Skip to content

Update politeness v2 #367

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
xehu merged 10 commits into from
Jan 12, 2026
Merged

Update politeness v2 #367

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
c3b4422
sync politeness updates
sundy1994 Dec 16, 2025
b388a32
add new WH/yesno questions to yeomans_test dataset
sundy1994 Dec 16, 2025
14dc294
Refactor Yes/No question detection and improve WH question handling i…
sundy1994 Jan 8, 2026
125d821
add 'am' to WH followers
xehu Jan 9, 2026
a6e99a7
remove commented-out thank you
xehu Jan 9, 2026
183d1e6
add am everywhere
xehu Jan 9, 2026
bafe275
update to question detection
xehu Jan 9, 2026
40aee79
an expanded and more robust version of the question function (with a …
xehu Jan 10, 2026
8e15934
fixing more edge cases
xehu Jan 10, 2026
a011781
update robustness of question detector and update tests
xehu Jan 11, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,4 +58,5 @@ node_modules/
# testing
/output
/vector_data
test.py
test.py
test.ipynb
57 changes: 53 additions & 4 deletions src/team_comm_tools/features/keywords.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
# reference: https://github.com/bbevis/politenessPy/blob/main/keywords.py
kw = {
"spacy_neg_only": {
"Negative_Emotion": [
Expand Down Expand Up @@ -7260,7 +7261,34 @@
" sorry ",
" woops ",
" whoops ",
" oops "
" oops ",
" apology "
],
"Third_Person": [
" he ",
" him ",
" his ",
" himself ",
" she ",
" her ",
" hers ",
" herself ",
" they ",
" them ",
" their ",
" theirs ",
" themselves "
],
"Contrast_Conjunction": [
" but ",
" however ",
" instead ",
" although ",
" even though ",
" despite ",
" and yet ",
" nevertheless ",
" nonetheless "
],
"Ask_Agency": [
" do me a favor ",
Expand Down Expand Up @@ -7365,7 +7393,6 @@
"Gratitude": [
" thank ",
" thanks ",
" thank you ",
" grateful ",
" gratitude ",
" cheers "
Expand Down Expand Up @@ -14419,25 +14446,47 @@
" cock ",
" crap ",
" damn ",
" dammit ",
" damnit ",
" dick ",
" dickhead ",
" dick-head ",
" dumb ",
" dumbass ",
" dumb-ass ",
" dumb ass ",
" dyke ",
" fuck ",
" fucking ",
" fucker ",
" goddam ",
" goddammit ",
" goddamed ",
" hell ",
" horshit ",
" homo ",
" jackass ",
" jackass ",
" motherfucker ",
" mother-fucker ",
" motherfucking ",
" nigger ",
" nigra ",
" piss ",
" prick ",
" pussy ",
" queer ",
" screw ",
" shit ",
" shite ",
" shitting ",
" sob ",
" sonofa ",
" suck ",
" sucked ",
" sucks "
" sucks ",
" twat ",
" wanker ",
" whore "
],
"Truth_Intensifier": [
" really ",
Expand Down
86 changes: 68 additions & 18 deletions src/team_comm_tools/features/politeness_v2_helper.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -218,29 +218,79 @@ def bare_command(doc):
def Question(doc):
"""
Counts the number of sentences containing question words and question marks.

Reference: https://github.com/bbevis/politenessPy/blob/main/strategy_extractor.py
Args:
doc (spacy.tokens.Doc): The spaCy Doc object containing the text to be analyzed.

Returns:
tuple: A tuple containing the counts of Yes/No questions and WH-questions.
"""

keywords = set([' who ', ' what ', ' where ', ' when ', ' why ', ' how ', ' which '])
tags = set(['WRB', 'WP', 'WDT'])

# doc = nlp(text)
sentences = [str(sent) for sent in doc.sents if '?' in str(sent)]
all_qs = len(sentences)

n = 0
for i in range(len(sentences)):
whq = [token.tag_ for token in nlp(sentences[i]) if token.tag_ in tags]

if len(whq) > 0:
n += 1

return all_qs - n, n
# POS tags for WH-words like who/what/where
search_tags = {'WRB', 'WP', 'WDT'}
# WH-words and common auxiliaries that follow them in real questions
wh_words = {'what', 'who', 'where', 'when', 'why', 'how', 'which'}
wh_followers = {
'what': {'are', 'is', 'do', 'does', 'can', 'should', 'might'},
'who': {'am', 'is', 'are', 'was', 'can', 'should'},

@xehu xehu Jan 9, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@sundy1994 I found a bug when testing this. It turns out that if you do not add 'am' to WH_followers, things like who am I or what am I supposed to do are not detected as WH_questions. But there are still some bugs with this... I'll follow up on Slack.

