Brief description: 

We researched a novel approach to address fundamental natural language processing (NLP) tasks such as syntactic and semantic analysis. Our approach outperforms all existing methods by a significant margin and represents the new state-of-the-art across various core NLP tasks. Key to our approach is unsupervised pre-training of character-level neural language models that are then utilized as a "general model of language understanding" in downstream NLP tasks. The novel approach is described and evaluated in detail in our full paper publication at the International Conference on Computational Linguistics (COLING 2018), titled "Contextual String Embeddings for Sequence Labeling". We made our approach publicly available in form of an open source project called "Flair". The library is easily installable from a central Python repository and includes pre-trained models for a wide range of NLP tasks such as part-of-speech tagging, named entity recognition and word sense disambiguation. It therefore allows third parties to integrate and leverage state-of-the-art NLP in their use cases. Flair has since received significant attention from industry and research and been used both within and outside Zalando for state-of-the-art NLP.

Main Features: 
  • A powerful NLP library. Flair allows you to apply our state-of-the-art natural language processing (NLP) models to your text, such as named entity recognition (NER), part-of-speech tagging (PoS), sense disambiguation and classification.
  • Multilingual. Thanks to the Flair community, we support a rapidly growing number of languages. We also now include 'one model, many languages' taggers, i.e. single models that predict PoS or NER tags for input text in various languages.
  • A text embedding library. Flair has simple interfaces that allow you to use and combine different word and document embeddings, including our proposed Flair embeddings, BERT embeddings and ELMo embeddings.
  • A Pytorch NLP framework. Our framework builds directly on Pytorch, making it easy to train your own models and experiment with new approaches using Flair embeddings and classes.
Areas of Application: 

Natural Language Processing (NLP)

Technological novelty: 

Flair outperforms the previous best methods on a range of NLP tasks: Named Entity Recognition, Part of Speech Tagging, Chunking.

Component / Service / App


fashion Brain
BDV Reference categories: 
Data processing architectures
Data Visualisation and Interaction
Information Service activities
Readiness Level: