DMA Discourse/pragmatic markers

Definition

Discourse markers and pragmatic particles: well, okay, actually, anyway, right, etc.

Detection Rules

Default discourse markers rule (based on MFTE). Refines RB.

p1
cql[word={words} & pos!="VB|VBD|VBG|VBN|VBP|VBZ"]
p2
cql[word="right" & pos="UH"]
p3
cql[word="right" & pos="JJ"]
p4
cql[word="well" & pos="UH"]
p5
cql[word="no" & pos="UH"]
p6
cql[word="sure" & pos="JJ|RB"]
p7
cql[word="of"] [word="course"]
combine: p1 | p2 | p3 | p4 | p5 | p6 | p7
Word list (13 items)
anyway anyways damn goodness gosh lol nope omg whatever wtf yeah yep yes

Requires: word, pos

Refines: RB

mfte

MFTE discourse/pragmatic markers. Two groups: (1) simple word list always tagged DMA, (2) context-sensitive patterns (sure, right, of course, mind you). “now” excluded because MFTE tags it as TIME first. “anyhow” excluded because MFTE tags it as RBother, not DMA. Anchor on “right” (not “all”) and “course” (not “of”) to match MFTE’s anchoring convention.

p1
cql[word={words} & pos!="VB|VBD|VBG|VBN|VBP|VBZ"]
p2
cql[word="right" & pos="UH"]
p3
cql[word="right" & pos="JJ"]
p4
cql[word="well" & pos="UH"]
p5
cql[word="no" & pos="UH"]
p6
cql[word="sure" & pos="JJ|RB"]
p7
cql[word="of"] [word="course"]
combine: p1 | p2 | p3 | p4 | p5 | p6 | p7
Word list (13 items)
anyway anyways damn goodness gosh lol nope omg whatever wtf yeah yep yes

Requires: word, pos

pybiber

pybiber replaces all tokens with dep_rel == “punct” with _punct_ in its blob, so any punctuation (including quotation marks) qualifies as the preceding token.

cql[dep="punct"] [word="well|now|anyhow|anyways"]

Requires: word, pos, dep

Normalization

Per words

Examples

Well, that changes everything.

Now, where were we?

Well no they didn’t say actually.

Source: le_foll_2024

Okay I guess we’ll see how things go right?

Source: le_foll_2024

Non-examples

She now understood the full extent of the problem.

now = time adverb, not discourse marker — no longer a false positive (mfte rule excludes “now”)

The well had dried up during the summer.

well = noun (a well) — no longer triggers detector (mfte rule requires UH POS)

Sources

  • biber_1988 — Biber, Douglas (1988) : Variation across Speech and Writing
  • mfte — Le Foll, Elen & Shakir, Muhammad (2023/2025) : Multi-Feature Tagger of English (MFTE) — Python version
  • pybiber — Brown, David West & Reinhart, Alex (2026) : pybiber — Python package for linguistic feature extraction and Multi-Dimensional Analysis
  • xiao_2009 — Xiao, Richard (2009) : Multidimensional analysis and the study of world Englishes
  • grieve_2023 — Grieve, Jack (2023) : Register variation explains stylometric authorship analysis