Relation extraction is the task of recognizing and characterizing a particular relationship between two or more entities in text. Depending on the languages in which annotated named entity relation corpora are available, relation extraction has been studied using different machine learning methods, including supervised, semi-supervised and even unsupervised method. Those studies focus on English language and other resources rich languages. However, relation extraction in Uyghur language, which is ethnic minority language that wildly used in Xinjiang Uyghur Autonomous region of China, there are two problems: 1) there are no studies have reported regarding with relation extraction method. 2) the existing annotated Uyghur named entity relation corpus size is relatively small. To address these issues, we utilized the existing Uyghur named entity relation annotated corpus which only contains a small amount of annotated news articles to propose a hybrid semi-automatic method of expanding existing annotated corpus. Our method is based on Conditional Random Fields(CRFs) followed by some rule based post processing and manual correction. In this way, we expanded the corpus size by three times than the existing one.
@InProceedings{ABIDEREXITI18.15, author = {Kahaerjiang Abiderexiti and Ayiguli Halike}, title = {Semi-Automatic Corpus Expansion for Uyghur Named Entity Relation based on a Hybrid Method }, booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)}, year = {2018}, month = {may}, date = {7-12}, location = {Miyazaki, Japan}, editor = {Erhong Yang and Le Sun}, publisher = {European Language Resources Association (ELRA)}, address = {Paris, France}, isbn = {979-10-95546-29-0}, language = {english} }
Summary of the paper