A Regional News Corpora for Contextualized Entity Discovery and Linking

A Regional News Corpora for Contextualized Entity Discovery and Linking Adrian M.P. Bras¸oveanu*, Lyndon J.B. Nixon*, Albert Weichselbraun†, Arno Scharl* * Department of New Media Technology, MODUL University Vienna, Am Kahlenberg 1, 1190 Vienna, Austria † Swiss Institute for Information Research, University of Applied Sciences Chur, Pulverm¨uhlestrasse 57, 7000 Chur, Switzerland {adrian.brasoveanu,lyndon.nixon,arno.scharl}@modul.ac.at albert.weichselbraun@htwchur.ch Abstract This paper presents a German corpus for Named Entity Linking (NEL) and Knowledge Base Population (KBP) tasks. We describe the annotation guideline, the annotation process, NIL clustering techniques and conversion to popular NEL formats such as NIF and TAC that have been used to construct this corpus based on news transcripts from the German regional broadcaster RBB (Rundfunk Berlin Brandenburg). Since creating such language resources requires significant effort, the paper also discusses how to derive additional evaluation resources for tasks like named entity contextualization or ontology enrichment by exploiting the links between named entities from the annotated corpus. The paper concludes with an evaluation that shows how several well-known NEL tools per- form on the corpus, a discussion of the evaluation results, and with suggestions on how to keep evaluation corpora and datasets up to date. Keywords: Named entity linking, corpora, contextualization, knowledge base population, television news transcripts, evaluation. 1 Introduction With an ever increasing amount of news and social media coverage comes a need to extract relevant information automatically. A method that can help in this endeavour is Named Entity Linking (NEL), the grounding to knowledge base entries of the mentions extracted from texts. Knowl- edge Base Population (KBP) (Ji et al., 2014), the process through which the entities identified by NEL systems are used to populate new knowledge bases, is another useful technique for exploring the relations between entities. However, since NEL or KBP evaluation tasks might require a new corpus or at least a new gold standard, and the creation of such resources requires significant effort, there is a desire to automate steps in the corpus creation process. By exploiting the knowledge graph built with named entities from a corpus, new gold standards can be created for specific tasks like KBP, entity contextualization or enrichment. In contrast to research focused on social media (Bontcheva and Rout, 2012), we used regional news and analyzed transcripts from the German regional broadcaster RBB (Rundfunk Berlin Brandenburg). This paper presents the corpus designed for KBP evaluations and shows how new datasets for specific evaluation tasks are automatically created by exploiting the links between the entities from the graph associated with the corpus. The rest of this paper is organized as follows: Section 2 presents the dataset and the process used to generate it: cre- ating annotations; calculating inter-annotator agreements; NIL clustering of the unlinked entities; and conversion to popular NEL evaluation formats such as TAC KBP (Ji et al., 2014) and NIF (Hellmann et al., 2013). Section 3 describes the knowledge engineering methods used for generating new datasets from the entity graph built from the corpus for tasks like ontology population and enrich- ment, contextualized entity extraction …

Content truncated to 500 words