.png){kind=icon}
Return
Decisive vector guided column annotation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2024.110958.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Column annotation aims to annotate tabular column data with semantic labels and plays an essential role in a variety of applications such as data cleaning, data discovery, schema matching, semantic search and data governance. Recently, deep learning techniques have been introduced for column annotation and have successfully achieved substantial gains in performance. However, their incapability of combating with noisy labels is becoming thorny increasingly, as training data corrupted with noisy labels is ubiquitous in the industrial scenarios. To cope with this issue, this paper develops a novel learning paradigm, which exploits decisive ( i.e. , semi-hard) vectors to guide the model training with noisy labels. Extensive experimental results on the same benchmarks have demonstrated the effectiveness of our method over the state-of-the-art alternatives.
Keywords:
Noisy labels
Decisive vector learning
Column annotation
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W
Organization
Cited Papers
ShareSaveWeighting and pruning based ensemble deep random vector functional link network for tabular data classification.png)
PATTERN RECOGNITION
OAAI
PATTERN RECOGNITIONIF7.6
OAAIShareSaveShareSave
ShareSaveShareSave
View more