Comments (1)
Hi Ashutosh! This is a good question and I'm not sure if what we're doing is the best strategy. For urls we are using twitterurl
and for user mentions twitteruser
. These get tokenized like so:
# twitterurl
['twitter', '##ur', '##l']
# twitteruser
['twitter', '##user']
The rationale here was to minimize the number of tokens needed to represent the fillers but at the same time avoid conflicts with other tokens. Also, we wanted to re-use the default vocab for BERT large in this case and not create a new one.
In general, I would argue that BERT will likely be able to learn the embedding for any arbitrary filler, given enough data.
FYI, the code to prepare the pretraining data can be found here.
Hope this helps!
from covid-twitter-bert.
Related Issues (18)
- Add link to preprint in readme HOT 1
- Failed to convert the TensorFlow model to a PyTorch model HOT 4
- Masked worked prediction returning "unused" tokens HOT 3
- CUDA out of memory Error from loading CT-Bert directly from Huggingface HOT 2
- License? HOT 2
- Failed to download dataset HOT 1
- Strategies for downstream NER tasks with more label
- Make one over CORD-19 HOT 1
- the model architecture of COVID-Twitter-BERT v2 MNLI
- No proper encodings for covid-related terms HOT 1
- Failed to find data adapter HOT 2
- A few observations. HOT 2
- TypeError HOT 1
- Logic behind [UNK] token for pronouns HOT 2
- COVID Category (CC) dataset has 4 invalid formats for Tweet ID HOT 1
- Loss HOT 2
- Download Twitter-Sentiment SemEval HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from covid-twitter-bert.