Text Lens — Browser-Based NLP Text Analysis

Research & Writing Tool · Browser-Based NLP

Text Lens

Inspect structure, readability, lexical patterns, repeated phrases, and a lightweight sentiment signal. Compare two texts side by side without sending either text to a server.

Private by design · analysis stays in your browser







Text A0 words
0 charactersAnalysis updates automatically
.txt, .md, or .csv · files stay in this browser

Concordance & corpus tools

Local analysis
Search a word or phrase in context, adjust the context window, and inspect collocations that recur more often than their component-word frequencies would suggest.

Manage loaded documents

Documents stay grouped by Text A/Text B. Reordering changes display/export order within a group; removing a document rebuilds that text from the remaining loaded files. Focus and custom-reference roles are labeled beside each document.

Enter a term or phrase to see up to 50 concordance lines.

Load one or more local files to see document-level corpus statistics.

Multiple local files are concatenated for whole-text analysis without inserting filenames into the analyzed text, while document identity is retained for corpus comparison. You can reorder/remove loaded documents, focus one document, select an explicit or custom multi-document keyness reference, and save/restore a source-free corpus setup. Concordance is exact-token matching after case normalization; it is not stemming or semantic search.

Import Word or PDF

Extract selectable text from a local .docx or PDF and analyze it in Text Lens. Files are read in this browser; scanned/image-only PDFs require OCR and are not supported. Review extracted text for lost columns, tables, or footnotes.

No document selected. Imports replace the selected text field.

Tone and rhetorical signals

Explore signals that contribute to a passage's rhetorical stance: hedging, certainty, intensification, direct address, questions, and affective vocabulary. These are transparent word-pattern counts, not diagnoses of author intent or automated assessments of literary quality.

Interpretation depends on genre, quotation, irony, negation, dialect, and context. Counts include quoted language and are not semantic classification. Use the concordance above to inspect occurrences in context.

Word frequency across the text

Track where a word appears in equal-length sections of a text. Counts are normalized per 1,000 words so sections can be compared. This measures distribution, not meaning.

Enter a word to trace its distribution.

Most frequent terms

Method note. Readability formulas and syllable counts are English-language heuristics. The sentiment panel is a transparent lexicon-and-negation signal, not a trained sentiment model and not a judgment about meaning, tone, quality, or intent. Frequency, collocation, concordance, and similarity measures describe surface lexical patterns; interpretation remains with the reader. PMI can over-emphasize rare pairs, so Text Lens requires a user-selected minimum occurrence count. Cross-document distinctive terms use signed log-likelihood G² with a separate minimum-frequency floor and can compare against the remainder of the corpus, one document, or a custom reference set; lexical similarity is not semantic similarity. Saved corpus setups contain settings and file metadata/order, not source text.