# Which token vector is used for Sentiment Analysis? **URL:** \ **Category:** Beginners\ **Created:** [March 4, 2022, 7:57pm UTC](https://discuss.huggingface.co/t/which-token-vector-is-used-for-sentiment-analysis/15372 "2022-03-04T19:57:13Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![MahdiA](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/mahdia/32/2700_2.png) [@MahdiA](https://discuss.huggingface.co/u/MahdiA)\ **Post date:** [March 4, 2022, 7:57pm UTC](https://discuss.huggingface.co/t/which-token-vector-is-used-for-sentiment-analysis/15372/1 "2022-03-04T19:57:13Z")
I see in most of Sentiment Analysis tasks which are implemented based on BERT, only the embedding of `[CLS]` is passed to classifier, while others are useless. What is the reason behind it? ---
**Author:** ![beneyal](https://sea2.discourse-cdn.com/hellohellohello/user_avatar/discuss.huggingface.co/beneyal/32/27988_2.png) [@beneyal](https://discuss.huggingface.co/u/beneyal)\ **Post date:** [March 4, 2022, 8:14pm UTC](https://discuss.huggingface.co/t/which-token-vector-is-used-for-sentiment-analysis/15372/2 "2022-03-04T20:14:52Z")
According to the paper, BERT’s `[CLS]` token aggregates the hidden states of the other tokens, which renders them “useless” for sequence classification tasks, as all relevant info is already pooled into `[CLS]`. ---
**Author:** ![DeepthiSankepalli](https://avatars.discourse-cdn.com/v4/letter/d/d78d45/32.png) [@DeepthiSankepalli](https://discuss.huggingface.co/u/DeepthiSankepalli)\ **Post date:** [February 16, 2024, 12:02pm UTC](https://discuss.huggingface.co/t/which-token-vector-is-used-for-sentiment-analysis/15372/3 "2024-02-16T12:02:47Z")
Can you please elaborate more on it? I see the transformers/BERT layers with Token Embeddings, Segment Embeddings, Position Embeddings available along with CLS, SEP