# Trigger-Based Language Modeling **Domain:** Statistical Natural Language Processing / Language Modeling **Doc Type:** Canonical Mechanism Node **Maturity:** Developed **Related:** [[wiki/Language Model|Language Model]], [[wiki/Distributional Semantics|Distributional Semantics]], [[wiki/Switchboard Corpus|Switchboard Corpus]], [[wiki/Broadcast News Corpus|Broadcast News Corpus]], [[wiki/Living Language — Proximal Frequency Research Reference|Living Language — Proximal Frequency Research Reference]] --- ## Definition **Trigger-Based Language Modeling uses long-range lexical associations to adjust the probability of a word after a related word has appeared in the discourse.** A trigger pair need not be adjacent or synonymous; the earlier term changes the later term's relevance over a wider context. ## Historical Context Trigger models were developed in early statistical language-modeling work, including research associated with Roni Rosenfeld and colleagues at Carnegie Mellon in the early 1990s. They extended short local histories by carrying topical information across longer spans. ## Living Language Context The source files described in [[projects/Ten Years Building a Symbolic Language Engine|Ten Years Building a Symbolic Language Engine]] appear consistent with trigger data derived from [[wiki/Switchboard Corpus|Switchboard]] and [[wiki/Broadcast News Corpus|Broadcast News]]. That identification remains a reconstruction until toolkit documentation is recovered. [[wiki/Living Language — Proximal Frequency Research Reference|Living Language — Proximal Frequency Research Reference]] repurposed trigger relations from prediction into traversal. Instead of asking only which word becomes more probable next, it asked whether weighted associations could trace human conceptual movement through language. ## Key Insight **A predictive relation can become a navigational relation when the system exposes it as a path rather than hiding it inside a probability.** ## See Also [[wiki/Language Model|Language Model]], [[wiki/Distributional Semantics|Distributional Semantics]], [[wiki/Corpus Engineering|Corpus Engineering]], [[wiki/Switchboard Corpus|Switchboard Corpus]], [[wiki/Broadcast News Corpus|Broadcast News Corpus]], [[wiki/Synset vs Concept Set|Synset vs Concept Set]] ## Sources / Provenance - Roni Rosenfeld and colleagues, early work on trigger-based language models. - Project corpus remnants and filename reconstruction.