menu

snorkel

Building and managing training datasets for machine learning

Channels
# All channels
view-forward
# announcements
view-forward
# api
view-forward
# applications
view-forward
# help
view-forward
# projects
view-forward
# tutorials
view-forward
Team
Posts
Members
Info
down-caret

Announcing Snorkel Flow!

We made an exciting announcement on snorkel.org today! We've reposted that message below: The Snorkel team is now focusing their efforts on Snorkel Flow, an end-to-end AI application development platform based on the core ideas behind Snorkel—check it out at snorkel.ai! The…

thumbsup
5
message-simple
6

How to use Snorkel on numeric data?

Hi - I'm trying to see if I can use Snorkel to label my numeric training dataset (and then looking to either do label classification or convert the labels to scores and do a regression classification). My question is if there are coding examples anywhere that show how to use…

thumbsup
0
message-simple
0

Widely varying results when training multi-task model on same inputs

Hello! I'm in the process of writing a paper comparing the efficacy of multi-task models with weak supervision to the performance of their single-task equivalents, and I've been using Snorkel v0.9.6 to explore this. However, I've been experiencing a strange issue in that when…

thumbsup
0
message-simple
2

Where to find this util package?

I am trying to follow this https://www.snorkel.org/get-started/. I am not able to locate from utils import load_unlabeled_spam_dataset. Can someone guide me on where to get this package? Appreciate your help and time.

thumbsup
0
message-simple
2

Why the probabilities are not higher in the positive label group?

Hi, I used LabelModel to label my train set, and I also obtained probability by using label_model.predict(L=L_train, tie_break_policy="abstain", return_probs=True). I assume that the probabilities should be relatively higher in the positive label group. However, that is not what…

thumbsup
0
message-simple
4

Is it possible to assign weights on label functions

Hi, dear Snorkel community, I am relatively new to using Snorkel. My purpose is to use Snorkel to generate labels for a dataset we got. So far, we came up with dozens of rules, which are mainly based on domain knowledge of the data. Domain knowledge gave us hints that these…

thumbsup
0
message-simple
3

Snorkel for Time Series Annotation

Hi everyone :) I was diving into snorkel methods, namely labeling functions and slicing functions (and using LabelModel) for labeling time series of IMU data (muck like the paper that got posted here, https://openreview.net/pdf?id=SJedYj5ruV), and I managed to get some labeling…

thumbsup
0
message-simple
1

What is the best redundant labeling strategy to maximize the performance of…

I have 50,000 sentences I need labeled for a custom NER model, with triple redudancy. I can have as many workers as I want do the labeling. I will combine their labels using Snorkel's LabelModel. What is the best strategy to employ to give the LabelModel the best performance…

thumbsup
2
message-simple
2

Training with ground-truth labeled and Snorkel-labeled data

Background: I'm writing a conference paper on some Snorkel work, so I want to make sure I understand why Snorkel is considered a weak supervision approach, but isn't considered a semi-supervised learning approach, even though many tutorials leverage some ground truth data in…

thumbsup
1
message-simple
2

How to use the mdoel probabilities output

I'm trying to use Snorkel for labeling social media text as topics. In the tutorial it says "As we saw in Section 4, the LabelModel outputs probabilistic (float) labels. If the classifier we are training accepts target labels as floats, we can train on these labels directly" …

thumbsup
0
message-simple
1

Shortlisting large number of labelling functions

I am trying to use snorkel in the context of credit risk modelling (binary classification). I have written lots of labeling functions. I do have some ground truth available so I want to inform labeling functions somehow. The question is: which ground-truth related metrics is it…

thumbsup
0
message-simple
1