Corpus Linguistics: Analysis Techniques

Definition of Corpus

• A corpus is a large, principled collection of naturally occurring language.
• Corpora are usually stored electronically.
• Texts in a corpus must be produced in a natural communicative setting.
• The texts are spoken or written for an authentic communicative purpose.
• The purpose is not to put them into a corpus.
• A corpus may consist of newspaper articles, even though they are included for convenience.
• Journalists write articles to communicate with readers, not to fill a linguistics corpus.

Definition of Corpus Linguistics

• The study of language in use using corpus analysis.
• It involves methods and principles of applying corpora in language studies and teaching/learning.

Definition of General/Generalized Corpora

• Very large corpora aim to represent language as a whole.
• They often contain more than 10 million words.
• A generalized corpus contains a variety of language to generalize findings.
• Examples:
  • British National Corpus (BNC)
  • American National Corpus (ANC)
  • COCA
  • Written texts (newspaper and magazine articles).
  • Works of fiction and nonfiction.
  • Writing from scholarly journals.
  • Spoken transcripts (informal conversations, government proceedings, business meetings).

Definition of Specialized Corpora

• Represents a particular part of a language.
• Examples include a particular subject field or dialect.
• It contains texts that represent the language of that type.
• They can be large or small.
• Examples:
  • Michigan Corpus of Academic Spoken English (MICASE)
  • CHILDES Corpus (language used by children)
  • Michigan Corpus of Upperlevel Student Papers
  • Medical corpus (language used by nurses and hospital staff)

Definition of Learner Corpora

• Contains written texts and/or spoken transcripts of student language.
• The students are acquiring the language.
• These corpora can be examined to see common student errors.

Definition of Pedagogic Corpora

• Contains language used in classroom settings.
• Includes textbooks, transcripts of classroom interactions, or any text learners encounter.
• Useful to make sure students learn useful language.
• They can also examine teacher-student dynamics, or as self-reflective tool for teacher development.

Definition of Data-Driven Learning (DDL)

• An approach to foreign language learning.
• Language is treated as data.
• Students are researchers undertaking guided discovery tasks, rather than following teachers and textbooks.

"Principled" in Corpus Compilation/Linguistics

• A corpus follows guidelines, which can differ depending on the corpus.
• The texts must be planned.
• The language must be chosen according to specific characteristics.
• Texts must be chosen to be useful for your research question/aim.
• Text choice depends on the corpus' goal.

Concrete Examples of Basic Corpus Linguistics Principles

• Context:
  • Words need context to convey meaning.
  • Isolated words can be interpreted differently.
• Language patterns:
  • Patterned nature is apparent through corpus analysis.
  • Collocations are words that usually co-occur.
  • Without one part of a collocation, the whole collocation does not make sense.
  • Tense and aspect: some verbs only stand with specific tenses or aspects.
• Language in use:
  • Corpus linguistics studies language in use.
  • It is not concerned with prescriptive "rules".
  • Some expressions/utterances may be strictly incorrect.
  • Language in use might not be the one expected.

Corpus Tools (Generally)

• Frequency counts (single words and multiword units).
• Concordancing.
• Pattern identification (collocation search).
• Corpus comparison.
  • Compare different language varieties (regional varieties, register).
  • Diachronic language comparison.
  • Compare general and specialized corpora to identify corpus-specific terminology.
  • For translation (multi-language corpora, e.g. linguee).

Research/Application Fields for Corpus Linguistics

• More theoretical:
  • Phraseology (multi-word units).
  • Lexicogrammar: connection between lexical and grammatical aspects.
  • Register, language change.
• More practical:
  • (Foreign) language teaching.
  • Lexicography.
  • Translation (studies).
  • Language for specific purposes.
  • Writing assistance/language reference.

"Authentic Language" and "Naturally Occurring Language"

• Every utterance (spoken, written, or transcribed) has been produced for communication.
• The purpose is not to put it into a corpus.
• Language has been produced in natural communicative settings.
• Authenticity also implies wanting language material produced by native speakers.
• Some corpora look at the language of English learners.

Accessing Corpora

• Download a corpus and corpus software.
• Download a corpus and access the corpus software online.
• Access a corpus and corpus software online.

Compiling Corpora

• Collect texts in accordance with the research interests.
• Use texts from books, newspapers, journals, transcriptions, texts by language learners.
• Use texts from the web.

