Find recurring combinations of adjacent words
Sequences counts two-, three-, or four-word combinations in the Active selection. It reveals repeated phrasing while preserving sentence, speaker-turn, and text boundaries, and it lets you open any sequence as an exact phrase in Concordance. This is especially useful for spotting potential collocations in your data.
A sequence is an uninterrupted window of adjacent words
Two, three, or four consecutive word tokens in their original order. No intervening word can be skipped.
One appearance of that complete combination. Repeated appearances contribute to its Frequency.
One distinct combination represented by one result-table row, regardless of how often it occurs.
Choose the scope with Active selection
Sequences uses the shared Active selection. Every count and normalized rate refers only to that scope.
Counts sequences throughout every imported corpus text.
Counts only the selected interviews or documents.
Counts sequences inside turns attributed to the selected speaker codes across their texts.
Counts the combined members of one or more custom text groups.
Counts turns attributed to members of one or more custom speaker groups.
Choose 2, 3, or 4 words and understand the boundaries
Illustration · Adjacent windows overlap. In a two-word analysis, the middle words can participate in more than one sequence.
Also called bigrams. This is the default and usually produces the broadest result set.
Also called trigrams. These capture more specific recurring phrasing.
Four-word sequences are more specific and often less frequent.
Uses the shared project rule. With it off, we’re contributes we and re; with it on, it contributes one word. Changing the rule requires Update Sequences.
Where a sequence must stop
- A sequence never crosses a sentence boundary marked by a period, question mark, or exclamation mark.
- A sequence never joins the end of one speaker’s turn to the beginning of another speaker’s turn.
- A sequence never crosses from one corpus text into another.
- A visual or imported line break alone does not stop a sequence when the same sentence and speaker turn continue.
- Speaker codes and parenthetical comments are excluded before sequences are counted.
- Letter case is combined. The displayed sequence uses normalized lowercase forms.
Generate sequences step by step
Opening the workspace does not begin counting automatically.
Verify the texts, speakers, or groups that answer your research question.
Select 2 words, 3 words, or 4 words.
Decide whether forms containing apostrophes count as one word or separate tokens.
Progress, percentage, and elapsed time remain visible. The button becomes Cancel while counting.
Exempla reports total occurrences and unique sequences above the completed table.
Read the three result columns
Illustration
The complete adjacent word combination in normalized form.
The exact number of occurrences in the generated scope.
Frequency ÷ all sequence occurrences of the selected length × 1,000.
Use sequence frequencies as a hypothesis list
A high-frequency sequence is a useful lead, not automatic proof of a collocation. Use it to identify candidates, then validate them in context.
Use frequency to prioritise phrases, then confirm whether recurrence is stable across texts, turns, and speakers.
Open candidates as exact phrases and review context before interpreting a sequence as a meaningful pattern.
Look for near-regular variants (for example, one extra function word) to avoid over-reading one frozen form.
Filter and sort completed results
Keeps sequences containing the entered characters anywhere. Because this is a text containment filter, a short fragment can also match inside a word.
Keeps sequences beginning with the entered complete word or words. Filtering you know can return you know that, but not you knowledge.
Keeps sequences whose raw Frequency is at least the entered value. The default is 2, which hides combinations occurring only once.
Clear Filter sequences and set Minimum to 1. The cached results reappear without generation.
- Filters are case-insensitive and do not change the completion totals.
- Changing Filter, mode, or Minimum updates the displayed table immediately.
- Export follows the currently displayed rows after filtering.
- Changing sequence length, Active selection, or apostrophe handling requires Update Sequences.
Sort from the headings
First click A–Z; second click Z–A.
First click highest to lowest; second click lowest to highest.
First click highest to lowest. Its order matches Frequency within one completed analysis.
Sorting applies to all displayed results, not only rows currently visible on screen.
Double-click a sequence to find its occurrences
Double-click any row. Exempla opens Concordance and runs an Exact phrase search for that sequence in the current Active selection.
Choose the recurring combination you want to inspect.
Concordance finds the adjacent words in the same order.
Review speaker, context, source text, and transcript line.
Export sequences as CSV
The CSV contains Sequence, Frequency, and Per 1,000.
Only displayed rows are exported. Clear the filter and use Minimum 1 for the complete generated table.
Exempla proposes a date-first filename containing the project name and Sequences.
At least 5,000 rows use a responsive progress window with Cancel. Cancelling leaves no incomplete CSV.
When sequence results are unexpected
Confirm that the scope contains enough adjacent spoken words for the selected length.
Scope, length, or apostrophe handling changed. Choose Update Sequences to recount.
Check whether it crosses a sentence, speaker turn, or text boundary, or contains a parenthetical comment.
Minimum defaults to 2. Set it to 1 to include sequences occurring once.
Contains can match inside words. Choose Begins with for a complete initial word or phrase.
Clear Filter sequences and set Minimum to 1.
Confirm the same Active selection and apostrophe rule, and turn off Include comments in Concordance.
The CSV follows the filtered table. Reset Filter and Minimum before exporting everything.
Was this chapter helpful?
Your response will help improve the guide.