Why only 1 LSTM unit in the NER system?

Hi, in week 3, we’re building an NER system using a single LSTM layer. Why are we using a single LSTM unit? If there is only a single LSTM unit, doesn’t that mean there is nothing being remembered? For a system to take into account previous inputs, should there be at least 2 LSTM units, with the first unit feeding into the second?

I’m just wondering how the NER system works if there’s really no memory.

Nevermind, the assignment uses an LSTM layer. Got hung up with the single unit in the reading/videos, which doesn’t look like that’s the actual architecture we’ll be using.

1 Like