Chinese Open Relation Extraction and Knowledge Base Establishment

Abstract

Named entity relation extraction is an important subject in the field of information extraction. Although many English extractors have achieved reasonable performance, an effective system for Chinese relation extraction remains undeveloped due to the lack of Chinese annotation corpora and the specificity of Chinese linguistics. Here, we summarize three kinds of unique but common phenomena in Chinese linguistics. In this article, we investigate unsupervised linguistics-based Chinese open relation extraction (ORE), which can automatically discover arbitrary relations without any manually labeled datasets, and research the establishment of a large-scale corpus. By mapping the entity relations into dependency-trees and considering the unique Chinese linguistic characteristics, we propose a novel unsupervised Chinese ORE model based on Dependency Semantic Normal Forms (DSNFs). This model imposes no restrictions on the relative positions among entities and relationships and achieves a high yield by extracting relations mediated by verbs or nouns and processing the parallel clauses. Empirical results from our model demonstrate the effectiveness of this method, which obtains stable performance on four heterogeneous datasets and achieves better precision and recall in comparison with several Chinese ORE systems. Furthermore, a large-scale knowledge base of entity and relation, called COER, is established and published by applying our method to web text, which conquers the trouble of lack of Chinese corpora.

Unsupervised Approaches for Textual Semantic Annotation, A Survey

“…-Consistencies that exist in different aspects, for example, linguistic (Cimiano et al 2004;Jia et al 2018;Sánchez et al 2011), syntactic, or structural (Krapivin et al 2008;Sleiman and Corchuelo 2014). -Information redundancy that exists in various sources (Agichtein and Gravano 2000;Brin 1999;Dill et al 2003b;Etzioni et al 2005).…”

“…-Many works go further into some specific domains, such as Sci-Tech compound entity recognition (Yan et al 2016), biomedical named entity recognition (Song et al 2018), entity extraction from clinical records (Alicante et al 2016;Boytcheva 2018;Henriksson et al 2015), chemical named entity recognition (Swain and Cole 2016; Zhang et al 2016), educational term extraction (Conde et al 2016), cybersecurity concepts extraction (Xiao 2017), Drug Name Recognition (Liu et al 2015b), medical named entity recognition (He and Kayaalp 2008;Kavuluru et al 2013;Skeppstedt 2014), and so forth. -In addition to English, unsupervised Named Entity Recognition has been studied in other languages, including Chinese (Jia et al 2018), Spanish (Copara et al 2016), French (Mosallam et al 2014), Italian (Alicante et al 2016), and Russian (Ivanitskiy et al 2016), as well as crosslingual (Abdel Hady et al 2014), even including a dead language like Sumerian (Liu et al 2015a).…”

“…Another complex practice is the Open Information Extraction (OIE) strategy (Elsahar et al 2017;Gamallo and Garcia 2015;Jia et al 2018), which extracts arbitrary open-domain facts of knowledge from massive corpora, instead of being constrained to an existing template stored in an ontology or knowledge base.…”

Reconstruction of ER Network from Specific Academic Texts for the Governance of MSW‐NIMBY Crisis in China

“…It is based on 24 kinds of dependencies proposed by the LTP (language technology platform) of HIT-SCIR (Harbin Institute of Technology) [71]. It can be generalized into the combination of words, POS-tags, dependency paths, and dependency labels on paths [72]. We defined logical and graphical expression to display sentence rules in accordance with the above labels in the template of semantic annotation as well as extracted ER triples by the certain rules, as shown in Table 2.…”

Triple Trustworthiness Measurement for Knowledge Graph

“…erefore, an increasing number of researchers are commi ed to productively extracting information directly from unstructured web text, such as ORE [2,9,19], NELL [7], and to automatically constructing large-scale knowledge graphs, such as Freebase [3], DBpedia [1], and Wikidata 1 . However, some noises and errors are inevitably introduced in the process of automation.…”