C6Text analysisStylisticLayer 2 (Contextual)

Local Coherence

Detects unnatural paragraph-to-paragraph transitions that feel mechanically connected rather than reflecting genuine logical flow between ideas.

Technical description

Analyzes inter-paragraph coherence by examining: entity chains (do subjects carry over between paragraphs?), lexical overlap patterns, discourse connective usage, and topic continuity. Uses dependency parsing to track argument structure across paragraph boundaries. Flags transitions that rely solely on generic connectives ('Furthermore', 'Moreover', 'Additionally') without substantive topic bridging.

How it works

Layer 2 (NLP): Parses each paragraph's main topic using dependency analysis. Tracks entity chains across paragraph boundaries. Measures topic overlap between consecutive paragraphs. Detects reliance on generic transition words without substantive connections. Flags sections where paragraphs share transitions but not topics.

Why this matters

AI-generated text often connects paragraphs using surface-level transition words without genuine topical continuity. The ideas in consecutive paragraphs may be loosely related but lack the tight argumentative threading that characterizes human academic writing. This creates a 'beads on a string' effect where paragraphs could be reordered without affecting perceived coherence.

Score thresholds

0-1: Strong topical continuity with natural transitions
2-3: Some mechanical transitions mixed with genuine flow
4-5: Paragraphs connected only by generic transition words

Limitations

Literature review sections naturally shift topics more frequently. Short paragraphs may not provide enough context for entity chain analysis. Some academic styles favor explicit transition markers, which may be flagged as formulaic.