Skip to content

Changelog

Major changes to TCSE by version. Maintenance releases that do not change search functionality may be omitted.

v13

Version 13 adds a layer of typed abstraction to Advanced Search: noun chunks are no longer just placeholders, but slots you can constrain and cross-reference. See Advanced Search Query Syntax for full details.

  • Typed noun-chunk slots (v13.1.0) — _{COND} constrains a chunk by its head (root) token: _{pron} matches chunks headed by a pronoun, _{-pron} its exact complement. Example: [give|send|tell] _{pron} _ retrieves ditransitives with pronominal recipients.
  • Entity-typed chunk conditions (v13.2.0) — _{%PERSON} matches a chunk headed by a PERSON entity; types combine with | (_{%PERSON|%ORG}) and mix with other conditions (_{pron|%PERSON} for "a human referent").
  • Lemma echo (v13.3.0, extended in v13.4.0) — Bind a slot's lemma with :N and refer to it later with =N, so that cross-slot identity becomes a single query: {noun:1} [after] _{=1} retrieves study after study, and {adv:1} [and] {=1&adv} retrieves over and over.
  • Chunk-head lemma conditions (v13.5.0) — _{[life]} matches chunks whose root lemma is life, so cognate objects come out directly: [live]{verb} _{[life]}.
  • Chunk-head lemma disjunction (v13.6.0) — the lemma condition accepts alternatives: _{[part|role]} matches chunks headed by part or role, so [do|play|take]{verb} _{[part|role]} covers light-verb pairs in one query. _{-[part|role]} matches chunks headed by neither.
  • Stable order of tied collocations (v13.6.2) — rows of a collocation table that tie on the selected measure are ordered by frequency, then alphabetically. Their order used to depend on the server software and changed when the server moved to Ruby 3.4.
  • Literal % and _ (v13.6.3) — a % typed in a plain, translation or talk search is matched as that character, and so is a _ where the search accepts it (translation and talk searches, the construction list; a plain search still refuses _). * remains the wildcard. Before, % matched any characters, so 50% also found 50 followed by anything.
  • Hash-based random order and faster search (v13.7.0) — plain searches run several times faster on common words. The random order is now computed from the seed and the result ID: the rows shown for a given seed differ from earlier versions, and adding talks no longer reorders existing results.
  • Advanced Search counts rows (v13.8.0) — the total counts segments or sentences, as in a plain search, instead of matches. A row with several matches counts once and shows all of them highlighted, so totals are smaller than before (by up to about a tenth in sentence mode) and the last pages are no longer empty. Exports record count_unit as segment or sentence.
  • Saved settings (v13.9.0) — the settings panel and the list or KWIC view are saved in the browser and restored on the next visit. Saved settings in use marks settings that differ from the defaults, and Reset to defaults clears them.

v11–v12

User Interface

  • Multilingual UI — Interface available in English, Japanese (日本語), Chinese (中文), and Korean (한국어). Use the language buttons on any page to switch.
  • Streamlined search workflow — The RESET button has been removed. You can now start a new search at any time by simply entering a query and clicking SEARCH, just like a regular search engine.

New Search Features

  • KWIC concordance view — Toggle between list and KWIC (Key Word In Context) display modes for linguistic analysis
  • Named entity (NER) search — Use %PERSON, %ORG, %GPE, %DATE, etc. in Advanced Search to find named entities
  • Collocation analysis — MI, t-score, and DP statistics with lemma-based grouping in Collocation mode
  • Collocation network — Interactive force-directed graph visualization of word co-occurrence relationships
  • Export search results — Download token search results as TSV (ZIP) or JSON with context, linguistic annotations, and translations for research use

Collocation & Visualization

  • NER-enhanced n-grams — Named entities replaced with %TYPE labels (e.g., %PERSON said, in %GPE) for entity-aware frequency analysis
  • Network zoom slider — Zoom control (50%–200%) replaces scroll-based zooming for more precise graph navigation
  • Text highlight default off — Keywords and discourse markers highlighting is now off by default for a cleaner transcript view

Construction & Data

  • Construction category filter — Patterns categorized into Phrasal Verb, Idiom, Grammatical, Collocation, Discourse, and NER Pattern, with filter UI
  • 1,167 construction patterns — Including pseudo-cleft, resultative, NER patterns, adjective-subjectivity constructions, and other linguistic structures
  • 6,419 TED Talks — Over 1,100 new talks added since v10