Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

BLSTM + BioWordVec #107

Open
michelole opened this issue Apr 28, 2019 · 2 comments
Open

BLSTM + BioWordVec #107

michelole opened this issue Apr 28, 2019 · 2 comments

Comments

@michelole
Copy link
Member

michelole commented Apr 28, 2019 •
edited
Loading

We need a common ground to compare LSTM with BioWordVec embeddings and Arnold's strategy with character trigrams. Either:

I suggest the second.
@michelole michelole added the P0 Highest priority issues, a MUST label Apr 28, 2019
@michelole
Copy link
Member Author

michelole commented May 31, 2019 •
edited
Loading

A quick-and-dirty approach for the second was tried before on https://github.com/michelole/n2c2/tree/issue-107, but led to Invalid shape: Requested INDArray shape [1, 0, 1750] contains dimension size values < 1 (all dimensions must be 1 or more).

Maybe we should try now the first (#110) and completely drop Bi-LTSM approach, because it is too complex.

@michelole michelole mentioned this issue May 31, 2019
Open
michelole added a commit to michelole/n2c2 that referenced this issue Jun 4, 2019
@michelole
Refactor code around new interface InputRepresentation
d524666 
Introduce the new interface `InputRepresentation` to separate logic of input representation (e.g. word embeddings, character trigrams) from iterators and classifiers. This allows new combinations required as part of bst-mug#107 and bst-mug#110.

Move data-dependent methods such as `initializeTruncateLength` and `loadFeaturesForNarrative` to the iterators.

Remove public and duplicate attributes to reduce complexity.
michelole added a commit to michelole/n2c2 that referenced this issue Jun 4, 2019
@michelole
Extract architecture code into own interface
2bc8a92 
This allows other combinations as required by bst-mug#107 and bst-mug#110.
@michelole michelole mentioned this issue Jun 4, 2019
Merged
@michelole
Copy link
Member Author

michelole commented Jun 5, 2019 •
edited
Loading

We decided to drop BLSTM method, so removing P0.

@michelole michelole removed the P0 Highest priority issues, a MUST label Jun 5, 2019
@michelole michelole changed the title Common deep learning method BLSTM + BioWordVec Jun 5, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@michelole