Update README.md
Browse files
README.md
CHANGED
@@ -12,22 +12,17 @@ license: apache-2.0
|
|
12 |
|
13 |
## luke-japanese
|
14 |
|
15 |
-
**luke-japanese** is the Japanese version of **LUKE** (**L**anguage
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
**luke-japanese**は、単語とエンティティの知識拡張型訓練済みモデル**LUKE**の日本
|
24 |
-
語版です。LUKE は単語とエンティティを独立したトークンとして扱い、これらの文脈を
|
25 |
-
考慮した表現を出力します。詳細については
|
26 |
-
、[GitHub リポジトリ](https://github.com/studio-ousia/luke)を参照してください。
|
27 |
|
28 |
### Experimental results on JGLUE
|
29 |
|
30 |
-
The
|
31 |
[JGLUE](https://github.com/yahoojapan/JGLUE) is shown as follows:
|
32 |
|
33 |
| Model | MARC-ja | JSTS | JNLI | JCommonsenseQA |
|
@@ -40,6 +35,8 @@ The performance of luke-japanese evaluated on the dev set of
|
|
40 |
| Waseda RoBERTa base | 0.962 | 0.901/0.865 | 0.895 | 0.840 |
|
41 |
| XLM RoBERTa base | 0.961 | 0.870/0.825 | 0.893 | 0.687 |
|
42 |
|
|
|
|
|
43 |
### Citation
|
44 |
|
45 |
```latex
|
|
|
12 |
|
13 |
## luke-japanese
|
14 |
|
15 |
+
**luke-japanese** is the Japanese version of **LUKE** (**L**anguage **U**nderstanding with **K**nowledge-based **E**mbeddings), a pre-trained _knowledge-enhanced_ contextualized representation of words and entities. LUKE treats words and entities in a given text as independent tokens, and outputs contextualized representations of them. Please refer to our [GitHub repository](https://github.com/studio-ousia/luke) for more details and updates.
|
16 |
+
|
17 |
+
This model contains Wikipedia entity embeddings which are not used in general NLP tasks. Please use the [lite version](https://huggingface.co/studio-ousia/luke-japanese-base-lite/) for tasks that do not use Wikipedia entities as inputs.
|
18 |
+
|
19 |
+
**luke-japanese**は、単語とエンティティの知識拡張型訓練済みTransformerモデル**LUKE**の日本語版です。LUKEは単語とエンティティを独立したトークンとして扱い、これらの文脈を考慮した表現を出力します。詳細については、[GitHub リポジトリ](https://github.com/studio-ousia/luke)を参照してください。
|
20 |
+
|
21 |
+
このモデルは、通常のNLPタスクでは使われないWikipediaエンティティのエンベディングを含んでいます。単語の入力のみを使うタスクには、[lite version](https://huggingface.co/studio-ousia/luke-japanese-base-lite/)を使用してください。
|
|
|
|
|
|
|
|
|
|
|
22 |
|
23 |
### Experimental results on JGLUE
|
24 |
|
25 |
+
The experimental results evaluated on the dev set of
|
26 |
[JGLUE](https://github.com/yahoojapan/JGLUE) is shown as follows:
|
27 |
|
28 |
| Model | MARC-ja | JSTS | JNLI | JCommonsenseQA |
|
|
|
35 |
| Waseda RoBERTa base | 0.962 | 0.901/0.865 | 0.895 | 0.840 |
|
36 |
| XLM RoBERTa base | 0.961 | 0.870/0.825 | 0.893 | 0.687 |
|
37 |
|
38 |
+
The baseline scores are obtained from [here](https://github.com/yahoojapan/JGLUE/tree/9f650417195ec54a080411f44e2395012979d42e#baseline-scores).
|
39 |
+
|
40 |
### Citation
|
41 |
|
42 |
```latex
|