The MERLIN corpus contains 2,286 texts for learners of Italian, German and Czech that were taken from written examinations of acknowledged test institutions. The exams aim to test knowledge across the levels A1-C1 of the Common European Framework of Reference (CEFR).
The texts are writing samples from TELC - language tests (German and Italian) and the exams of the University of Prague (Czech). The tasks are standardized and aligned to the Common European Framework of Reference for Languages (CEFR). ...
Both institutions, TELC as provider of the European Language Certificates and the Test Centre of the Institute of Language and Preparatory Studies at Charles University in Prague, are full members of ALTE. They offer internationally recognized language examinations in line with the high ALTE standards.
more details on data preparation for annotations
To ensure an immediate relation to the CEFR, specially trained testers re-rated all exam texts using the MERLIN rating grid that was developed within the project. ...
The reliability of the ratings was subjected to rigorous statistical verification procedures. As a result, a reliable rating profile is created for each text in the corpus. The profile reflects both a general holistic overall level and the individual rating criteria detailed below:
general linguistic range | vocabulary range | vocabulary control | grammatical accuracy | coherence | sociolinguistic appropriateness | orthography
We provide a comprehensive overview of the test tasks by target language and CEFR level tested. ...
The level of the test may differ from the level that the learner text received in the re-ratings.
The tasks are represented using a grid that was developed for these purposes by ALTE (Association of Language Testers in Europe, www.alte.org). The grid contains detailed information about the tasks and the specific characteristics of the intended text, e.g. regarding topic, register, domain (author: Olaf Bärenfänger).
General notes on task descriptions 
Please cite the task descriptions as: MERLIN project, task description: <name of the task>, 2014, http://merlin-platform.eu
Each text in the corpus is described with the following metadata:
Information about the text author:
age, gender, mother tongue (L1)
Information about the test:
task ID and topic, CEFR level of the test the written production was extracted from
Ratings:
Overall rating: CEFR level the test received in the re-rating
Fair CEFR level according to single rating criteria: general linguistic range | vocabulary range | vocabulary control | grammatical accuracy | coherence | sociolinguistic appropriateness | orthogaphy
You can define a subset of learner texts using these metadata in » Define a subcorpus. In this way you can analyse phenomena of the learner language for certain groups of learners.
Metadata can be displayed for each text:
» Simple search: click on "view learner info and ratings" in your search result
» Advanced search: click on the [i] in your search result
» Define a subcorpus: click on "view learner info and ratings" in your search result
Number of texts per CEFR level of the test (test level) compared to the number of texts per CEFR level assigned in the re-rating (fair average)
Test Level |
Fair Average |
|||
Czech |
|
|
A1 |
1 |
A2 |
111 |
A2 |
189 |
|
B1 |
143 |
B1 |
165 |
|
B2 |
188 |
B2 |
81 |
|
|
|
C1 |
2 |
|
Italian |
A1 |
207 |
A1 |
29 |
A2 |
202 |
A2 |
378 |
|
B1 |
201 |
B1 |
394 |
|
B2 |
201 |
B2 |
2 |
|
German |
A1 |
206 |
A1 |
57 |
A2 |
209 |
A2 |
297 |
|
B1 |
210 |
B1 |
331 |
|
B2 |
204 |
B2 |
293 |
|
C1 |
204 |
C1 |
42 |
|
|
|
C2 |
4 |
|
Total |
2286 |
2265 |
||
Number of annotated texts per annotation layer
|
Czech |
German |
Italian |
Total number of texts |
442 |
1033 |
813 |
Texts that received TH1 |
440 |
1033 |
813 |
Texts that received EA1 |
361 |
752 |
754 |
Texts that received TH2 |
231 |
275 |
154 |
Texts that received EA2 |
198 |
258 |
85 |
