TCSE: Ted Corpus Search Engine

Change Log

2026/11/01
Saved settings (v13.9.0): the settings panel and the list or KWIC view are kept in the browser; a label marks settings that differ from the defaults, with a button to reset them
2026/11/01
Advanced search counts sentences or segments, not matches (v13.8.0): a row with several matches counts once and shows all of them highlighted, so the totals are smaller than before and the last pages are no longer empty
2026/11/01
jQuery 3.7.1 and Bootstrap 3.4.1; fonts are served from this site instead of Google Fonts (v13.7.2)
2026/11/01
Faster plain, translation and advanced searches (v13.7.1)
2026/11/01
Faster searches (v13.7.0); with a given random seed, the rows shown differ from earlier versions and no longer shift when talks are added
2026/11/01
Robustness (v13.6.4): clearer messages for malformed requests, session cookie marked Secure and SameSite, search box accepts up to 200 characters
2026/10/31
Security hardening (v13.6.3): Content-Security-Policy added; % and _ in search keys no longer act as wildcards (* still does)
2026/10/30
Runtime updated to Ruby 3.4 (v13.6.2); tied rows in collocation tables ordered by frequency, then alphabetically
2026/08/11
Head-lemma disjunction in chunk conditions (v13.6.0): _{[a|b]}, e.g. [do|play|take]{verb} _{[part|role]}
2026/08/10
Chunk-head lemma condition (v13.5.0): _{[lemma]}, e.g. [live]{verb} _{[life]} for cognate objects
2026/08/10
Lemma echo on plain token slots (v13.4.0): =N in any {} condition, e.g. {adv:1} [and] {=1&adv}
2026/08/09
Lemma echo (v13.3.0): bind a lemma with :N and refer to it with =N, e.g. {noun:1} [after] _{=1}
2026/08/09
Entity-typed noun-chunk slots (v13.2.0): NER types in chunk conditions, e.g. by _{%PERSON}, _{%PERSON|%ORG}
2026/08/08
Typed noun-chunk slots added to Advanced Search (v13.1.0): _{COND} constrains a chunk by its head token, e.g. [give] _{pron} _
2026/08/08
Internal: data-fetch pipeline hardened (bounded redirect following, timeouts) and test suite expanded
2026/04/10
Export search results: token search results downloadable as TSV (ZIP) or JSON with context and annotations
2026/04/10
Help icons added throughout the interface linking to relevant documentation pages
2026/03/16
Construction patterns updated: adjective-subjectivity constructions added (1,160 patterns total)
2026/03/09
Performance: composite indexes, tsvector full-text search for talk lookup, zoom slider for network graph
2026/03/09
CDN dependencies removed: all JS/CSS/fonts bundled locally
2026/03/09
Text highlight default changed to off for cleaner transcript view
2026/03/09
N-gram mode renamed to Collocation: interactive network graph, collocation tables, and n-gram frequency lists
2026/03/07
Construction taxonomy revised: 6 categories (collocation, discourse, grammatical, idiom, ner_pattern, phrasal_verb)
2026/03/07
NER construction patterns added: %ENTITY slot constructions (e.g., "in %GPE", "%PERSON said")
2026/03/07
Chinese (中文) and Korean (한국어) user interface languages added
2026/03/03
KWIC (Key Word In Context) concordance view added: toggle between list and KWIC display modes
2026/03/03
POS tag shorthand aliases supported: {v}, {n}, {a}, {r}, {pr}, etc. in Advanced Search
2026/03/03
User interface modernized: responsive design, accessibility improvements, and print-friendly layout
2026/03/03
Documentation updated and moved to GitHub Pages
2026/02/25
NER (Named Entity) search added to Advanced Search: use %PERSON, %ORG, %GPE, %DATE, etc. to find named entities
2026/02/25
Collocation search with MI (Mutual Information) and t-score statistics added to Collocation mode
2026/02/25
Translation concordance: search translation text and view corresponding English segments
2026/02/25
Construction patterns categorized into phrasal_verb, idiom, grammatical, and discourse with filter UI
2026/02/25
New information structure patterns added: pseudo-cleft, resultative, and others
2026/02/25
Construction database cleaned up: duplicates and low-frequency patterns removed (1,101 patterns)
2026/02/25
