Advanced English II (BFE19013)

AntConc Hands-on Practice

Instructor: Miao Zhang 張淼

Today’s plan

Three parts

  1. Part 1 (~30 min): Have a review on what we learned previously and design your own mini research question
  2. Short break
  3. Part 2 (~80 min): AntConc tutorial — the core workflow
  4. Short break
  5. Part 3 (~40 min): Your own mini research with AntConc

What you will be able to do today

By the end of this session, you can:

  • review what a corpus is and why linguists use it
  • load a corpus in AntConc and run searches with different search strategies
  • use the KWIC, Cluster, N-Gram, Collocate, Word List, and Keyword tools
  • use corpus evidence from AntConc to answer a small research question

Part 1: Review

Quick warm-up

(Without looking at your notes)

  • What is a corpus? What constitutes a good corpus?
  • What are the limitations of corpus research?
  • What types of corpora are there?
  • What tools can you use to analyze a corpus? What does each tool do?

A corpus is MORE than a collection

A corpus is:

  • large enough to be useful
  • systematic and organized
  • machine-readable
  • designed for analysis
  • representative of a language variety or purpose

What makes a GOOD corpus?

  1. Representativeness — it represents the language variety it studies
  2. Finite size — it has a design and a collection goal
  3. Machine-readable — searchable and analyzable by computer
  4. Standard reference — other researchers can reuse it and repeat the study

Why use a corpus?

Even expert speakers:

  • do not know everything
  • may be biased by what they notice
  • cannot remember every pattern
  • cannot easily quantify language use

A corpus helps us test patterns in real data.

The nature of corpus research

  • It is empirical and data-driven.
  • It involves systematic analysis of language use in real life.
  • It provides objective evidence for linguistic (theoretical) claims.

The limitations

  • There is no negative evidence in a corpus.
    • Absence of evidence is not evidence of absence.
  • Corpora do not automatically provide explanations.
  • The corpus and the research question(s) determine the quality of the study.
  • Corpus research needs to be complemented by experimental studies at times.
    • Observational studies only tests associations, not causation.

Types of corpora

  • Sample vs. monitor — fixed in time vs. keeps growing
  • General vs. specialized — broad picture vs. one genre/group/domain
  • Written vs. spoken
  • Raw vs. annotated — plain text vs. tagged/labeled text

Basic corpus tools

  1. Concordancers — show all examples of a word with context
  2. Aligners — compare parallel texts (translation studies)
  3. POS taggers — assign grammatical categories
  4. Parsers — analyze structure and function

Today we focus on the first one — with a tool that does much more.

Where do corpora come from?

  • Online corpora — ready-made, e.g. COCA, BNC, COHA, GloWbE, NOW (english-corpora.org)
  • Build your own — collect texts that fit your research question (commercial tools like Sketch Engine, or free tools like AntConc)

Quick check

  1. You want to study how you and your classmates’ English proficiency improve over time. Do you need a sample or a monitor corpus?
  2. You want to know whether record is used more often as a noun or a verb. Can your intuition answer this reliably?

Design your own research question

Think of one question about language you could answer with a corpus. Some examples:

  • A word or phrase (Do people say X or Y more?)
  • A pattern (What words come after “the ___ of”?)
  • How pronouns are used in spoken vs. written English
  • How modal verbs are used in academic writing vs. speaking corpora?

Part 2: AntConc tutorial — the core workflow

What is AntConc?

AntConc is a freeware, cross-platform software tool for:

  • corpus linguistics research
  • corpus-based teaching
  • data-driven language learning

Developed by Laurence Anthony (Waseda University, Japan).

It runs on Microsoft Windows, macOS, and Linux.

Text and multimodal corpora

AntConc works with:

  • text-only corpora — plain text you load or build yourself
  • multimodal corpora — text documents linked to image, audio, video, or YouTube media, so concordance results can be checked against the original media context

Getting started

  • Windows (installer): double-click the installer, then launch from the Start menu
  • Windows (portable): unzip the package and double-click AntConc.exe
  • macOS: open the disk image and drag the AntConc app to Applications
  • Linux: install the Flatpak bundle, then run flatpak run org.antconc.AntConc

Do it now

Download and install AntConc on your laptop.

Open the program and make sure it starts.

Raise your hand when you see the main window.

The ten tools

AntConc contains ten tools, each with its own tab in the tool window:

# Tool What it does
1 KWIC concordance lines — usage in context
2 Plot barcode-style distribution of hits across texts
3 File View full source texts with hits highlighted
4 Cluster frequent word sequences around a query
5 N-Gram frequent n-word sequences in the corpus

The ten tools

# Tool What it does
6 Collocate words that co-occur near a query
7 Word List frequency lists of words
8 Keyword List statistically distinctive words vs. a reference corpus
9 Wordcloud visualizes results as a word cloud
10 ChatAI interact with an AI model using tool output as context

The same starting step everywhere

Every tool follows the same pattern:

  1. Load a corpusFile → Open Corpus Manager, or File → Open File(s) as a Quick Corpus
  2. Enter a query in the tool’s search box
  3. Set the search mode — “Words”, “Case”, “Regex”, or “Full Regex”
  4. Read the summary counts and inspect the results table

Remember this pattern — it repeats for all ten tools.