Useful Corpora for Language Teaching/Learning

• Use online tools like Sketch Engine or downloadable tools like #LancsBox.
• General corpora
  • authentic language in context (BNC, enTenTen, English Corpus for SkELL, Open American National Corpus)
  • for working on lexical and grammar skills
• Specialized corpora
  • for working on current topics or bilingual subject teaching (Brexit, ScienceBlogs, Environment corpus, e-flux, EcoLexicon English Corpus)
• Literary corpora
  • for working on literary skills (Project Gutenberg English, English Drama Corpus, Shakespeare English Drama Corpus)

Main #LancsBox and Sketch Engine Differences

• Accessibility:
  • Sketch Engine is online.
  • #LancsBox is local.
• Pre-Loaded Corpora:
  • Sketch Engine: >700.
  • #LancsBox: 7.
• Languages:
  • Sketch Engine: >100. -#LancsBox: 2.
• Pricing:
  • Sketch Engine: Around 100€ per year.
  • #LancsBox: Free.
• Corpus building:
  • Sketch Engine: Up to 1 million for your own.
  • #LancsBox: No word limits for your own.
• Usability:
  • Sketch Engine: Basic knowledge is sufficient.
  • #LancsBox: Sound knowledge is helpful.

Sketch Engine's Concordance Tool

• A collection of word-form occurrences in its textual environment.
• In its simplest form, it is an index.
• References the place of occurrence in a text.
• It finds words, phrases, tags, documents and displays the results in context.
• The concordance can be sorted, filtered, counted, and processed further.
• Too many results from large corpora may find can be tedious.
• KWIC=Keywords in context -> concordance lines

Concordance Tool's Query Types

• Simple:
  • typing go will find goes, going, gone, went
• Lemma: finding all forms of the lemma (base form)
• Phrase:
  • finding phrases exactly as typed
• Word:
  • this option will find a word form as typed
• Character:
  • finding tokens with the character(s)
• CQL:
  • Corpus Query Language for complex criteria making use of part-of-speech tags and regular

Concordance Tool's Use in Corpus Analysis

• Subcorpus:
  • limiting the search
• Macro: saving searches for future use
• Filter context:
  • keeping lines fulfilling conditions
• Text types:
  • excluding or including specific documents or corpus parts

Tags

• A label is assigned to each token in an annotated corpus.
• They indicate part of speech.
• They often also show grammatical categories and morphological information.

Sketch Engine's "Wordlist" Tool

• Used to generate frequency lists of all kinds.
• Examples are words, lemmas, nouns, verbs, tags, words containing/not containing certain characters.
• To compile word frequency counts:
  • Log in and select a corpus.
  • Select corpus for the list to derive from.
  • Leave default settings.
  • Or, set criteria or use a custom search.
  • Click GO.
• You can compile wordlists for...
  • Nouns, verbs, adjectives and other parts of speech; words beginning, ending, containing certain characters; word forms, tags, lemmas and other attributes; or a combination of all of the above.

Sketch Engine tool “Keywords”

• Shows which words appear more in the focus corpus than would be expected.
• General language is represented by the reference corpus.
• Any token can qualify if it is used more in a focus corpus than in reference corpus.

How Keywords Can Be Helpful

• Sketch Engine combines statistics with linguistic criteria to extract keywords and terms.
• Useful in language learning or applied linguistics
• Can define or understand the main corpus topic.
• Easy to compare different corpora.

N-Gram Lists

• Frequency lists of multiword expressions (MWEs) or lexical bundles.
• N-gram = a sequence of items.
• An item can refer to anything.
• In linguistics, n-grams are sometimes multiword expressions.

Key Differences Between N-Gram and MWE Lists

• N-gram lists
  • Focus: Frequency of word sequences.
  • Output: Common fixed word (bigrams, trigrams, etc.)
  • Use Case: Find common phrases and collocations.
• Keyword + MWE
  • Focus: Significance compared to a reference corpus.
  • Output: Keywords and fixed expressions with semantic significance.
  • Use Case: Identify domain-specific or unusual words/phrases

N-Gram List Compilation

• Log in, choose a corpus, click N-GRAMS and select the number of items.

Analyzing Lexico-Grammatical Features

• With the help of the Word Sketch tool.

