Archive
How to use archive.org’s US-English news collection as a language learning corpus with QUIK-like speaking samples
- Much of TV news nowadays seems to amount to not much more than a constant stream of sound bites – however, exactly this brevity,
- the large archive and simple search interface:
- the research/browsing capabilities visible on the left here, including the varied sources – of which Arabic and French and other European TV likely provide a somewhat different perspectives on Edward Snowden –
- and the caption-like transcription, make it all the more accessible for intermediate learners of English.
- video clips of only 30 seconds length is hardly enough for instruction, however, you can have students work with corpus-QUIK-like spoken samples, and have them string a news history together if you design webquest-like research assignments – with the major added benefits, that this corpus is spoken and trains listening.
- For more background info on archive.org’s transcribed TV news, consult this NYTimes article.
Query treebanks with Fangorn for English SLA?
To provide inductive empirical examples, SLA classes have benefitted from query interfaces to target language text corpora in SLA. But corpora are usually POS-tagged – and queried – at best, which constitutes a certain “impedance mismatch” to what SLA classes actually teach. The Fangorn very large treebank query language beta demonstration page
looks already interesting for analyzing English in SLA (hover over tree elements to highlight the corresponding text), including, thanks to its capability of editing and refining queries graphically from the search results, for demonstrations during face-to-face classes. Wondering whether other corpora than Penn Treebank, Wikipedia (5k and 5000k sentences) will be made available online, and other languages but English will be supported.
A search interface to the EuroParl corpus
(Note the on my IE9, the text in the right column appears “blacked-out” – Select it to view, or use a different webbrowser).
Setting up European Union translation memories and document corpora for SDL-Trados
-
SDL-Trados installation allows the translation program to teach this industry-standard computer-aided translation application . So far, however, we had no actually translation memory loaded into this translation software.
-
The European Union is a powerhouse for translation and interpreting – at least for the wide range of their member languages many of which are world languages – , and makes some of their resources – which have been set up for translation and interpreting study use here before – available to the community free of charge as reported during a variety of LREC’s.
-
This spring, the Language Technology Group at the Joint Research Centre of the European Union this spring updated their translation memory offer DTG-TM can fill that void at least for the European Languages that have a translation component at UNC-Charlotte.
-
We download on demand (too big to store: http://langtech.jrc.ec.europa.eu/DGT-TM.html#Download)
-
Is the DGT-TM 2011 truly a superset of the 2007, or should both be merged? probably too much work?
-
-
and extract only the language pairs with English and the language only the languages “1”ed here : “G:\myfiles\doc\education\humanities\computer_linguistics\corpus\texts\multi\DGT-tm\DGT-tm_statistics.xlsx” (using “G:\myfiles\doc\education\humanities\computer_linguistics\corpus\texts\multi\DGT-tm\TMXtract.exe”)
-
and convert
-
English is the source language by default, but should be the target language in our programs,
-
The TMX format this translation memory is distributed provided in, should be “upgradeable ” to the SDL Trados Studio 2011/2011 SP1 format in the Upgrade Translation Memories wizard”.,
-
TBA:where is this component?
-
-
-
configure the Trados to load the translation memory
-
how much computing resources does this use up?
-
how do you load a tm?
-
can you load in demand instead of preload all?
-
- Here are the statistics for the translation memories for “our” languages
-
uncc Language Language code Number of units in DGT – release 2007 Number of units in DGT – release 2011 1 English EN 2187504 2286514 1 German DE 532668 1922568 1 Greek EL 371039 1901490 1 Spanish ES 509054 1907649 1 French FR 1106442 1853773 1 Italian IT 542873 1926532 1 Polish PL 1052136 1879469 1 Portuguese PT 945203 1922585 Total 8 8 7246919 15600580
-
-
Would it be of interest to have the document-focused jrc-acquis distribution of the materials underlying the translation materials available on student/teachers TRADOS computers so that sample texts can be loaded for which reliable translation suggestions will be available – this is not certain for texts from all domains – and the use of a translation memory can be trained in under realistic conditions?
-
“The DGT Translation Memory is a collection of translation units, from which the full text cannot be reproduced. The JRC-Acquis is mostly a collection of full texts with additional information on which sentences are aligned with each other.”
