Curate before you cluster: LLM-consensus filtering for marketing text analytics

Purpose Social listening is a core market-sensing capability, yet firms and researchers often struggle to isolate the conversation they intend to analyze. During high-attention events, campaign-related discourse becomes intertwined with broader political, cultural, and brand-related discussion, creating a corpus construction problem that can distort downstream analytics. This paper examines how firms can construct event-bound listening corpora while maintaining semantic coherence and transparency. Design/methodology/approach We propose a three-stage workflow that combines (1) multi-model large language model (LLM) consensus labeling to establish contextual judgments of relevance, (2) classifier distillation to create a scalable high-precision seed set, and (3) similarity-governed semantic expansion to recover additional relevant discourse while controlling semantic drift. The approach is demonstrated using 73,564 Reddit posts surrounding Nike’s 2018 Dream Crazy campaign. Findings The proposed workflow substantially improves retrieval quality while maintaining low contamination. Human audits and temporal falsification converge on an empirically observable semantic boundary, showing that corpus construction is fundamentally a boundary-management problem. Effective social listening depends on governing the trade-off between coverage and contamination before downstream analyses are conducted. Practical implications The framework provides managers with a transparent procedure for defining what consumer chatter “counts” when monitoring campaigns, controversies, crises, or other high-velocity events. This improves the reliability of listening dashboards and reduces the risk of basing decisions on semantically contaminated data. Originality/value By combining contextual LLM judgments with explicit validation and threshold governance, the paper advances a design-oriented perspective on AI-assisted listening in interactive marketing.

Authors

Institutions

Publication Details

Journal
Journal of Research in Interactive Marketing
Published
2026-11-16
DOI
https://doi.org/10.1108/jrim-11-2025-0719
Primary Topic
Digital Marketing and Social Media
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Curate before you cluster: LLM-consensus filtering for marketing text analytics

Lavy Khoushinsky
3 links
Journal of Research in Interactive Marketing
Digital Marketing and Social Media
article

Curate before you cluster: LLM-consensus filtering for marketing text analytics

Lavy Khoushinsky
article en

Abstract

Purpose Social listening is a core market-sensing capability, yet firms and researchers often struggle to isolate the conversation they intend to analyze. During high-attention events, campaign-related discourse becomes intertwined with broader political, cultural, and brand-related discussion, creating a corpus construction problem that can distort downstream analytics. This paper examines how firms can construct event-bound listening corpora while maintaining semantic coherence and transparency. Design/methodology/approach We propose a three-stage workflow that combines (1) multi-model large language model (LLM) consensus labeling to establish contextual judgments of relevance, (2) classifier distillation to create a scalable high-precision seed set, and (3) similarity-governed semantic expansion to recover additional relevant discourse while controlling semantic drift. The approach is demonstrated using 73,564 Reddit posts surrounding Nike’s 2018 Dream Crazy campaign. Findings The proposed workflow substantially improves retrieval quality while maintaining low contamination. Human audits and temporal falsification converge on an empirically observable semantic boundary, showing that corpus construction is fundamentally a boundary-management problem. Effective social listening depends on governing the trade-off between coverage and contamination before downstream analyses are conducted. Practical implications The framework provides managers with a transparent procedure for defining what consumer chatter “counts” when monitoring campaigns, controversies, crises, or other high-velocity events. This improves the reliability of listening dashboards and reduces the risk of basing decisions on semantically contaminated data. Originality/value By combining contextual LLM judgments with explicit validation and threshold governance, the paper advances a design-oriented perspective on AI-assisted listening in interactive marketing.

Journal of Research in Interactive Marketing
Université Laval (CA)
Digital Marketing and Social Media
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon