Annotation guidelines for entity extraction projects
In an entity extraction project, you must annotate all entities that appear in the document.
Q: If an entity is annotated in one document, does it need to be annotated again in another document?
A: Yes, it does.
Annotated entities cannot contain line feeds ("\n") or periods (.). If they do, the entities will not be detected. However, this does not affect model training.
Are there advanced rules engine settings for entity extraction projects?
Q: My documents have a fixed format, and the fields that I need to extract have clear and consistent context. Is there a fast and effective extraction method that does not require extensive annotation?
A: Yes, there is. We offer a rules engine module to improve the performance and efficiency of entity extraction models. To use this feature, contact us to be added to the whitelist. The DingTalk group number is 23142693. You can configure the rules engine when you create a model. The rules engine interface is shown below:
You can configure rules that combine regular expressions, dictionaries, wildcard characters, and entities extracted by the model. Use the rule preview to test whether your rules work as expected.
Rule configuration example:
I need to extract the name, gender, and date of birth of the plaintiff and the defendant from the following court document:
Plaintiff: Cheng Xiaoer, female, resident of Yuhang District, Hangzhou, Zhejiang Province.\n\nDefendant: Wang Moumou, male, born on October 1, 2019, Han ethnicity.
You can use the rules engine to configure a regular expression that extracts these fields:
Add rule group -> Set Entity Type to Plaintiff -> Set Rule Element to Regular Expression -> Enter (?<=Plaintiff:)([^,]+)(?=,) in the Value field.
After you configure the rules, you can click Rule Preview to check them:
Are there common fields available for extraction?
The platform can extract common fields, such as dates and phone numbers. You can select the fields you need, and the model you create will extract them. Model creation requires annotated data. To use this feature, you can annotate 10 data entries to train a basic model. In the advanced settings, make sure to select the common fields that you want to use.
To access this feature, go to Entity Extraction Project > Model Hub > Create Model > Advanced Settings: