Auto-labelling entities in low-resource text: a geological case study
| Autor: | Paul Duuring, Eun-Jung Holden, Majigsuren Enkhsaikhan, Wei Liu |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Information retrieval
Computer science business.industry Deep learning 02 engineering and technology computer.software_genre Technical documentation Domain (software engineering) Human-Computer Interaction Jargon Annotation Named-entity recognition Artificial Intelligence Hardware and Architecture 020204 information systems Labelling 0202 electrical engineering electronic engineering information engineering Artificial intelligence Transfer of learning business computer Software Information Systems |
| Zdroj: | Knowledge and Information Systems. 63:695-715 |
| ISSN: | 0219-3116 0219-1377 |
| Popis: | Studies on named entity recognition (NER) often require a substantial amount of human-annotated training data. This makes technical domain-specific NER from industry data especially challenging as labelled data are scarce. Despite English as the surface language, technical jargon and writing conventions used in technical documents render the low-resource language challenges where techniques such as transfer learning hardly work. Relieving labour intensive annotations using automatic labelling is thus an important research topic, seeking ways to obtain labelled data quickly and consistently. In this work, we propose an iterative deep learning NER framework using distant supervision for automatic labelling of domain-specific datasets. The framework is applied to mineral exploration reports and produced a large BIO-annotated dataset with six geological categories. This quality-labelled dataset, OzROCK, is made publicly available to support future research on technical domain NER. Experimental results demonstrated the effectiveness of this approach, further confirmed by domain experts. The generalisation ability is verified by applying the framework to two other datasets: one for disease names and the other for chemical names. Overall, our approach can effectively reduce annotation efforts by identifying a much smaller subset, that is challenging for automatic labelling thus requires attention from human experts. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink