Effective named entity recognition for idiosyncratic web collections

Named Entity Recognition (NER) plays an important role in a variety of online information management tasks includ-ing text categorization, document clustering, and faceted search. While recent NER systems can achieve near-human performance on certain documents like news articles, they still remain highly domain-specific and thus cannot effec-tively identify entities such as original technical concepts in scientific documents. In this work, we propose novel ap-proaches for NER on distinctive document collections (such as scientific articles) based on n-grams inspection and clas-sification. We design and evaluate several entity recogni-tion features—ranging from well-known part-of-speech tags to n-gram co-location statistics and decision trees—to clas-sify candidates. In addition, we show how the use of exter-nal knowledge bases (either specific like DBLP or generic like DBPedia) can be leveraged to improve the effectiveness of NER for idiosyncratic collections. We evaluate our sys-tem on two test collections created from a set of Computer Science and Physics papers and compare it against state-of-the-art supervised methods. Experimental results show that a careful combination of the features we propose yield up to 85 % NER accuracy over scientific collections and substan-tially outperforms state-of-the-art approaches such as those based on maximum entropy