A Corpus Study and Annotation Schema for Named Entity Recognition and Relation Extraction of Business Products

A Corpus Study and Annotation Schema for Named Entity Recognition and Relation Extraction of Business Products Saskia Sch¨on, Veselina Mironova, Aleksandra Gabryszak, Leonhard Hennig DFKI GmbH Berlin, Germany {firstname.lastname}@dfki.de Abstract Recognizing non-standard entity types and relations, such as B2B products, product classes and their producers, in news and forum texts is important in application areas such as supply chain monitoring and market research. However, there is a decided lack of annotated corpora and annotation guidelines in this domain. In this work, we present a corpus study, an annotation schema and associated guidelines, for the annotation of product entity and company-product relation mentions. We find that although product mentions are often realized as noun phrases, defining their exact extent is difficult due to high boundary ambiguity and the broad syntactic and semantic variety of their surface realizations. We also describe our ongoing annotation effort, and present a preliminary corpus of English web and social media documents annotated according to the proposed guidelines. Keywords: Named Entity Recognition, Relation Extraction, Information Extraction 1. Introduction Recognizing non-standard entity and relation types is an important task in many real-world information extraction applications like relation extraction, knowledge base con- struction and question answering. In areas such as mar- ket research and supply chain management, many com- panies would benefit from systems that automatically and continuously acquire up-to-date information about produc- ers, vendors and other suppliers of specific parts, products, new technologies and components. Similarly, the construc- tion of knowledge graphs that store supplier and vendor re- lationships would clearly benefit from information extrac- tion approaches by reducing the manual effort required to create and maintain such databases. For example, in both scenarios it would be useful to extract information about e.g. a CompanyProvidesProduct relation from a news text like “Sensata Technologies’ products include speed sen- sors, motor protectors, and magnetic-hydraulic circuit break- ers”, where the product argument refers to a non-consumer product or product class entity such as “speed sensors” or “magnetic-hydraulic circuit breakers”. However, when it comes to such specific domains, devel- oping named entity recognition algorithms is severely ham- pered by the lack of publicly available training data and the difficulty of accessing existing dictionary-type resources, such as product catalogs. Many available named entity recognition corpora consist of general news articles (Tjong Kim Sang and De Meulder, 2003; Doddington et al., 2004; Weischedel et al., 2013), while information about B2B products is typically available on non-journalistic, special- ized web portals and forums. Product mentions, as in the example above, are often general noun phrases, instead of proper names, which increases the difficulty of detecting them using gazetteer-based approaches. In addition, rela- tional information about companies and their products is very limited in freely available knowledge bases (KB), such as Freebase (Bollacker et al., 2008), Wikidata (Vrandeˇci´c and Kr¨otzsch, 2014), or DBpedia (Auer et al., 2007), since these KBs are in large parts based on Wikipedia, which aims to exclude commercial, non-encyclopedic informa- tion. For example, DBpedia contains only approximately 60, 000 triples for the …

Content truncated to 500 words