Notice
Recent Posts
Recent Comments
Link
| 일 | 월 | 화 | 수 | 목 | 금 | 토 |
|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | |
| 7 | 8 | 9 | 10 | 11 | 12 | 13 |
| 14 | 15 | 16 | 17 | 18 | 19 | 20 |
| 21 | 22 | 23 | 24 | 25 | 26 | 27 |
| 28 | 29 | 30 | 31 |
Tags
- 기계학습
- gradient descent
- BFS
- 파이썬
- Linear Model
- LeetCode
- machinelearning
- GPT
- transformer
- 코테
- 프롬프트
- Deeplearning
- attention
- Linear Regression
- LLM
- 머신러닝
- rnn
- Python
- prompt engineering
- 프로그래머스
- 부스트캠프
- 알고리즘
- 일기
- dl
- Django
- deque
- NLP
- 코딩테스트
- Programmers
- ChatGPT
Archives
- Today
- Total
크크루쿠쿠
[NLP]Word Embedding 본문
Word Embedding
- word를 vector로 변환하는것.
- 'cat' 과 'kitty' 는 비슷한 단어 -> short distance
- 'hamburger' 와 'cat'은 유사X -> far distance
Word2Vec
- 같은문장 내에서 인접한 단어는 유사도가 높을것이다
ex) 'The cat purrs.','This cat hunts mice'
- "cat" 단어 주변 단어의 확률 분포를 갖게됨

How Word2Vec Algorithm Works
Sentence: "I study math.", Vocabulary: {"I","study","math"}, Input: "study" [0,1,0], Output: "math" [0,0,1]

word vector를 공간에 표현하면 word간의 관계를 볼 수 있음.
GloVe: Global Vectors for Word Representation
- 입출력 단어 쌍에 대해서 사전에 Window 내에서 얼마나 나타나는지 계산함
- 그 값에 가까워지도록 새로운 loss function을 사용함

- 빈번한 단어 쌍 자체를 loss 로 넣어 중복되는 계산 제거 -> Word2Vec과의 차이점
- Fast training
-

성별의 차이가 비슷한 vector로 나타남.
'AI > 부스트캠프 AI Tech' 카테고리의 다른 글
| [NLP] LSTM and GRU (0) | 2021.09.07 |
|---|---|
| [NLP]Recurrent Neural Network and Language Modeling (0) | 2021.09.07 |
| [NLP]Intro to NLP, Bag-of-Words (0) | 2021.09.06 |
| Pytorch Troubleshooting (0) | 2021.08.20 |
| Pytorch Hyperparameter Tuning (0) | 2021.08.20 |
Comments