This preprint presents three linked pilot studies on relationship-relevant text as an empirical bridge to a prior conceptual framework on recurrence-aware conversation decision support. The studies move from adjacent-turn dialogue carryover, to emotional and interactional structure in relationship-help posts, to coarse episode-level trajectory classification in update-style relationship narratives. Across all three levels, simple lexical and n-gram methods recover interpretable but limited signal. The paper is foundational and does not directly test the full conceptual framework.
Mariia Yakovleva (Fri,) studied this question.