Corpora downloadable in AntConc

Open “File”, check the “Open Corpus Manager” option, and click the “Corpus Library” tab.

Find the corpora you want to download, and click the “Download” icon. After downloading, click “Tar” to set it as the Target Corpus.

A brief introduction to regular expression

Regular expressions (“regex”) are a compact language for describing patterns of text, not just exact words.

Instead of searching for one word, you search for a shape that many words or phrases can fit.

E.g., we want all forms of a verbwalk, walks, walked, walking, walker, to walk — not just one form. One regex can catch them all in a single search.

Why learn a little regex?

  • One pattern can catch many forms at once — all forms of a verb (walk, walks, walked, walking), singular and plural nouns…
  • You can find things a normal search cannot: “any word repeated twice in a row”, “a digit followed by a word”, etc.
  • In AntConc you meet regex in the “Regex” and “Full Regex” search modes.

The basic building blocks

Symbol Means Example pattern Matches
. any one character b.g bag, big, bug (not boog)
* previous item, 0+ times be*t bt, bet, beet, beeet
+ previous item, 1+ times be+t bet, beet (not bt)

The basic building blocks

Symbol Means Example pattern Matches
? previous item, 0 or 1 time colou?r color, colour
[ ] one of these characters analy[sz]e analyze, analyse
[a-z] one character in the range [aeiou] any one vowel
\| or cat\|dog cat or dog

Anchors and word boundaries

Symbol Means Example Matches
^ start of the token ^un unhappy, unfair
$ end of the token ing$ going, reading
\b word boundary \bcan\b can, not cannot or scan

These are how you say “the whole word”, “starts with”, or “ends with” inside a regex.

Regex for corpus questions

  • walk(s|ed|ing|er)? — catch walk, walks, walked, walking, walker in one search: parentheses group the endings, | means or, and ? makes the whole ending optional
  • \d+ — any number (3, 2026) — \d means a digit
  • \b(\w+)\s+\1\b — the same word twice in a row (“the the”, “is is”)

Practice Regex

Go to regex101.com

Paste the text I uploaded to moodle into the Test String box.

In the Regular Expression box, try these patterns:

  • \d+
  • \w+ and \W+
  • [\w.]+@[\w.]+
  • [\w']+,\s[A-Z]{1}.[\s,]

What did you catch with the above Regex?

Try figure out how to find these:

  • Find all years (4 digit numbers).
  • Find all speaker labels/time stamps/URLs.
  • Find all ISBNs in the format 978-0-521-80487-5.

Advanced search

  • Find all cases when the speaker 2 asked a question.
  • Find all dates.
  • Find all citations.

Wildcards in action

  • teach* — words starting with a sequence
  • *ing — words ending with a sequence
  • student|pupil or [student,pupil] — alternatives
  • Can I help_[?] — a word followed by a non-word marker (“Can I help?”)

Wildcard searches are fast and flexible — the best first step beyond simple lookup.

Regex in action

Token-level regex — each space-separated token is a pattern:

  • colou?r — matches color or colour
  • data.* driven — two tokens, first starts with data

Full-text regex — one regex against the whole document text (raw-files corpora only):

  • \bcorpus\s+linguistics\b — multi-word sequence, flexible spacing
  • \b(\w+)\s+\1\b — backreference: the same word repeated

Full Regex has the largest performance cost — use it with care on very large corpora.

Try it in AntConc

Using the KWIC tool:

  1. Search with a wildcard: how many words in the corpus start with un-?
  2. Search with a token-level regex: find all forms of analy[sz]e — which spelling is more common in this corpus?

Tool 1: KWIC (Key Word In Context)

What KWIC is for

The KWIC tool generates concordance lines for a search query — each hit is shown with its left and right context.

Designed for:

  • fast pattern discovery
  • close reading of usage
  • easy movement from summary results to full-text evidence in the File View tool

KWIC: core workflow

  1. Enter a query in the KWIC search box
  2. Set the search mode: “Words”, “Case”, “Regex”, or “Full Regex”
  3. (Optional) Adjust “Results Set” and “Context Size” — defaults are fine for a first run
  4. Click Start (or press Enter)
  5. Read “Total Hits” and inspect the concordance lines

Key move: double-click a KWIC row to open that hit in the File View tool.

KWIC: practical examples

  • Learning: compare near-synonyms (big vs. large) with separate KWIC searches and inspect the differences in context

Try it now

Load the practice corpus and open the KWIC tab.