@xehu xehu Jan 10, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

UPDATE - I realized that it actually doesn't make much sense to separate the wh_followers from the other auxiliaries, so I refactored this so that we only use a consistent set of auxiliaries

'where': {'is', 'are', 'can', 'should'},
'when': {'is', 'are', 'can', 'should'},
'why': {'is', 'are', 'do', 'does', 'can', 'might', 'would'},
'how': {'is', 'are', 'do', 'does', 'can', 'should', 'would'},
'which': {'is', 'are', 'was', 'can', 'should'}
}
# Auxiliaries that typically initiate Yes/No questions
yesno_aux = {
'do', 'does', 'did', 'have', 'has', 'had',
'can', 'could', 'will', 'would',
'may', 'might', 'shall', 'should',
'is', 'are', 'was', 'were', 'am'
}
# Pronouns that often follow auxiliaries in Yes/No questions
pronoun_followers = {'i', 'you', 'we', 'he', 'she', 'they', 'it'}

wh_count = 0
yesno_count = 0
counted_sentences = set()
for sent in doc.sents:
sent_text = sent.text.strip()
sent_tokens = list(sent)
if not sent_tokens:
continue
# Method 1: Find question sentences by checking for '?' at end

@xehu xehu Jan 7, 2026

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So there's a bug with this logic, which is that, if the sentence doesn't end with a question mark, the fallback method is very error prone. For example:

Question(nlp("can you tell me what is your name?")) yields YesNo = 1, WH = 0 (correct)
Question(nlp("can you tell me what is your name")) yields YesNo = 0, WH = 1 (incorrect)

Why can't we use logic that looks for the question words, but then applies the logic with the search_tags, which helps to get around some of these problems?

if sent_text.endswith('?'):
# try to find the first WH-word in the sentence
wh = False
for token in sent_tokens:
if token.text.lower() in wh_words and token.tag_ in search_tags and token.dep_ not in {"relcl", "acl"}\
and token.i < sent.root.i:
wh = True
break
if wh:
wh_count += 1
else:
# Fallback: no WH in the sentence → treat as Yes/No question
yesno_count += 1
counted_sentences.add(sent.start)
continue
# Method 2: For remaining sentences, apply lexical rule-based detection --- Extract tokens and their metadata for fast access
for i in range(len(sent_tokens) - 1):
tok1 = sent_tokens[i]
tok2 = sent_tokens[i + 1]
t1_lower = tok1.text.lower()
t2_lower = tok2.text.lower()
if sent.start in counted_sentences:
break # already counted
# Yes/No pattern
if t1_lower in yesno_aux and t2_lower in pronoun_followers:
yesno_count += 1
counted_sentences.add(sent.start)
break
# WH pattern
if t1_lower in wh_words and tok1.tag_ in search_tags and tok1.dep_ not in {"relcl", "acl"}\
and tok1.i < sent.root.i and t2_lower in wh_followers.get(t1_lower, set()):
wh_count += 1
counted_sentences.add(sent.start)
break
return yesno_count, wh_count


def word_start(keywords, doc):
Expand Down
8 changes: 7 additions & 1 deletion tests/data/cleaned_data/test_chat_level.csv
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1437,11 +1437,17 @@ yeomans_test,yeomans_user_b,I guess they almost complete it.,Hedges_receptivenes
yeomans_test,yeomans_user_a,"Ma'am, no my lady, do you know Mr. Smith?",Formal_Title_receptiveness_yeomans,2
yeomans_test,yeomans_user_b,"I agree, this is correct",Agreement_receptiveness_yeomans,2
yeomans_test,yeomans_user_a,This is for you. Why you don't understand it's for you?,For_You_receptiveness_yeomans,2
yeomans_test,yeomans_user_b,"Thank you, I'm really grateful",Gratitude_receptiveness_yeomans,3
yeomans_test,yeomans_user_b,"Thank you, I'm really grateful",Gratitude_receptiveness_yeomans,2
yeomans_test,yeomans_user_a,We here you. We totally understand,Acknowledgement_receptiveness_yeomans,2
yeomans_test,yeomans_user_b,"Shit. You dumb asshole, what the hell? Who's that bastard? Suck my dick.",Swearing_receptiveness_yeomans,7
yeomans_test,yeomans_user_a,"Hey hello good morning, oh actually good evening.",Hello_receptiveness_yeomans,4
yeomans_test,yeomans_user_b,Are you sure? Is this the guy? Did he lie to you?,YesNo_Questions_receptiveness_yeomans,3
yeomans_test,yeomans_user_a,"Did you finish the report, which was due today?",YesNo_Questions_receptiveness_yeomans,1
yeomans_test,yeomans_user_b,"Did you finish the report, which was due today?",WH_Questions_receptiveness_yeomans,0
yeomans_test,yeomans_user_a,"We can start here. What is the question?",WH_Questions_receptiveness_yeomans,1
yeomans_test,yeomans_user_b,"Has she met the teacher who helped you last year?",WH_Questions_receptiveness_yeomans,0
yeomans_test,yeomans_user_a,"Do you know what time it is?",WH_Questions_receptiveness_yeomans,0
yeomans_test,yeomans_user_b,"Have you read the article that explains why this happens?",WH_Questions_receptiveness_yeomans,0
yeomans_test,yeomans_user_a,I'm sorry I sincerely apologize.,Apology_receptiveness_yeomans,2
yeomans_test,yeomans_user_b,Wow! Amazing! Perfect!,Affirmation_receptiveness_yeomans,3
yeomans_test,yeomans_user_a,I love you. My friend,First_Person_Single_receptiveness_yeomans,2
Expand Down