stefan-it commited on
Commit
5c9b42a
·
verified ·
1 Parent(s): c47824b

docs: add initial version

Browse files
Files changed (1) hide show
  1. README.md +45 -3
README.md CHANGED
@@ -1,3 +1,45 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ ---
4
+
5
+ # TurBLiMP Evaluations
6
+
7
+ This dataset hosts the TurBLiMP evaluation results on my [Turkish Model Zoo](https://huggingface.co/collections/stefan-it/turkish-language-models-653c3146fbe250285fccec0f).
8
+
9
+ More about the TurBLiMP benchmark:
10
+
11
+ > TurBLiMP is the first Turkish benchmark of linguistic minimal pairs, designed to evaluate the linguistic abilities of monolingual and multilingual language models (LMs).
12
+ > This benchmark covers 16 core grammatical phenomena in Turkish, with 1,000 minimal pairs per phenomenon.
13
+ > Additionally, it incorporates experimental paradigms that examine model performance across different subordination strategies and word order variations.
14
+
15
+ I've modified the [original](https://github.com/ezgibasar/TurBLiMP/blob/297de13fb7a0ce524fe32e8b175c6b5255d66960/evaluation.py) evaluation script and extended it for the Turkish Model Zoo. My evaluation script can be found [here](own_evaluation.py).
16
+
17
+ # Results
18
+
19
+ After running the evaluation script, all results can be parsed with this [notebook](ParseEvaluations.ipynb) to print-out a nice overview table:
20
+
21
+ | Phenomenon | [`dbmdz/electra-small-turkish-cased-generator`](https://huggingface.co/dbmdz/electra-small-turkish-cased-generator) | [`dbmdz/electra-base-turkish-cased-generator`](https://huggingface.co/dbmdz/electra-base-turkish-cased-generator) | [`dbmdz/electra-base-turkish-mc4-cased-generator`](https://huggingface.co/dbmdz/electra-base-turkish-mc4-cased-generator) | [`dbmdz/electra-base-turkish-mc4-uncased-generator`](https://huggingface.co/dbmdz/electra-base-turkish-mc4-uncased-generator) | [`dbmdz/bert-base-turkish-cased`](https://huggingface.co/dbmdz/bert-base-turkish-cased) | [`dbmdz/bert-base-turkish-uncased`](https://huggingface.co/dbmdz/bert-base-turkish-uncased) | [`dbmdz/bert-base-turkish-128k-cased`](https://huggingface.co/dbmdz/bert-base-turkish-128k-cased) | [`dbmdz/bert-base-turkish-128k-uncased`](https://huggingface.co/dbmdz/bert-base-turkish-128k-uncased) | [`dbmdz/distilbert-base-turkish-cased`](https://huggingface.co/dbmdz/distilbert-base-turkish-cased) | [`dbmdz/convbert-base-turkish-cased`](https://huggingface.co/dbmdz/convbert-base-turkish-cased) | [`dbmdz/convbert-base-turkish-mc4-cased`](https://huggingface.co/dbmdz/convbert-base-turkish-mc4-cased) | [`dbmdz/convbert-base-turkish-mc4-uncased`](https://huggingface.co/dbmdz/convbert-base-turkish-mc4-uncased) |
22
+ |----------------------|-----------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------|
23
+ | Anaphor Agreement | 74.1 | 94.3 | 94.3 | 92.8 | 96.7 | 97.3 | 97.3 | 97.7 | 96.9 | 58.1 | 44.3 | 44.6 |
24
+ | Argument Str. Tran. | 86.6 | 99.6 | 99.4 | 98.7 | 99.7 | 99.6 | 99.8 | 99.1 | 97.5 | 51.9 | 58.1 | 51.3 |
25
+ | Argument Str. Ditr. | 79.3 | 96.1 | 95.5 | 95.2 | 99.8 | 96.1 | 96.1 | 96.1 | 95.4 | 64.6 | 58.6 | 64.5 |
26
+ | Binding | 70.7 | 96.2 | 91.4 | 89.6 | 99.9 | 98.5 | 97.7 | 99 | 93 | 89.1 | 49.4 | 78.4 |
27
+ | Determiners | 91.8 | 99.3 | 98.2 | 99.1 | 99.9 | 100 | 99 | 99.3 | 82.9 | 0 | 0 | 0 |
28
+ | Ellipsis | 10.6 | 49.7 | 46.3 | 49 | 87.4 | 73.6 | 96.6 | 87.5 | 13.6 | 54.7 | 57.8 | 67.9 |
29
+ | Irregular Forms | 98.7 | 97.9 | 99 | 99.8 | 98.8 | 100 | 99.9 | 99.6 | 94.1 | 82.9 | 86.6 | 95.2 |
30
+ | Island Effects | 39.1 | 35.3 | 41.8 | 44 | 49.4 | 39.8 | 60.9 | 51.2 | 47.4 | 96.7 | 99.4 | 100 |
31
+ | Nominalization | 90 | 96.6 | 97 | 95.4 | 97.4 | 97 | 98.9 | 97.4 | 95.6 | 55.2 | 59.2 | 60.6 |
32
+ | NPI Licensing | 90.9 | 96.1 | 95 | 98 | 98.2 | 97.6 | 97.2 | 95 | 92.1 | 82.1 | 95.6 | 71.9 |
33
+ | Passives | 100 | 91.2 | 93.6 | 91.6 | 82.2 | 78.1 | 84.4 | 81.3 | 98.8 | 100 | 100 | 99 |
34
+ | Quantifiers | 97.9 | 98 | 98 | 97.6 | 95.7 | 94.6 | 98 | 98.4 | 98.4 | 99 | 99 | 99 |
35
+ | Relative Clauses | 79.9 | 90.7 | 92 | 91.6 | 97.7 | 97.5 | 97 | 98.5 | 92 | 53.4 | 53.7 | 56.9 |
36
+ | Scrambling | 99.5 | 100 | 100 | 99.8 | 100 | 100 | 99.6 | 100 | 99.8 | 38.7 | 59.3 | 63.3 |
37
+ | Subject Agreement | 82.8 | 99 | 97.2 | 96.1 | 98.3 | 99.2 | 99.1 | 98.8 | 97 | 47.7 | 43.9 | 56.4 |
38
+ | Suspended Affixation | 97.5 | 99 | 99.1 | 98.8 | 100 | 100 | 100 | 100 | 100 | 25.4 | 12.8 | 23.2 |
39
+ | Model Average | 80.6 | 89.9 | 89.9 | 89.8 | 93.8 | 91.8 | **95.1** | 93.7 | 87.2 | 62.5 | 61.1 | 64.5 |
40
+
41
+ # Summary
42
+
43
+ The [TurBLiMP paper](https://arxiv.org/abs/2506.13487) used the [`dbmdz/bert-base-turkish-128k-uncased`](https://huggingface.co/dbmdz/bert-base-turkish-128k-uncased) model showing strong performance.
44
+
45
+ My evaluation here showed, that the [`dbmdz/bert-base-turkish-128k-cased`](https://huggingface.co/dbmdz/bert-base-turkish-128k-cased) even performs better on the TurBLiMP benchmark.