-
It remains to be seen how easily one can transfer documents from this distribution into Trados to work with the translation memory
-
Here is where to download:
-
uncc
lang
inc
1
de
1
en
1
es
1
fr
1
it
1
pl
1
pt
-
The JRC-Acquis comes with these statistics:
-
-
uncc
Language ISO code
Number of texts
Total No words
Total No characters
Average No words
1
de
23541
32059892
232748675
1361.87
1
en
23545
34588383
210692059
1469.03
1
es
23573
38926161
238016756
1651.3
1
fr
23627
39100499
234758290
1654.91
1
it
23472
35764670
230677013
1523.72
1
pl
23478
29713003
214464026
1265.57
1
pt
23505
37221668
227499418
1583.56
Total
7
164741
247374276
1588856237
10509.96
-
- What other multi corpora are there (for other domains and other non-European languages)?
Corpus del Español Actual (CEA)
-
Link:
- Example of KWIC view result:
- Based on Europarl, Wikicorpus (2006!), MultiUN. From their metadata page:
Metadata for Corpus del Español Actual
Corpus name
Corpus del Español Actual
CQPweb’s short handles for this corpus
cea / CEA
Total number of corpus texts
73,010
Total words in all corpus texts
539,367,886
Word types in the corpus
1,680,309
Type:token ratio
0 types per token
Text metadata and word-level annotation
The database stores the following information for each text in the corpus:
There is no text-level metadata for this corpus.
The primary classification of texts is based on:
A primary classification scheme for texts has not been set.
Words in this corpus are annotated with:
Lemma (Lemma)
Part-Of-Speech (POS)
WStart (WStart)
The primary tagging scheme is:
Part-Of-Speech
Further information about this corpus is available on the web at:
- To use, “consult the IMS’s brief description of the regular-expression syntax used by the CQP and their list of sample queries. If you wish to define your query in terms of grammatical and inflectional categories, you can use the part-of-speech tags listed on the CEA’s Corpus Tags page.”
- Also provides frequency data (based on word forms or lemmas, and others – up to a 1000):
- Examples of a frequency query result (click for full-size image. Note that a lemmatized list was requested here which links all inflected forms back to the lemma, and vice versa, upon clicking the lemma, displays a KWIC view containing all forms subsumed under that lemma, see picture above):
MS-Bing Dictionary for Chinese learners of English–and vice versa?
- Link: http://dict.bing.com.cn/?ulang=EN-US&tlang=ZH-CN#%3Ahome, powered by Engkoo:
- This looks like a pretty evolved learning tool: It has instant suggestions that include usage information and translations.
- Rich results that include contextual, parallel web-as-corpus matches in text and text-to-speech (that, on spot-checking, seems barely noticeably computer-generated).
Questions? Read the About. Or just ask me a quick
Your message has been sent
FAQs for LRC student staff or for students or for teachers. To search our FAQs, in the browser addressbar, add after "https://plagwitz.wordpress.com/feed/?tag=faqs+/" "+TAG1" (from tag cloud below) OR "https://plagwitz.wordpress.com/feed/tag=faqs
&category_name=" "CAT1" (from category hierarchy below). OR search both categories and tags, and multiple TAGs/CATs (connect with "," for OR-search, with "+" for AND-search), like so: https://plagwitz.wordpress.com/feed/?tag=TAG1+TAG2+...TAGn&category_name=CAT1
+CAT2+...CATn"
Other ways to find help
If you cannot find it here, look there: 5,500 Language-Learning Links and Programs for learning or teaching 150 languages
Shortcuts:Our Lists, Our Maps, LRC Staff Moodle Site,LRC Project Moodle Site, 49erexpress, UNCC Moodle, Student Recordings: s:claslcslrcsanakostudent
Learning usage samples: Sanako oral exam, Kaltura webcam presentation, Dictation with speech recognition, Sanako written exam, Chinese and Japanese interactive stroke-order practice
Test the Sanako Installer, Webbrowser Popup Konfigurator for XP, or Windows7, faster LRC TeacherPC Log-in Let MS facilitate diacritics writing by installing for you US-International keyboard layout
This is my personal blog (Google+). The views expressed on these pages are mine alone and not those of my employer. The information in this weblog is provided “AS IS” with no warranties, and confers no rights.
Thomas' Work Space 