SRCB at SemEval-2023 Task 2: A System of Complex Named Entity Recognition with External Knowledge

Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 671–678 July 13-14, 2023 ©2023 Association for Computational Linguistics SRCB at SemEval-2023 Task 2: A System of Complex Named Entity Recognition with External Knowledge Yuming Zhang, Hongyu Li, Yongwei Zhang, Shanshan Jiang, Bin Dong Ricoh Software Research Center (Beijing) Co., Ltd. {Yuming.Zhang1, Hongyu.Li, Yongwei.Zhang, Shanshan.Jiang, Bin.Dong}@cn.ricoh.com Abstract The MultiCoNER II shared task aims at de- tecting semantically ambiguous and complex named entities in short and low-context settings for multiple languages. The lack of context makes the recognition of ambiguous named en- tities challenging. To alleviate this issue, our team SRCB proposes an external knowledge based system, where we utilize 3 different types of external knowledge retrieved in different ways. Given an original text, our system re- trieves the possible labels and the descriptions for each potential entity detected by a mention detection model. And we also retrieve a related document as extra context from Wikipedia for each original text. We concatenate the original text with the external knowledge as the input of NER models. The informative contextual representations with external knowledge signif- icantly improve the NER performance in both Chinese and English tracks. Our system win the 3rd place in the Chinese track and the 6th place in the English track. 1 Introduction The task of Multilingual Complex Named Entity Recognition (MultiCoNER) (Malmasi et al., 2022b; Fetahu et al., 2023b) aims to deal with the complex named entity recognition problem. Unlike the or- dinary entities in most Named Entity Recognition (NER) tasks, these complex entities can be com- posed of any form of a language. Compared with the 6 categories defined in MultiCoNER (Malmasi et al., 2022a), MultiCoNER II (Fetahu et al., 2023a) furtherly defines 33 fine-grained categories for the complex entities and provides them with short and uncased texts (in short and low-context settings). Moreover, at the stage of test, some spelling mis- takes are added into the test set, causing data dis- turbance. Recognizing complex named entities in such settings is challenging for NER systems. In practice, for a professional annotator, a solu- tion to deal with these complex entities is using external knowledge. The external knowledge in- cludes professional knowledge of their own, infor- mation presented by search engines and contents in databases. Retrieving such related external knowl- edge can help to determine whether a fragment of sentence is an entity or not and eliminate the ambiguity of complex entities (Wang et al., 2019). Meanwhile, (Wang et al., 2022) proposed a general knowledge-based NER system retrieve the related documents of the input sentence as external knowl- edge which is proved to be very effective in short and low-context named entity recognition. There- fore, we believe that introducing external knowl- edge to NER systems is still a straight-forward and effective way to improve the performance of com- plex entity recognition. In this paper, we propose an external knowledge based system, which is composed of a knowledge retrieval module and an NER module. In the knowl- edge …

Content truncated to 500 words