Search for a word you are curious about.

  1. How many total hits do you get?
  2. Double-click a line to see it in File View.

Tool 2: Plot

What Plot is for

The Plot tool shows where search hits occur across files, as barcode-style distribution plots along each document’s timeline.

Designed for distribution analysis: is a word evenly spread, or concentrated in specific documents or sections?

Plot: practical examples

  • Learning: overlay plots for two near-synonyms and compare where each tends to appear

Plot: core workflow

  1. Enter a query in the Plot search box
  2. Set the search mode (“Words”, “Case”, “Regex”, or “Full Regex”)
  3. (Optional) Adjust “Plot Zoom” — defaults are fine for a first run
  4. Click Start
  5. Read “Total Hits” and “Total Files With Hits”, then inspect the plot rows

Click a plotted hit line to jump to its location in File View.

Practice 1

Using the Plot tool, search for a keyword from this week’s reading (e.g., corpus).

  1. In which file is the word most concentrated?
  2. Does that match what you would expect from the text?

Tool 3: File View

What File View is for

The File View tool displays the full text of a selected file and highlights search hits directly in the text.

Designed for close reading: inspect local context and move quickly between hits.

File View: practical examples

  • Learning: check whether your target expression appears naturally in authentic texts

File View: core workflow

  1. Select a file in the “Target Corpus” list (double-clicking a file also opens it)
  2. (Optional) Enter a query to highlight hits; if blank, the file loads without highlights
  3. Set the search mode
  4. Click Start
  5. Use “Hit Location” to step through hits — the selected hit is emphasized

File View connects to KWIC: you can send a highlighted hit to KWIC to search the whole corpus for it.

Tool 4: Cluster

What Cluster is for

The Cluster tool finds recurring multi-word sequences around a search term and summarizes them as ranked cluster types.

Designed to reveal phrase-level patterning: which combinations are frequent, widespread, and worth inspecting in KWIC.

Cluster: practical examples

  • Research: identify recurring phrase frames around a keyword (in terms of) and compare how often they recur across documents
  • Teaching: show learners typical phrase extensions before or after a target word — chunking and formulaic language
  • Learning: generate clusters for a key term to learn common phrase patterns for writing

Cluster: core workflow

  1. Enter a search query in the Cluster search box
  2. Set the search mode (“Words”, “Case”, and/or “Regex”)
  3. Set the main parameters: “Cluster Size”, “Min. Freq”, and “Min. Range”
  4. Choose “Search Term Position” — On Left, On Right, or On Left/Right
  5. Click Start

Practice 2

Using the Cluster tool, search for the.

Set cluster size to 3, search term position “On Right”.

  1. What are the most frequent 3-word clusters starting with the?
  2. Double-click one cluster to see it in KWIC.

Tool 5: N-Gram

What N-Gram is for

The N-Gram tool extracts recurring multi-word sequences of a specified length and ranks them by frequency and range.

With “Open Slots” enabled, you can examine patterned variation inside sequences (e.g., the ___ of).

N-Gram: practical examples

  • Learning: explore typical multi-word patterns in a genre and reuse them for more natural writing

N-Gram: core workflow

  1. Set the main parameters: “N-Gram Size”, “Open Slots”, “Min. Freq”, and “Min. Range”
  2. (Optional) Enter a query to filter n-grams
  3. Set the search mode
  4. Click Start

Practice 3

Using the N-Gram tool, set n-gram size to 3 (no query).

  1. What are the most frequent 3-word sequences in the corpus?
  2. Set “Open Slots” to 1 and search for the * of. What words fill the slot?

Tool 6: Collocate

What Collocate is for

The Collocate tool identifies words that co-occur with a search term within a defined left/right window span.

Designed for co-occurrence analysis: compare directional frequencies, range, and association strength to find meaningful collocational patterns.

Collocate: practical examples

  • Learning: build a collocation list for a topic word, then check each item in KWIC

Collocate: core workflow

  1. Enter a search query in the Collocate search box
  2. Set the search mode
  3. Set “Window Span” (From/To), “Min. Freq”, and “Min. Range”
  4. Click Start

Sort by Likelihood or Effect to find the strongest associations, not just the most frequent.

Practice 4

Using the Collocate tool, search for important.

  1. Set the window span to 3 left / 3 right. What are its top collocates?
  2. Which collocate has the highest Effect score? What does that mean?

Tool 7: Word List

What Word List is for

The Word List tool generates a frequency list of word entries from the target corpus, ranked by frequency, range, or lexical form.

Designed for lexical profiling: how often items occur, and how widely they are distributed across files.

Word List: practical examples

  • Learning: check common words and forms in a genre, then open KWIC to see typical usage

Word List: core workflow

  1. Enter a query only if you want to filter the list; leave it blank for a full word list
  2. Set the search mode
  3. Set “Min. Freq” and “Min. Range”
  4. Click Start

Practice 5

Using the Word List tool, leave the query blank and run a full word list.

