LEWIP-informal model is designed to transfer formal text into informal keeping the important slots from the source text. The slots can be either pre-defined or detected automatically.
The model is based on LEWIP (Levenshtein editing with the parallel corpus). It exploits the ability of T5 model to fill the slots between the known texts with undefined number of tokens.
How to use
Installation
!pip install lewip_informal
Generating content preserving informal paraphrases
When the important entities are known in advance
from lewip_informal import LEWIP
model = LEWIP(use_cuda = True)
text = 'I want to go to NY'
ent = ['NY']
model.generate(text, ent)
# expected output 'i wanna go to NY'
When the important entities are NOT known in advance
In case the important slots are not known, they are automatically detected with auxiliary tagger model.
from lewip_informal import LEWIP
model = LEWIP(predefined_entities = False, use_cuda = True)
text = 'I really want to travel to NY'
model.generate(text)
# expected output 'I really want to go to NY'
You may want to have a look at the slots which were labeled as important by the tagger. Use 'show_template' variable
model.generate(text, show_template = True)
# expected output
# I <extra_id_0> want to <extra_id_1> to NY
# 'I really want to go to NY'
- Downloads last month
- 4
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.