What can a corpus tell us about discourse
5 likes1,902 views
The document discusses the concept of discourse, its multifaceted definitions, and how corpus linguistics can aid in understanding discourse features through quantitative analysis. It emphasizes the importance of context in interpreting discourse and outlines limitations of corpus methods, stressing that cohesive ties and semantic relationships are complex and not easily captured by quantitative data alone. Furthermore, it highlights practical steps for a corpus-based approach to discourse analysis and the significance of obtaining relevant texts for thorough examination.
What can a corpus tell us about discourse
- 1. What can a corpus tell us about discourse? Mª Paz Muñoz Elisabet Martínez Ainoa Ortiz Ester Ortega Kautar Ouatik 2º Grado en Estudios Ingleses
- 2. 1. What is discourse? The term “discouse” has two meanings: Slippery: it eludes neat definition Baggy: it embraces a wide range of linguistic and social phenomena. Two basic senses of the term “discourse”: Discourse as conected text -> formal sense Discourse as language in use -> functional sense
- 3. The reference to the context and the participants can include not just the inmediate context of situation but the larger social and cultural context as well. Acording to Schiffrin “ To understand the language of discourse…we need to understand the world in which it resides” We can distinguish between two broad areas under “ the describable internal relationships”: Cohesion across sentences and utterances, using grammatical and lexical devices (Halliday and Hasan 1976) The organisation and management of discourse, including the distribution of given and new information, topic management. ( Brow and Yule 1983, Coulthard 1985 and McCarthy 1991)
- 4. 2. What can a corpus tell us about discourse? Analysts would need to use quantitative methods with the aim of producing findings that are both descriptive and explanatory. The descriptive findings are generated by searching for particular discourse features in a corpus - a collection of texts of a specific register, a single extended text, such as a textbook or a novel - using computational means.
- 5. Explaining the frequency, significance and use of these features generally involves reference to context, either the immediate co-textual environment, or to other text or other corpora of texts. For, as Stubbs (2001a) remind us, ' in corpus work, context means two rather different things: co-text (a short span of a few words within one single text), and inter-text ( repeated occurrences, often a very large number, of similar patterns across different, independent texts').
- 6. The analyst may compare and contrast an individual text, or a sub-corpus of texts of a specific type, with text of another type, or with a larger and more general corpus. Corpus-derived frequency information has revolutionized language description at the level of lexis and grammas, so too has the study of discourse hugely benefited from the kinds of quantitative data that corpora yield.
- 7. Using corpus tools to identify what makes individual texts cohesive, or to track their internal organization through the use of discourses makers is more problematic. From a discourse perspective, typically involves identifying the micro-features of specific text types and from this extrapolating textual macro-features.
- 8. Corpus tools cannot easily detect cohesive tries, such as pronominal reference, unless they have been tagged as such. It is another matter to identify what a device is cohesive with. More amenable to corpus analysis are features of lexical cohesion ( Halliday and Hasan 1976) including reiteration ( the direct and indirect repetition of words, the use of synonyms, near synonyms and general terms) and collocation.
- 9. Studies combining corpus data and other research tools, such as discourse completion tests (DCTs), provide insights into how the corpus data are realized in specific context and between specific participants. They combine corpus-based procedures with research methods from other disciplines, such as genre analysis, phraseology, pragmatic and ethnography. Research into the third level of discourse, discourse-as-social-practice.
- 10. 3. What are the limitations of using a corpus in the study of discourse and how might we overcome them? Discourse analyst have been slow to embrace the opportunities offered by corpus linguistics, due to the perception that corpora consist of de-contextualised text fragment. Discourse analysis requires whole texts, often of the same type. Specialised corpora of specific registers, including spoken language, have proliferated, and most corpora of general English, including many that are freely available online, are tagged for text-type and register.
- 11. But there is a more fundamental problem facing the discourse analyst. While corpus tools allow researchers to track, tally and plot the surface features of discourse. They do not necessarily correlate with the underlying semantic relations between part of a text. This is a limitation of the study of cohesion in general. Halliday and Hasan: “cohesion expresses the continuity that exists between oe part of the text and another” In the end, quantitative data alone are not going to answer all the questions that analyst bring to the study of discourse. A comination of computation and interpretation offer the most promising way forward.