Sketch Engine tool "Word Sketch"

• Words collocated. Words and other words in its surroundings.
• Summary of the word's grammatical and collocational behavior.
• Results are organized into grammatical relations.

Sketch Engine tool "Thesaurus"

• Words used in similar contexts develop into synonyms.
• Identify collocations of both a given word and other words from the same part of speech. Compare sketches to those of the searched word, and find whose closest matches. It suggests similar meaning.
• Beware: not all synonymous
• May help with the lexical variability of students

Sketch Engine tool “Word Sketch Difference”

• Used to make comparisons by contrasting collocations.
• You can compare:
  • Lemma
  • Word Forms
  • Subcorpora

Use of Word Sketch Difference

• Helps understand the differences between similar words.
• Shows which collocations in which combinations appears.

Corpus Analysis Benefits

• Relevance
  • Content must be relevant.
  • Important that it appears frequently.
• Authenticity
  • Language produced by native speakers.
  • Language that appears in natural context.
  • Authentic language always appears in patterns.
• Response
  • Respond directly to your learner's needs (E.g. the writing assistance).
  • Allows you to look up more exercises with the help of corpora.
• Autonomy
  • Learners get aware of learning process.
  • The more aware, the more language can be learned.
• Availability
  • Corpus= available native speaker.
  • Data is available for everyone.
• Sustainability
  • Easily memorized due to the relevant topic.
  • It sticks longer in your head.

Corpus Linguistics (Indirect)

• Syllabus design
  • Which language patterns are important.
  • What to teach and when to teach it.
• Teaching materials design, reference work design
  • Always corpus-based.
  • Teach language actually spoken.
• Evaluating existing pedagogical material
  • Base material on literature.
• Selection of language phenomena.
• Presentation of selected items and structures.
  • Items interact with one another.

Data-Driven Learning Combination

• Language learners are "linguistic researchers" or "language detectives".
• Learners interact with corpus data pre-processed by a teacher.
• Teachers select concordance lines that students work.

Corpus Linguistics (Direct)

• Lexical skills: skills of synonymity, collocation, register
• Interferences between learners’ first and second language.
• Grammar structures in use
• Literature Lessons
• Responding to concrete learner needs

Benefits of Data-Driven Learning

• Learning how to learn.
• Personalized learning.
• Topics appear relevant.
• Communication and teamwork.
• Language is natural to learners.
• Lexico-grammatical Approach.
• Enhanced Noticing Skills
• Extended Cognitive Abilities
• Learner Autonomy

The Negative Aspects of Data-Driven Learning

• Requires great effort in the beginning.
  • Preparing takes a lot of time.
• Special resources need to be in place.
• License to use corpus programmes must be purchased.
• The curriculum is full already.

Impact of Corpus Analysis

• Prescriptive vs. descriptive.
• Corpus analysis allows to confirm actual use, disprove tradition.
• Reveals there is rarely commercial teaching material
• Corpus informs commercial teaching material in textbook.
• Data driven vs simplifications & lack of register variations

The Definition of lexico-grammatical approach”?

• Lexis and grammar are integrated The role of phraseology and formulaic language (e.g. phrasal verbs, collocations, light-verb constructions)

3 Categories to Phraseological Item Count

Lexico - Grammatical Continuum: Language Psycholinguistic View: Language Characteristics of Phraseological Items: phrase meaning

Types of Phrases

In BNC ”smile” verb main past simple tense The phrase is “love“ aspect of the verb.

Aspects of Usage-Based

Allow for integration of corpora into language pedagogy Focus on authentic texts Focus on communication

Reasons for Integrating Corpus Use

• Focus on the language itself.
• Empirical answers to language questions are empirical. Near synonyms
• Contrastive aspects -word equivalents Lexical features (e.g., collocations, discourse markers, text
• Data driven learning and people exploration of a language
• Reference point and up-to-date usage

Examples of Direct Corpus Use

• Respond to learners' information needs. Learners come across items and ask questions. Learners come grammar structures and ask questions Have explore structures Have learners explore patterns

Explore key aspects of corpus linguistics, including the significance of collocations, handling grammatically incorrect utterances, and comparative corpus analysis. Learn about the lexicogrammatical contributions and practical applications of corpus comparison. Understand how tools like 'Concordance' enhance corpus analysis through subcorpora and macros.