corpus linguistics

corpus linguistics

The Fascinating World of Corpus Linguistics: Theory, Methods, and Applications

1. Introduction to Corpus Linguistics

At first glance, field linguistics may not seem to have much in common with modern analyses of data from large machine-stored corpora. However, we argue that field linguistics can learn a great deal from corpus statistics. The model for analyzing the data is now referred to as corpus statistics, which is derived from a usually very large machine-readable collection of texts known as a corpus.

Second, “corpus” is often understood as a collection of specific types of texts. However, it is important to note that researchers are not required to limit the elements of their collections to texts. Spoken language can take various forms and may include things like TV programs, personal communication, trial testimonies, or parliamentary debates. We suggest that the definition of “corpus” should be more inclusive. It can consist of records of language that are received and stored in any form that allows for linguistic analysis, not just those composed of texts.

There are a few points that need to be made at the outset. First, we use the term “approach” instead of “method” because it conveys the idea that corpus linguistics is a flexible toolkit and that users have different choices when selecting specific techniques.

Corpus linguistics is a methodological discipline that can go a long way towards helping field linguists in their investigative work. It is a generally applicable methodology primarily used in linguistic research. It has been developed to support linguists in their research with various purposes and objectives, which explains its widespread popularity and rapid development.

2. Building and Annotating Corpora

To prime the data-cleaning part of your Pythonic corpus-building process, two useful tools are the Beautiful Soup HTML/XML parser and the NLTK. The Beautiful Soup requires only a few lines of Python code to create a soup that is aware of the nested attributes of an HTML file or of the graphic-pixel coordinates of a PDF file. Interfacing the soup with specific HTML or PDF elements sorts words and punctuation marks—so-called tokens—into a list of input for, say, a widely used natural-language toolkit.

In the field of corpus linguistics, the single most conspicuous activity is the preparation and studying of text corpora. The term corpus is derived from the Latin word that means “body.” In English, the word is used to denote “a collection of written texts, especially the entire works of a particular author or a body of writing on all subjects.” A corpus linguist’s corpus is, essentially, a large electronic database of written texts. A text incorporated in a corpus is usually termed a document. Development of corpora is limited only by the interest, money, and time of the linguist collecting the data. The advantage of having a corpus is that it is a manageable collection; one can index it, search for specific features, and dodge the problems of analysis inherent in manual processing.

3. Quantitative Analysis Techniques in Corpus Linguistics

Despite the great variety of available tests common in other fields, they are relatively infrequently reported in CL studies compared to the frequency of their application. The most common tests are chi-squared tests of independence, such as the chi-test for tabulations and analysis of variance for word length, in addition to the parametric t-test or its non-parametric equivalents, such as the Mann-Whitney test. Differences exceeding specific magnitudes are their standard of judgement. Nonetheless, if differences fall short of this level yet follow the same pattern in each separate text, they may suggest directionality not detectable by these statistics.

Although qualitative analysis is essential for the closer examination of the concordance, it is quite limited in studying relationships of different types between elements of the concordance. Quantitative types of analysis examine the frequency of different elements of concordances, seeking to discover common and atypical regularities. This assumption underpins the algorithms and the technology of all linguistic sorts. The most common word, bigram, and dispersion concordance analyses are the most common. Nevertheless, it is often descriptive and shallow in practice: it is assumed that differences identified in these analyses are significant, while arguments from inferential statistics are weak if they appear at all.

4. Qualitative Analysis and Interpretation of Corpus Data

This chapter provides an overview of available techniques of and ideas for how to approach qualitative analysis. To varying degrees, they are applicable both to spoken and written language, synchronous dialogue and sequential monologue interaction, and regardless of the specific type of language use: socio-pragmatic genre characteristics as well as author-speaker style and idiolect can be valid objects of investigation. The techniques that are spelled out in more detail, however, shed light on other levels of (spoken) language use more clearly, such as lexis or syntax. This is on purpose, since the primary interests of this course are focused on exactly these aspects, and since many aspects of qualitative analysis will be introduced in the respective chapters. Finally, for reasons of space, the discussion of some crucial aspects of qualitative analysis such as replicability or the choice of appropriate research questions and different analytical perspectives is left out, and a brief outlook concludes the chapter.

One of the major goals of corpus linguistics is to test or build new theories of language, preferably theories of cognitive, communicative, and social aspects of language that have gained empirical validity. Qualitative analysis and interpretation are crucial for this purpose, and in many ways the most characteristic feature of corpus linguistics, either compared to computational text analysis or to experimental psycholinguistics pursuing similar goals. The development of comprehensive qualitative methods, and introductory textbooks that make these methods more widely accessible, is an important research task for many subfields of corpus linguistics, but it by no means overshadows the existing qualifications of the majority of corpus techniques to feed careful and fine-grained qualitative analysis. In most cases, it is not an either/or question, but an intelligent combination of both that is necessary and helps researchers make the best use of their data.

5. Applications of Corpus Linguistics in Linguistic Research and Beyond

This study emphasizes the external attributions of successful Chinese learning strategies for beneficial modeling by Chinese learners. The results also provide valuable multitier language teaching resources for practical applications and propose responsive strategies in intermediate Chinese language education.

The main purpose of this article is to introduce a way to explore the Chinese college students’ mental state by using errors contained in the CCUS and clarify the correlation between the students’ errors and their anxiety and satisfaction in achieving a desirable Chinese proficiency level. 447 essays submitted by 13 non-Chinese-major Chinese learners were analyzed in this study. Research results show that high satisfaction learners tend to use complex sentences, while the tenuous, anxious ones easily produce grammatical errors. Data from the CCUS for causation interpretation and expression proves to be more challenging, which is a major reason leading to low metacognitive strategies, high levels of anxiety, and low completion satisfaction.

Corpus linguistics is a research method that is applicable to various fields. After the foundation of the British National Corpus in the early 1990s, since 2019, 4.27 billion Chinese words have been included in the Corpus of Chinese University Students (CCUS), which opens up an unprecedented way to study Chinese college students’ language learning and language use from a data-driven perspective. This helps to reveal learners’ writing characteristics, learning strategies and difficulties, learning motivation, and mental state, and to inquire into the adaptive learning model. However, corpus-based studies in the Chinese as a foreign language field have just started.