- 12. 4. How does a corpus-based approach work in practice? It’s necessary to follow these steps to do the approach (analysis): Frequency list & words researches Identify regularities of a corpus of linguistics. Constructs a provisional schematic (It can be checked and refined) Account the contextual and cultural factors of texts. Researcher is in a position of speculate (how the formal features of the texts encode their communicative and social functions)
- 13. An example of a small corpus (10.000 words) of teenage written narratives: the Cringe Text Corpus, it was compiled using an online teenage magazine as the source. It was used: To compile a list of the most frequent words in the corpus. To search these for linkers (eliminate instances of linkage at the phrase level , I was so embarrassed )
- 14. For example: Paratactic syntax (If it is used “and” many times) Hypotactic style (Subordinate clauses, academic abstracts)
- 15. Dispersion plot: Graphically tool, where the position of the item in each of the sixteen texts in which it occurs is visually represented by a short vertical line in the column headed Plot .
- 16. These thirty words alone are a strong indicator as to the thematic content of the narratives that comprise the Cringe Text Corpus. The fact that many are semantically related, either because belong to the same lexical set such as school, locker, walked, ran or to the same word family: friend, friends, walking, walked..
- 17. The semantic relations are typically instantly in the form of cohesive chains there are two types : identity chain and similarity chain Identity chains typically run the length of the whole text. Similarity chain of words that are related simply because they commonly co-occur is called a cluster.
- 18. An identity chain is a set of items that are co-referential: every member of the set refers to the same person or event, (I, me) The items in a similarity chain “belong to the same general field of meaning, referring to actions, events, and objects and their tributes) Cross-checking with the individual texts in the corpus we can clearly see that they follow a narrative structure that shares characteristics with the structure described by Labov and Waletzky Abstract-orientation-complication-evaluation- resolution-coda
- 19. In short, story-telling is the way that women and teenage girls perform their gender. A more critical analysis might argue that such discursive practices maintain and reproduce asymmetrical power relations in society, and that the teenage girls magazines are complicit in a process of discursively positioning their readership as the helpless and disempowered objects of male derision. In this point, the statistical data that a corpus approach delivers can serve to corroborate the findings of a more impressionistic approach, to confirm or disconfirm, hunches, and to suggest new directions for further interrogation of the texts themselves. This cyclical alternation between counting and interpreting accurately characterises the application of corpus analysis to discourse.
- 20. 5. Wat kind of data do you need to study discourse? Obviously, discourse analysis needs texts; if not whole conversations, at least pretty long fragments. Most discourse analysis focuses on textual features of specific text-type, so it is needed sufficient examples of these to provide general data. However, it does not mean it has to be enormous. As Partington says, in a recompilation of texts of similar type, the interactional processes and the contexts remain fairly constant. Therefore, O'Keeffe claims that 'specialised lexis and structures are likely to occur with more regular patterning and distribution even with relatively small amounts of data'.
- 21. One of those kinds of corpus (that targets a specific text-type and register) would be the Cringe Text Corpus. Another specific, small corpus that was compiled using texts available on the internet is one of 24,000 words (from an academic journal) which consists of 139 texts, with about 174 words each. For descriptive and pedagogical purposes, a corpus this size can give us enough information. For a more specific study, some form of tagging (grammatical, semantic or phonological) is almost a requirement. Nevertheless, Baker says that first, corpus builders have to think about what type of questions they want their corpus answer, and then decide if some particular forms of tagging are necessary or not.
- 22. The advantage of a small and homogeneous corpus is that the context is precise and specific, and this is crucial if the goal is the study of discourse-as-language-in-context or the investigation of discourse-as-social-practice. Mahlberg notes that ‘the way in which an analysis of corpus data can be related to social situations depends on the information that is available on the origins and context of the texts’. Moreover, ‘if the texts in a corpus are selected according to transparent criteria and information on their contexts is stored together with the texts, corpora can provide useful insights into meanings that are relevant to a society and indicative of the ways in which society created itself’.
- 23. Developments in the application of complex system theory to language acquisition and use suggest that we are experiencing the conjunction of two disciplines, corpus linguistics and psycholinguistics, which used to work in parallel, even though both were concerned with frequency effects and the important of usage. The association of these two fields has influenced, for instance, in Ellis and Lu’s research of formulaic expressions. As well, investigation may demonstrate the frequency of occurrence of specific discourses, and of variations in their texture, both influence and are influenced by the performance of these discourses, by individuals and across whole socio-cgroupsultural.