|
Title:
|
EVALUATION OF NAMED ENTITY RECOGNITION
FOR THE GERMAN E-COMMERCE DOMAIN |
|
Author(s):
|
Sergej Denisov and Frederik S. Bäumer |
|
ISBN:
|
978-989-8704-34-4 |
|
Editors:
|
Pedro Isaías and Hans Weghorn |
|
Year:
|
2021 |
|
Edition:
|
Single |
|
Type:
|
Full |
|
First Page:
|
226 |
|
Last Page:
|
230 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Large online marketplaces offer search engines as an important navigation aid for their customers to navigate through the
enormous number of different products and merchants. The quality of the search results depends to a large extent on the
product information provided by the retailers. One way to improve search quality is to perform linguistic enhancement of
the product data. For this, we use Named Entity Recognition to identify specific e-commerce entity types and give them
higher weighting in the search. Typical entity types for the e-commerce domain are products, brands, and various product
attributes. Recognition of e-commerce entity types for the German language remains a challenge due to the limited
availability of existing resources and linguistic complexity. We address this challenge by acquiring data from two online
e-commerce marketplaces to create six NER datasets based on German product titles and descriptions. Across these
datasets, we evaluate the NER performance of the state-of-the-art models mBERT, GermanBERT, and XLM-RoBERTa.
As a result, the XLM-RoBERTa model archived the best performance with an F1 score of 0.8611 averaged over all
datasets. |
|
|
|
|
|
|
