Custom Entity Extraction is table-stakes to effectively infuse domain customization capabilities into your AI application. With Watson Discovery, you can extract business-specific entities using rapid customization tools such as Dictionaries, Patterns, Regular Expressions, and now, across all Watson Discovery v2 Public Cloud Plans, machine-learning based Entity Extractor (beta).
Training a machine-learning based Entity Extraction model is typically a very time-consuming task requiring significant data preparation, data labeling, knowledge of named entity recognition algorithms, as well as model orchestration. The Watson Discovery Entity Extraction Beta feature is business-user friendly and provides a single unified experience to define your entity types, iteratively label and train the model, analyze the model performance, deploy the model, and apply it your collection!
With this tool, the domain expert of the organization can label examples in a rapid and consistent way using an active learning-based component such as Suggestions, as well as Bulk Labeling. The labeling, training and evaluation experience is seamlessly coupled to enable rapid model improvement. Further, applying the model to your collection results in automatic creation of Search Facets, also powering your advanced search use-cases. Coming soon is the ability to visualize these insights from the Custom Entity Extractor on your original PDF documents using the Enhanced Document Preview.
This video details the best practices of iteratively labeling & training a robust Entity Extraction model. The datasets used in this video can be found here.
As with all our Beta releases, we'd love to get your feedback. Let us know about your experience with the Custom Entity Extractor and happy building!
For more context about this feature & recent improvements, you can refer to a recent press release.