No abstract

No abstract

No abstract

No abstract

Named entity recognition (NER) is one of the tasks in natural language processing that can greatly benefit from the use of external knowledge sources. We propose a named entity recognition framework composed of knowledge-based feature extractors and a deep learning model including contextual word embeddings, long short-term memory (LSTM) layers and conditional random fields (CRF) inference layer. We use an entity linking module to integrate our system with Wikipedia. The combination of effective neural architecture and external resources allows us to obtain state-of-the-art results on recognition of Polish proper names. We evaluate our model on data from PolEval 2018 1 NER challenge on which it outperforms other methods, reducing the error rate by 22.4% compared to the winning solution. Our work shows that combining neural NER model and entity linking model with a knowledge base is more effective in recognizing named entities than using NER model alone.A PREPRINT -NOVEMBER 27, 2018 recently, state-of-the-art NER systems employed word representations based on pre-trained language models, either replacing classic word embedding approaches [28,29] or using both representations jointly [1].In Polish, early tools for named entity recognition were based on heuristic rules and pre-defined grammars [30,14] or CRF models with hand-crafted features [38,34,20,21]. Pohl [32] used OpenCyc and Wikipedia to build purely knowledge-based NER system. Only recently methods involving deep learning were introduced [3, 23].While modern named entity recogntion methods have made considerable progress in exploiting contextual information and long term dependencies in text, in some cases it is not sufficient to accurately recognize a named entity. When the context does not provide enough information, model should be able to use external knowledge to help with the detection and classification. Such a need exists, for example, in the case of abbreviations or highly ambiguous phrases that can refer to several different entities. Therefore, we believe that the problem of integrating knowledge sources with NER models should be explored. In this work, we focus on named entity recognition for Polish language. We show how such model can be integrated with Wikipedia and how can we improve its performance by using an external knowledge base. The method proposed in this publication may also be used in other languages. ContributionsOur contributions are the following: 1) We propose a named entity recognition system for Polish that combines deep learning architecture with knowledge-based feature extractors, achieving state-of-the art results for this task. 2) We propose a method utilizing an entity linking model based on Wikipedia to improve the accuracy of named entity recognition. Additionally, we release a tool for efficient labeling of Wikipedia's articles. 3) We make the source code of our method available, along with pre-trained models for NER, pre-trained Polish Word2Vec [24] embeddings and ELMo [29] embeddings, labeled data set of articles ...