The other online corpora utilized are: (1) the Corpus of Contemporary American English (COCA; www.americancorpus.org, which at time of writing is 410 million words); (2) the British National Corpus (using two further online interfaces, but not BNCweb – corpus.byu.edu/bnc and www.natcorp.ox.ac.uk); (3) the Michigan Corpus of Spoken Academic English (MICASE; micase.elicorpora.info); (4) the TIME corpus (corpus.byu.edu/time). Using a range of corpora is a strength for the approach of the book, as it exposes readers to different types of corpora, but also a challenge as each tool (aside from the Brigham Young University corpora, that is, COCA, TIME, BYU-BNC and recently the 400-million-word Corpus of Historical American English, COHA, that share a common interface; see corpus.byu.edu) has a different search interface and functionality. The authors manage this situation well by focusing on specific linguistic questions, selecting an appropriate corpus for each and providing step-by-step instructions for how to carry out the specific search task. They use task boxes throughout for this purpose.
Chapter 1 provides an introduction to the five online English corpora utilized in the book, a brief discussion of what a corpus is and issues in corpus design. Chapter 2 is a concise overview of corpus analysis and quantitative methods with a strong focus on the interpretation of results obtained from searching (online) corpora. Some of the discussion in the chapter might be confusing for those new to corpus analysis, but there are example tasks throughout the chapter to reinforce the concepts. The focus of chapter 3 is the analysis of lexis, which is the natural starting point for corpus analysis. Anderson & Corbett point out that ‘the study of words and phrases is most accessible for students who have little or no previous experience of using a corpus’ (p. 45). The concordance as a tool for examining lexical items in context and the concepts of collocation, colligation (grammatical collocation) and semantic prosody are explained and illustrated. There is also a discussion of idioms and semi-fixed expressions and the surprising difficulty of retrieving them in corpora on account of the creative variation of language users (see Moon Reference Moon1998; Granger & Meunier Reference Granger and Meunier2008). The use of corpora such as SCOTS and MICASE allow illustration of the importance of considering mode and genre in lexical analysis. They present an example of the differences in the use of the word totally in the spoken and written components of SCOTS. Chapter 4 turns attention to ways in which grammar can be analyzed using the online corpora and will be of particular value as a model for integrating corpus analysis into grammar courses. It begins with continued attention to lexis, presenting the word blame as an extended example, used both as a noun and a verb. The chapter then moves on to show how grammatical items, specifically prepositions and pronouns, which are more numerous than the previously analyzed lexical items, can be handled using corpus analysis. Next, types of phrases (noun, verb, prepositional, adverb) are treated in turn, with directed tasks to support each of the structures discussed. This is followed by a discussion of clause structure, using both SPOCA (Subject-Predicator-Object-Complement-Adverbial) analysis and traditional categories of verb valency, and on to complex sentence types. Analysis of levels above the word is easier if corpora have at least part-of-speech (POS) and perhaps syntactical (i.e. a treebank or parsed corpus) annotation. Few online corpora are annotated beyond the POS level, and therefore it is not possible to do the kinds of analysis offered by tools such as ICECUP with its Fuzzy Tree Fragment queries (Nelson et al. Reference Nelson, Wallis and Aarts2002; see O'Donnell Reference O'Donnell, Römer and Schulze2008 for an evaluation of ICECUP and parsed corpora from the Survey of English Usage). However, the discussion and examples presented in the chapter on grammar demonstrate just how much can be investigated using the online tools available.
Chapter 5 moves up another level to explore the ways in which corpus analysis can be applied to the study of discourse. The term ‘discourse’ can mean very different things depending on whether one is interested in surface patterning (lexis and grammar) above the level of the sentence (i.e. features such as cohesion, clausal relations, etc.; see Hoey Reference Hoey2005) or more at a pragmatic level (i.e. issues of language and power, group identity, institutional structures, etc.; see Baker Reference Baker2005). Although local lexicogrammar is the field that has been most fruitfully plowed in corpus research, recent studies have demonstrated how a combination of corpus and discourse analysis promises a similar yield of insights (Partington et al. Reference Partington, Morley and Haarman2004; Baker Reference Baker2006; Hoey & O'Donnell Reference Hoey, O'Donnell and Lewandowska-Tomaszczyk2008). As the authors state, to study discourse through corpus analysis your corpus should be made up of full texts (the ‘whole-text’ policy in corpus compilation long advocated by Sinclair (Reference Sinclair1991: 19), among others) and your tool should support a full text view and allow analysis beyond basic concordances and a fixed word window collocation analysis. Unfortunately, while ‘many online corpora have adopted a whole-text policy in their design, it is not always possible for reasons of copyright for the complete texts to be accessible through the online version of the corpus’ (p. 109). Fortunately, two of the chosen online corpora, SCOTS and MICASE, do not have such restrictions, so the authors are able to provide examples of exploring lexical and grammatical patterning across paragraph and text units. Other online corpora that provide full text access include the student writing corpora MICUSP (Römer & O'Donnell forthcoming) and BAWE (Alsop & Nesi Reference Alsop and Nesi2009). The latter is available through another online corpus tool, The SketchEngine (sketchengine.co.uk; Kilgarriff et al. Reference Kilgarriff, Rychly, Smrz and Tugwell2004), that is briefly mentioned in this volume and is another tool worthy of investigation for teaching and research, particularly as it allows users to upload their own corpora. Among the many topics covered in this chapter are adjacency pairs (‘mutually relevant turns which are normally found together in discourse’, p. 112) in spoken discourse and paragraph collocates such as assume and in fact in written discourse.
Chapter 6 is worthy of mention as it highlights the use of (online) corpora in the study of pronunciation. This is a particularly understudied area in corpus linguistics because of the limited availability of phonetically transcribed or multimodal corpora. Examples of how SCOTS can be searched for instances of accent variation are useful. Also, the Speech Accents Archive (accent.gmu.edu) is offered as an online resource for the teaching and study of phonetics. Chapter 7 demonstrates the use of contextual information available with certain online corpora (SCOTS, MICASE and the BNC, although it is not so accessible in the BYU tool used in this volume) for sociolinguistic investigation. There is a short final chapter that discusses some of the issues involved in incorporating the use of online corpora in teaching and research and prospects for future developments and research. The appendix contains details of a range of online English corpora and there is a glossary of both the corpus and general linguistic terminology used in the book.
The two volumes discussed here share in common a focus on introducing teachers, students and researchers to the use of language corpora online and orienting readers to the basic principles and concepts of corpus analysis. The BNCweb volume is clearly first and foremost a handbook and introduction to the use of a corpus tool, whereas the Exploring English volume is organized around linguistic levels as they might be taught in an introductory course. They also differ slightly in their intended audience and probable usage. In teaching I would recommend the Anderson & Corbett volume for use in courses on English grammar and composition, and courses with a focus on spoken discourse or conversational analysis. The volume by Hoffmann et al. is particularly suited for use in a corpus linguistics (or even computational linguistics) module. It would also be valuable for a graduate student or researcher seeking to acquire skills in corpus analysis or access to rich usage data from a large corpus. Both volumes contain a range of tasks and exercises suitable both for class assessment and for self-study. Despite their common focus, the two volumes complement each other well and could be utilized side by side. I recommend them both to corpus and non-corpus linguists alike and, for the latter group in particular, would hope that they might encourage the formerly daunted to dip their toes into the waters of corpus analysis.
