Conclusions

Abstract

Entities and relationships are important structures that can be extracted from a text corpus to represent the factual knowledge inside the corpus. Effective and efficient mining of entity and relation structures from text helps gaining insights from large volume of text data (that are infeasible for human to read through and digest), and enables many downstream applications on understanding, exploring and analyzing the text content. Data analysts and government agents may want to identify person, organization and location entities in news everyday news articles and generate concise and timely summary of news events. Biomedical researchers who cannot digest large amounts of newly-published research papers in relevant areas would need an effective way to extract different relationships between proteins, drugs, and diseases so as to follow the new claims and facts presented in the research community. However, text data is highly variable: corpora covering topics from different domains, written in different genres or languages have typically required for effective processing a wide range of language resources such as grammars, vocabularies, gazetteers. The massive and messy nature of text data post significant challenges to creating tools for automated structuring of unstructured content that scale with text volume.