A web prototype for detecting chemical compounds and drugs Daniel Sánchez-Cisneros 1 , Sara Lana-Serrano 2 , Isabel Segura-Bedmar 1 , Leonardo Campillos 3 , Paloma Martínez 3 1 ComputerScienceDepartment, Universidad Carlos III de Madrid, Spain {dscisner, isegura,pmf}@inf.uc3m.es 2 Universidad Politécnica de Madrid, Spain slana@diatel.upm.es 3 Universidad Autónoma de Madrid leonardo.campillos@uam.es Abstract. This paper introduces a web prototype for named entity recognition of chemical compounds and drugs. The tool is based on a system developed to participate in the ChemDNER task organized as part of Biocreative 2013 work- shop. The system combines the ChemSpot tool as well as a set of semantic- based rules, which were defined according to the guidelines provided to task participants. The prototype is available at http://multimedica.uc3m.es:8080/biocreative2013demo/ Keywords: Drug named entity recognition, information extraction 1 Introduction Most research on named entity recognition (NER) in the biomedical domain are based on dictionary based methods and Supervised Machine Learning (SML) methods. The main problems with the former approach are their domain dependency and their ina- bility to recognize terms not included in the dictionaries. Machine learning techniques build classification models based on annotated corpus and produce the best results [1], although they require annotated corpora. Current trends try to develop hybrid systems that combine best of two approaches. In this work we present a prototype that combines existing systems such as ChemSpot [2] and Metamap [3] with gazetteers extracted from biomedical resources such as