SQL query optimization and database index improvements
2026/02/25
Number of searchable talks exceeds 6,400
2023/12/26
Minor modification to UI
2023/07/30
Number of searchable talks exceeds 5,000
2022/09/03
TCSE Bibliography page created
2022/09/01
Problem in pagination fixed
2022/09/01
Improved video/transcript sync
2022/09/01
User interface in Japanese added
2022/09/01
Noun chunk search function enabled
2022/09/01
Replace English POS-Tagger/Syntactic Parser with spaCy
2022/09/01
Burmese, Kurdish (Central), Northern Kurdish (Kurmanji), Slovak, and Hindi translations
2021/02/05
Video type selection functionality added
2021/02/05
Result randomization functionality added
2021/02/01
Existing data reanalyzed and many new data added
2019/03/08
Extracted keyword and discourse marker highlighting implemented (experimental)
2019/02/18
Documentation released
2019/02/12
Related talk recommendation based on cosine similarity was implemented
2019/02/06
TF-IDF keywords have become searchable in "Search Talk Info"
2019/02/04
Keywords extracted using TF-IDF are viewable by clicking on icon
2019/02/04
Playback speed control and video quality control are hidden unless 'TED' is the selected video source
2018/12/01
Swedish language translation added
2018/06/07
A big increase of the number of talks with translation available
2018/03/11
Usages of syntax updated: 'negative match' and 'one word wild card'
2017/09/21
Improved video/transcript sync
2017/09/20
Changed from auto-play to click-to-play / Video control simplified / Query timeout enabled
2017/09/18
Better wpm calculation
2017/09/08
Window styles largely modified
2017/09/05
N-gram presentation algorithm changed; Lots of internal fixes
2017/08/27
Startup time and language setting detection improved
2017/05/12
Video and transcript sync problem largely resolved
2017/05/06
Integrated help has been implemented
2017/05/04
Talk speed and readability/difficulty are shown in talk info and list
2016/11/28
Improved video sync on both TED and YouTube video playing modes
2016/11/23
Improved user experience in full screen video playing mode
2016/11/19
"I'm feeling lucky" button added to play a talk randomly selected
2016/11/10
TED Video quality is made selectable from normal (180k) or high (450k)
2016/11/02
More responsive design
2016/11/01
Shows speaker info when available
2016/10/30
Segment repeat-play button added
2016/10/29
Right-to-left languages (i.e. Arabic, Hebrew, and Persian) are supported (partial and experimental)
2016/10/29
More translation languages added
2016/10/26
Shows list of all available talks
2016/10/22
Video transcript view control added
2016/09/04
Video source gets selectable (again) between TED and YouTube
2016/05/31
Translation language is semi-automatically set using language code in request header
2016/05/30
Various construction-related functions added (yet to be documented)
2016/04/09
Video player positioning fixed and user experience on smartphones much improved
2016/04/07
Video quality is now selectable from normal (180k) or low (64k)
2015/10/08
Many more translations available: those whose line counts differ from the English transcript are no longer excluded
2015/05/06
Fallback search of videos implemented
2015/05/05
Version 3.0 released: transcript shown in sync with the video (fine-tunable); talks in languages other than English removed
2015/04/25
Controllable play speed / Automatic video stop
2015/04/24
Small improvements on advanced search / Problems playing video due to YouTube API change fixed
2015/02/25
Pop-up on the English transcript shows POS analysis and all available translations; stability improved
2015/02/21
Version 2.0 released: translations in 15 languages searchable; advanced search in Japanese removed; many bug fixes
2014/11/28
Video now starts where the segment starts; “play video 10 seconds earlier” option added
2014/11/14
Video search accuracy improved
2014/11/13
Regular search now returns items containing matches beginning from the onset of a word, not in the middle of a word
2014/11/01
Initial release