corpus-discovery-dialogue

What it does

Guides users through a multi-phase, Socratic dialogue to turn vague interest in a text corpus into concrete research questions and an analytical roadmap. It structures inquiry into Phase 0 (context and sample review), Phase 1 (corpus understanding with targeted questions), Phase 2 (interest exploration to surface tacit goals and hypotheses), and Phase 3 (formulation of candidate research questions). It prompts for sample texts, corpus characteristics (type, size, time span, authorship, format, metadata), and data readiness, then syntheses outputs like a Corpus Profile and an Interest Map before producing Candidate Research Questions. It relies on structured prompts and scripted blocks to elicit specifics, wait for user responses at each step, and require user confirmation before advancing. The process emphasizes phases, checkpoints, and explicit grounding in actual texts, with announcements and phase descriptions included for clarity. The skill is designed to be invoked when starting corpus analysis with unclear research questions or when structuring exploratory research, and it asks for optional inputs such as corpus_file_path and domain_context.

How it works

• Initiates with a start-announcement describing five phases and expected duration (20-40 minutes).
• Phase 0: Context Establishment & Sample Review:
  • Requests 3-5 sample texts if not provided.
  • Reviews samples to extract initial observations across length, style, content type, terminology, temporal markers, and structure.
  • Presents initial observations in a structured template and waits for user validation before Phase 1.
• Phase 1: Corpus Understanding (5-10 minutes):
  • Asks a sequence of targeted questions about basic characteristics, corpus size, temporal span, authorship, format, and data metadata.
  • Each sub-question has defined answer options (A–F, etc.) and prompts for confirmation before proceeding.
  • Outputs a structured "Corpus Profile" synthesis after Phase 1 concludes with user confirmation.
• Phase 2: Interest Exploration (5-10 minutes):
  • Probes motivation, puzzles, surprises, disappointment, and stakes through open-ended prompts.
  • Generates an "Interest Map" capturing primary draws, puzzles, surprises, disappointments, stakes, and implicit hypotheses.
• Phase 3: Question Formulation (5-10 minutes):
  • Prompts to present 8-12 candidate research questions organized by type, based on the Corpus Profile and Interest Map.
• Output: Each phase yields structured sections (e.g., Corpus Profile, Interest Map) and culminates in Candidate Research Questions tailored to corpus size, temporal span, authorship, and interests.

When to use it

• Invocation phrases include: "Help me explore this corpus", "What questions should I ask of these [texts/reviews/documents]?", "Guide me through corpus analysis", "I have text data but don't know where to start", "Design a research plan for this corpus".
• Suitable for starting corpus analysis with unclear questions or needing a structured exploratory framework.
• Skip Phase 1–3 when research questions are already clearly defined or when using other specialized dialogues (e.g., hypothesis-testing-dialogue, analysis-interpretation-dialogue).

What it can touch

• Requires: Optional inputs corpus_file_path, domain_context, research_constraints.
• Data touches include: corpus samples (3-5 examples), metadata fields (dates, categories, author info), and format (JSON, CSV, TXT, PDF, DOCX, HTML).
• It prompts for and relies on user-provided corpus descriptions, sample texts, and metadata availability to tailor profile and questions.

Caveats

• Based on user-provided responses; outputs depend on accuracy and completeness of inputs.
• The process emphasizes discovery before execution and requires sequential confirmation to proceed between phases.
• No outcomes are promised beyond formulation of questions and a roadmap; actual analysis results depend on later steps and data readiness.