(C) 2019-2020 by Damir Cavar
Contributors to previous versions: Oren Baldinger, Maanvitha Gongalla, Anurag Kumar, Murali Kammili
Brought to you by the NLP-Lab.org!
Flair v 4.5 wrapper for JSON-NLP. Flair provides state-of-the-art embeddings, and tagging capabilities, in particular, POS-tagging, NER, shallow syntax chunking, and semantic frame detection.
We provide a FlairPipeline
class, with the following parameters for ease of use:
lang
: defaults toen
. Different languages support different models, see the Flair Docs for details.use_ontonotes
: defaults toFalse
. Whether or not to use 4-class (True) or 12-class (False) NER tagging.fast
: defaults toTrue
. Whether or not to use the smaller, faster, but slightly less accurate versions of the models.use_embeddings
: defaults to ''. Passingdefault
will map toglove,multi-forward,multi-backward
, the recommended stacked-embedding configuration.char_embeddings
: defaults toFalse
. Whether or not to include character-level embeddings.bpe_size
: defaults to 0. If you want to include Byte-Pair Encodings, set this value to 50, 100, 200, or 300. See more at Flair Embeddings Docs.pos
: defaults toTrue
. Whether or not to include language-specific part-of-speech tags.sentinment
: defaults toTrue
. Whether or not to include sentiment analysis, if it is available for the given language.
Tagging and Embedding models are downloaded automatically the first time they are called. This may take a while depending on your internet connection.
The JSON-NLP repository provides a Microservice class, with a pre-built implementation of Flask. To run it, execute:
python flairjsonnlp/server.py
Since server.py
extends the Flask app, a WSGI file would contain:
from flairjsonnlp.server import app as application
The microservice exposes the following URIs:
- /expressions
- /token_list
These URIs are shortcuts to disable the other components of the parse. In all cases, tokenList
will be included in the JSON-NLP
output. An example url is:
http://localhost:5000/expressions?text=I am a sentence
Text is provided to the microservice with the text
parameter, via either GET
or POST
. If you pass url
as a parameter, the microservice will scrape that url and process the text of the website.
The additional Flair parameters can be passed as parameters as well.
Here is an example GET
call:
http://localhost:5000?lang=de&constituents=0&text=Ich bin ein Berliner.