Word embedding - Wikipedia
9
公开标注数
6
参与人数
2026-07-18 09:16:47
首次 Whisper
本页的公开 Whisper
划选高亮2026-07-18 15:40:47
原文高亮摘录
“Word embeddings may contain the biases and stereotypes contained in the trained dataset”
Whisper 随想笔记
Garbage in, garbage out—models just learn what we feed 'em.
划选高亮2026-07-18 12:37:47
原文高亮摘录
“words with multiple meanings are conflated into a single representation”
Whisper 随想笔记
But don't we just average senses anyway? Works fine for most tasks honestly.
划选高亮2026-07-18 12:28:47
原文高亮摘录
“words with multiple meanings are conflated into a single representation”
Whisper 随想笔记
Yeah that's why transformers killed static embeddings, context is everything.
划选高亮2026-07-18 10:21:11
原文高亮摘录
“a word embedding is a representation of a word”
Whisper 随想笔记
I always thought embeddings were just for search, but this sounds way broader.
划选高亮2026-07-18 10:12:11
原文高亮摘录
“a word embedding is a representation of a word”
Whisper 随想笔记
If similar words are close in space, does that mean 'hot' and 'cold' are neighbors?
划选高亮2026-07-18 10:03:11
原文高亮摘录
“a word embedding is a representation of a word”
Whisper 随想笔记
So it's basically a dictionary but for computers, got it.
划选高亮2026-07-18 09:34:47
原文高亮摘录
“a real-valued vector that encodes the meaning of the word”
Whisper 随想笔记
Makes sense, like when I tried to build a chatbot and it kept mixing up synonyms.
划选高亮2026-07-18 09:25:47
原文高亮摘录
“a real-valued vector that encodes the meaning of the word”
Whisper 随想笔记
But how do you actually measure 'closer in space'? Seems kinda fuzzy to me.
划选高亮2026-07-18 09:16:47
原文高亮摘录
“a real-valued vector that encodes the meaning of the word”
Whisper 随想笔记
So basically it's just a fancy way to say words become numbers that mean stuff.
分享本页 Whisper
短链接
https://domwhisper.com/s/645c023a67b4嵌入代码
<iframe src="https://domwhisper.com/embed/645c023a67b4" width="100%" height="480" style="border:0;border-radius:16px" loading="lazy"></iframe>