  1. What are the top 10 words in the corpus?
  2. Are they mostly function words or content words? Why?

Tool 8: Keyword List

What Keyword List is for

The Keyword tool compares a target corpus against a reference corpus to find items that are unusually frequent (or, optionally, unusually infrequent) in the target.

Depending on the selected Source, it can generate keywords for words, clusters, n-grams, or collocates.

Keyword List: practical examples

  • Learning: build a keyword list for your assignment corpus, then verify key items in KWIC

Keyword List: core workflow

  1. Load a target corpus and a reference corpus via File → Corpus Manager
  2. Select a Source — Word, Cluster, N-Gram, or Collocate
  3. Enter a query only if you want to filter; otherwise leave it blank
  4. Set “Min. Freq” and “Min. Range”, then click Start
  5. Read “Keyword Types” and “Keyword Tokens”, then inspect the ranked rows

Key idea: a keyword is not simply a frequent word — it is distinctive compared to a reference.

Tool 9: Wordcloud

What Wordcloud is for

The Wordcloud tool turns text or table outputs into a weighted word cloud — each item’s size reflects its value (e.g., frequency).

Sources: a Scratchpad text area, or exported results from KWIC, File View, Cluster, N-Gram, Collocate, Word, and Keyword.

Wordcloud: practical examples

  • Learning: cloud from Word or N-Gram outputs to spot high-salience vocabulary when preparing reports

Wordcloud: core workflow

  1. Select a Source — Scratchpad free text, or another tool’s results
  2. Set the Output label (the word column) and Output value (the size column)
  3. Set Max. Words to control how many items are drawn
  4. (Optional) Paste text into the Scratchpad panel
  5. Click Start and review the cloud

Search strategies

Four ways to search

AntConc offers four search strategies — choosing the right one moves you from quick exploration to detailed pattern analysis.

Strategy When to use it Example
Simple exact word or phrase lookup language, corpus analysis
Wildcard prefix/suffix families, alternatives teach*, *ing, student\|pupil
Regex (token level) precise token-level patterns colou?r, data.* driven
Full Regex multi-word spans, cross-token patterns \bcorpus\s+linguistics\b

The full wildcard list is in Global Settings → Searches.

Part 3: Your own mini research

The task

Work alone or in pairs.

  1. Choose one topic from the list (or propose your own)
  2. Turn it into a question you can answer with AntConc
  3. Run your searches — use at least two different tools
  4. Take notes and screenshots of your key results
  5. Prepare a 2–3 minute report for the class

What to report

Your mini report should answer:

  1. Question — what did you want to find out?
  2. Method — what did you search? Which tools did you use?
  3. Findings — what did the corpus show? (numbers + 1–2 example lines)
  4. Reflection — was the result surprising? What is the limitation of your search?

Topic list: words and phrases

  1. big vs. large — run separate KWIC searches. What kinds of nouns does each appear with?
  2. make vs. do — use Cluster to find common phrase patterns for each verb.
  3. important vs. significant vs. vital — compare their top collocates. Do they mean the same?
  4. -ful vs. -less words (wildcards: *ful, *less) — which ending is more common in this corpus?

Topic list: patterns

  1. Frequent 4-grams — use the N-Gram tool. What formulaic sequences dominate this corpus?
  2. Open-slot patterns — search the * of or a * of. What words fill the slot, and why?
  3. Distribution — pick a key term from this week’s reading and plot it. Is it spread evenly or concentrated?
  4. A word’s behavior — pick a verb and find its strongest collocates (sort by Effect).

Tips before you start

  • Write down your exact search strings — your study should be repeatable
  • Look at Type vs. Token counts, and check Range (how many files an item appears in)
  • Sort collocates by Effect or Likelihood, not just raw frequency
  • Read actual concordance lines — numbers alone can mislead you

Tips before you start

  • Double-click almost anything to jump to KWIC and check it in context
  • A surprising result is a good result — report what you found, not what you expected
  • A zero result can also be a finding: ask why

Research time

You have about 40 minutes to search and prepare your report.

Ask for help any time.

Share your findings

Each person/pair: –10 minutes.

  1. Your question
  2. Your searches
  3. Your findings (show a screenshot if you can)
  4. One thing that surprised you

Wrap-up

  • learned what AntConc is and how to load a corpus
  • walked through the core workflows
  • saw how to build, manage, and compare corpora
  • answered a real question with real language data

More info on AntConc

Please see more detailed documentation in the AntConc help file.

  • Anthony, L. (2026). AntConc (help documentation, July 20, 2026 version) [Computer software documentation]. Waseda University. https://www.laurenceanthony.net/software/antconc/ — PDF in the practice folder.