lxyuan

/

distilbert-base-multilingual-cased-sentiments-student

like 258

Hi,

For this fine-tuned sentiment model, the underlying architecture is distilbert-base-multilingual-cased, which supports 104 languages, including Korean. In other words, this model has the capability to process Korean text to some extent. Therefore, I suggest you first evaluate this model on your Korean sentiment dataset to gauge its performance.

If you wish to further improve the model's performance on the Korean language, you can fine-tune it using Korean-specific sentiment analysis datasets. For a practical example, please refer to the notebook link attached to my model card.

Quick test on korean:

저는이 영화가 좋아요
>>> [
  [
    {
      "label": "positive",
      "score": 0.9288509488105774
    },
    {
      "label": "neutral",
      "score": 0.04762842506170273
    },
    {
      "label": "negative",
      "score": 0.02352064475417137
    }
  ]
]


나는 이 영화를 싫어한다.
>>> [
  [
    {
      "label": "negative",
      "score": 0.5530788898468018
    },
    {
      "label": "positive",
      "score": 0.2614102363586426
    },
    {
      "label": "neutral",
      "score": 0.18551082909107208
    }
  ]
]