Flair is a powerful open-source library for natural language processing. It is mainly used to get insight from text extraction, word embedding, named entity recognition, parts of speech tagging, and text classification. All these features are pre-trained in flair for NLP models. It also supports biomedical data that is more than 32 biomedical datasets already using flair library for natural language processing tasks. Easily integrated with Pytorch NLP framework for embedding in document and sentence.
Humboldt University of Berlin and friends mainly develop flair. The Humboldt University of Berlin maintains the Flair library and has already done more than a hundred industry project implementations and research-based projects using Flair.
Let’s look at Flair’s performance based on the nlp task such as named entity recognition, parts of speech tagging, and chunking with their accuracy in the table below.
pip install flair
conda install -c bioconda flair
First, import sentences from flair’s data library, then import the model for SequenceTagger. Make a sentence using the Sentence object, then load Named entity recognition on SequenceTagger, then run the code.
For an example of the flair model, see the code below.
from flair.data import Sentence from flair.models import SequenceTagger # make a sentence sentence = Sentence('I love India .') # load the NER tagger tagger = SequenceTagger.load('ner') # run NER over sentence tagger.predict(sentence)
Flair has the following pre-trained models for NLP Tasks:
- Name-Entity Recognition
- Parts-of-Speech Tagging
- Text Classification
- Training Custom Models
In the flair library, there is a predefined tokenizer using the segtok library of python. To
use the tokenization just the “use_tokenizer” flag value is true. If not want to implement the write false. We can also define the label of each sentence and its related topic using the function add_tag.
For example, see the code below:
from flair.data import Sentence # Make a sentence object by passing an untokenized string and the 'use_tokenizer' flag untokenized_sentence = Sentence('The grass is green.', use_tokenizer=False # Print the object to see what's in there print(untokenized_sentence)
In this case, no tokenization occurs use_tokenizer is false.
Here is the list of embedding in the library. We will learn about flair library in detail, and there code implementation.
Effective embeddings are contextual string embeddings that capture latent syntactic-semantic data that goes beyond standard word embedding. The main differences are:
(1) Without any clear notion of vocabulary, they are educated and thus essentially model words as character sequences.
(2) they are contextualized by their surrounding text, meaning that depending on their contextual use, the same word will have distinct embeddings.
from flair.embeddings import FlairEmbeddings # init embedding flair_embedding_forward = FlairEmbeddings('news-forward') # create a sentence sentence = Sentence('The grass is green .') # embed words in sentence flair_embedding_forward.embed(sentence)
Training a Text Classification Model:
We are training a text classifier over the TREC-6 corpus, using a combination of simple GloVe embeddings and Flair embeddings.
from flair.data import Corpus from flair.datasets import TREC_6 from flair.embeddings import WordEmbeddings, FlairEmbeddings, DocumentRNNEmbeddings from flair.models import TextClassifier from flair.trainers import ModelTrainer # 1. get the corpus corpus: Corpus = TREC_6() # 2. create the label dictionary label_dict = corpus.make_label_dictionary() # 3. make a list of word embeddings word_embeddings = [WordEmbeddings('glove')] # 4. initialize document embedding by passing a list of word embeddings # Can choose between many RNN types (GRU by default, to change use rnn_type parameter) document_embeddings = DocumentRNNEmbeddings(word_embeddings, hidden_size=256) # 5. create the text classifier classifier = TextClassifier(document_embeddings, label_dictionary=label_dict) # 6. initialize the text classifier trainer trainer = ModelTrainer(classifier, corpus) # 7. start the training trainer.train('resources/taggers/trec', learning_rate=0.1, mini_batch_size=32, anneal_factor=0.5, patience=5, max_epochs=150)
We learn about the Flair open-source library for NLP problems. We also covered the area about NLP and the use of Flair to solve the tasks and their use in the industry. Some important Flair pipelines and their code in the development of pre-trained NLP models.
Subscribe to our NewsletterGet the latest updates and relevant offers by sharing your email.
Join Our Telegram Group. Be part of an engaging online community. Join Here.
Amit Singh is Data Scientist, graduated in Computer Science and Engineering. Data Science writer at Analytics India Magazine.