ConductSpeech

Validation and source transparency

SALT-Compatible Analysis Methodology

ConductSpeech uses documented SALT-style conventions, regression tests, and source-aware norm reporting to make language sample analysis easier to review. This page explains what is validated, what references inform the work, and where clinicians should still apply judgment.

What ConductSpeech validates

The SALT-compatible workflow is validated as a product system, not as a claim that ConductSpeech is the SALT desktop product. The goal is to support the conventions clinicians expect while keeping the transcript editable and the metric outputs visible before report generation.

Transcript conventions

Speaker lines, mazes, pauses, omissions, errors, unintelligible words, and overlap markers are parsed into reviewed transcript data.

C-units and SI

C-unit segmentation and clause counts are checked against reference examples and surfaced in the dashboard and clinical reports.

Maze summaries

Parenthesized maze material is counted separately and excluded from core word and morpheme totals.

Verbal facility

Words per minute and pause summaries are computed only when word-level timing exists.

Protocol and norms

Samples preserve protocol and grade context, and norm comparisons disclose source and confidence.

Source transparency for norms

Norm comparisons are only useful when the user understands what comparison is being made. ConductSpeech shows the selected age, grade, protocol, expected range, observed score, source, and confidence level. If an exact grade and protocol reference is not available, the interface explains that it is using the closest supported comparison.

This wording matters for clinical trust. User-facing pages and reports should describe the practical meaning of each comparison: exact match, closest supported comparison, descriptive comparison, or unavailable.

Product validation

Recent QA verified a real user flow through typed sample entry, persuasion protocol selection, grade norm display, dashboard metric review, SALT editor handoff, suggestion acceptance, re-analysis, and report generation. The smoke test confirmed SI, C-units, clauses, maze summary, and verbal facility status appear where a clinician would expect them.

Important limitations

  • ConductSpeech is SALT-compatible, not SALT-certified or SALT-licensed.
  • Verbal facility requires timed audio with word-level timestamps.
  • Automated C-unit, clause, and maze results should be reviewed on clinically complex samples.
  • Norm comparisons depend on available age, grade, and protocol reference data.
  • Reports are drafts for clinician review, not automatic clinical sign-off.

References and source material