Accepted paper · ASONAM 2026
Social computing · NLP · India
A Social and Legal Discourse Analysis of Women's Safety Narratives in India
A longitudinal analysis of how Reddit and YouTube discussions assign responsibility, support victims, and preserve public memory across major women's-safety cases.
RQ1 · Triggering incidents
Political implication changes who receives blame.
Mann–Whitney U = 59p = 0.0005r = 0.786Large case-level effect
Political implication Non-political
RQ2 · Evolving narratives
Attention shifts, but not along one universal curve.
Among 15 eligible cases with at least 20 rows in both windows, the acute-to-retrospective change was not statistically uniform.
Critical of government Critical of society Neutral reporting Supportive of victim
General baseline
comparison corpusAcute
0–90 daysSustained
91–365 daysRetrospective
366+ daysRQ3 · Institutional response
Association map, not a causation claim.
The response taxonomy is descriptive. It does not establish that online discourse caused policy or legal action.
Legislative
Nirbhaya 2012 · Kathua 2018
Regulatory / state action
Uber Delhi 2014 · Hyderabad Vet 2019 · RG Kar 2024
Executive / judicial only
Remaining cases in the paper taxonomy
Post-submission local audit
Adaptation improved classification, not every reliability measure.
This panel reports the later fixed 3,000-row validation matrix. It is deliberately separated from the submitted paper's pilot results.
Notes
Evidence and limitations
Corpus scale, 16-case framing, 2012–2024 span, paper-facing tests, and annotator agreement mirror the accepted manuscript.
The model audit uses later reproducibility artifacts and is labeled separately so it does not rewrite the submitted pilot.
Only aggregate statistics appear here. Raw usernames, comment text, and individual-level records are not published in this dashboard.
Platform participation is not representative of India as a whole. Observed discourse–response patterns are associations, not causal estimates.