Comtura has launched on Product Hunt check it out if you want to just dictate your notes into Salesforce, instead of fighting the Salesforce UX
https://t.co/B3ZH19KiAQ
@philoTechnica@paulg I think great programmers also need to write redundant code and make mistakes. Great programmers never stop learning. Learning goes hand in hand with mistakes and redundant code.
I shared my data annotation cookbook and tips and tricks with Alexey Georgev from DataTalks
The key takeaways:
1. Involve stakeholders in creating your data model asap
2. Rapid iteration is key
3. Active learning and distant supervision for efficiency
https://t.co/UCUPvHpl6L
Had a great talk on https://t.co/w0E6SaHc7V about dataset creation https://t.co/s2rKb8p086
I go into my conceptual framework for dataset creation and talk about:
1. in-house vs crowdsourcing
2. weak supervision to supercharge your annotation
3. tool recommendations
I will be talking on https://t.co/w0E6SaHc7V tomorrow 16:00 UTC+1 about Dataset Creation and Curation.
Tune in to hear about:
1. my process for annotation covering in-house vs crowdsourcing
2. how I use inter-annotator agreement
3. weak supervision
https://t.co/igziMwLMV6
As weak supervision is becoming more and more popular there are some interesting open source tools out there for leveraging noisy data:
https://t.co/tMsTNDJ2dH
https://t.co/EjhbQ0gRKs
Originally Snorkel was an open source project for programmatic data labelling https://t.co/aEY0uz084D
Today the team is focusing more and more on Snorkel flow, a commercial offering to reduce the cost of data labelling.
I came across Snorkel a couple of years ago it demonstrated a couple cool ideas around programmable labels:
1. thinking about programmable labels as labelling functions
2. use spacy linguistic features in these functions.
Functions help me conceptualise distant supervision
Had the pleasure to attend the Future of Data-Centric AI from Snorkel today
The core ideas of #datacentricai:
Data curation is the bottleneck, compute and models are commodities
To solve this problem re-weight and combine programmable weak supervision signals