nasa-petal PeTaL-labeller issues

nasa-petal / PeTaL-labeller

The PeTaL labeler labels journal articles with biomimicry functions.

https://petal-labeller.readthedocs.io/en/latest/

The Unlicense

6 stars 3 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Add files via upload

#89 ARalevski closed 2 years ago
0
Preprocess_golden.py

#88 pjuangph closed 1 year ago
0
Update README.md

#87 pjuangph opened 2 years ago
0
Match vocab file PeTaL.emb

#86 pjuangph opened 2 years ago
1
taxonomy update

#85 dsmith111 closed 2 years ago
0
Explore tokenizer optimization strategies to improve precision/recall.

#84 bruffridge opened 2 years ago
0
Change output of labeller from absolute "selection" to "ranking" using confidence scores.

#83 bruffridge closed 2 years ago
1
Match with petal

#82 elkong closed 2 years ago
0
Try using a tree of multilabel classifiers

#81 bruffridge opened 2 years ago
1
Look into Label Studio to help increase the size of our labelled dataset for training

#80 bruffridge opened 2 years ago
0
Golden Gate

#79 elkong closed 2 years ago
0
Try Google Cloud AutoML Natural Language's multilabel text classification on golden dataset

#78 bruffridge opened 2 years ago
1
Add a description of metrics to MATCH's README

#77 bruffridge closed 2 years ago
1
try to use a language model like GPT-2 for its general-purpose language understanding capabilities (and then integrate it with the MATCH classification task somehow)

#76 bruffridge opened 2 years ago
0
See if using an ensemble method with MATCH improves performance.

#75 bruffridge opened 2 years ago
0
Use a large ensemble of single-label classifiers (so, treat all of the labels independently, ignore the hierarchy, and we have a hundred separate yes/no tasks) and see if this works better than MATCH

#74 bruffridge opened 2 years ago
0
Run Match on just level 1 labels

#73 bruffridge closed 2 years ago
4
See how replacing random weights with pretrained and fine-tuned weights in MATCH affects performance

#72 bruffridge opened 2 years ago
4
auto-labeler/MATCH ought to look better now

#71 elkong closed 3 years ago
0
Create a multi-label classification model for the most used labels.

#70 bruffridge closed 2 years ago
0
Create a binary classification filtering model for the top 25% of labels

#69 bruffridge opened 3 years ago
0
rerun ablation study

#68 bruffridge closed 3 years ago
1
match, migration, mag, mesh

#67 elkong closed 3 years ago
7
weights, biases, and ablations

#66 elkong closed 3 years ago
0
Use open-source Snorkel to create labelling functions to expand our training dataset.

#65 bruffridge opened 3 years ago
2
MATCH with PeTaL: relevance thresholds, MAG and MeSH terms, flat taxonomy

#64 elkong closed 3 years ago
0
52 analyze data

#63 dsmith111 closed 3 years ago
0
Add 'leaf only' option to lensCleaner

#62 dsmith111 closed 3 years ago
0
Produce metrics to show which labels are being classified correctly and which aren't, and how they're being misclassified.

#61 bruffridge closed 2 years ago
5
Compare performance of only including leaf labels in the dataset.

#60 bruffridge closed 3 years ago
1
Replace MAG topics with another topic taxonomy

#59 bruffridge closed 3 years ago
1
Do k-fold cross validation to generate ablation study results for including MAG and MeSH labels.

#58 bruffridge closed 3 years ago
1
cleaned up and added work and analysis from Colab and GRC HPC

#57 elkong closed 3 years ago
0
Compare MATCH results to auto-sklearn

#56 bruffridge opened 3 years ago
1
Look into using a relevancy threshold vs. top k for labelling

#55 bruffridge closed 3 years ago
1
More DOIs

#54 dsmith111 closed 3 years ago
0
Does adding MeSH terms and/or MAG fields of study improve accuracy?

#53 bruffridge closed 3 years ago
3
Analyze our training dataset to discover ways to improve it to improve labelling accuracy

#52 bruffridge closed 3 years ago
1
Add DOI search

#51 dsmith111 closed 3 years ago
3
Only 409 papers with labels in cleaned_lens_output.json, should be 701

#50 bruffridge closed 3 years ago
1
Taxonomy format

#49 dsmith111 closed 3 years ago
0
Verify input format expected by MATCH for unlabelled papers

#48 bruffridge closed 3 years ago
3
Create a label hierarchy for MATCH input

#47 bruffridge closed 3 years ago
0
Python MATCH data preparation from Lens

#46 dsmith111 closed 3 years ago
2
Build a data pipeline for running unlabelled papers through the labeller

#45 bruffridge opened 3 years ago
5
Create a POC for integrating a Colab classification model with weights and biases

#44 bruffridge closed 3 years ago
0
Prepare train/test data for MATCH

#43 bruffridge closed 3 years ago
0
Look into using MATCH to improve the labeler

#42 bruffridge closed 3 years ago
6
Look into how we might use SPECTER to improve our labeller

#41 bruffridge opened 3 years ago
4
Plot a graph that shows how much adding additional training data improves labeller accuracy.

#40 bruffridge closed 3 years ago
1