LITERATURE · EVIDENCE · INTERPRETATION
Read what stands out.
A working desk for parallelism and possible deviation across eight levels. Every candidate needs close reading; literary meaning is yours to interpret.
01 / PREPARE
Your text, your context.
Selectable-text PDF only; no OCR. PDF extraction can alter line breaks and reading order. CoNLL-U adds supplied grammatical evidence.
Without a comparable reference, “deviation” means a local change or a cue to inspect—not a claim about a language-wide norm.
Pronunciation hints (advanced)
Built-in English hint lexicon is intentionally small. Add tab-separated rows: word, initial sound, rhyme key. Example: night[TAB]n[TAB]aɪt. Dialect and stress still need review.
Desktop ceiling: 5 million characters / 750,000 words / 100,000 nonempty lines per text. Mobile/low-memory: 2 million / 300,000 / 40,000. Findings retained: 5,000. Analysis runs in a worker; actual capacity depends on device memory.
02 / OBSERVE
Foregrounding workspace.
Run analysis to inspect evidence.
03 / INTERPRET
Close reading & notes.
Select “Locate in text” on a finding to highlight its line. All evidence offsets refer to the extracted / pasted text, not PDF page coordinates.
Your text will appear here after analysis.
Local session only. Reports include your notes and review decisions. Rejected candidates remain documented.
Download prepared file04 / KNOW THE METHOD
A companion to interpretation.
Foregrounding involves patterns of parallelism and deviation in relation to a norm. Counts and surface patterns can guide close reading; they cannot establish literary effect or authorial intent.
Scope: Rule-based English cues, not an AI critic or a full linguistic parser. Unicode text is accepted, but English cue lists do not generalize to all languages. For automatic syntax/grammar patterns supply CoNLL-U annotations. Phonological detection uses only words with pronunciation hints; no automatic phonemic transcription, stress or meter. Semantic and pragmatic results require human interpretation.
Descriptive frequency contrasts are not significance tests. Reference differences can reflect topic, genre or corpus size. Low-frequency words are excluded from automatic reference contrasts. Units follow punctuation and line breaks, so abbreviations can create extra boundaries.
Conceptual reading: Miall & Kuiken (1994), Foregrounding, defamiliarization, and affect; Lancaster University foregrounding checksheet. These sources inform the approach; they do not validate this implementation.