Amorphic Entity Recognizers provides the capability of training a custom entity recognizer model that can be used to generate AI content on unstructured dataset - text, pdf, images and audio datasets without any prior machine learning knowledge.
New Entity Recognizers¶
This section will display the list of active (TRAINED) or inactive (TRAINING) stages.
In order to get started with training a custom entity recognizer in Amorphic, we need the following information:
1. Training Dataset : Amorphic Dataset containing the raw text of data on which the CER model will be trained on. As a general rule, you should include as much relevant data as possible. The more data that you add, the more context the model can have on which to train itself. This raw text can be ingested in an Amorphic Dataset. This data can have different document text all in one file seperated by a new line or can be seperated into different files as well.
2. Entity Lists: Provides only a limited context. It only uses a selection from the specific entities list so that Amazon Comprehend can train on identifying the custom entity in the backend. For example, Text - iPhone X, Type - DEVICE
The relevant text are categorized as an entity type in this data.
Following information is required to train an entity recognizer in Amorphic.
- Create a entity recognizer training dataset
- Go to the Amorphic Machine Learning > Entity Recognizer section and Create a new entity recognizer
- Provide a training dataset from the dropwdown
- Upload the entity list file
- Add new custom entity types that the Amorphic CER should recognize for example, DEVICE.
- Click on Create
This will create an Amorphic CER object which will go through Submitted, Training and Trained phases. Only a Trained Amorphic CER object is be available for selection.
Once the Amorphic ER object is trained, the user can use the ER object on a dataset file as following:
Following are the steps for the same:
- Go to the Dataset Files
- Select “Apply ML” on the dataset files.
- Select “Entity Recognizer” as Type.
- Select the Amorphic Entity Recognizer object.
- Click on “Submit”
6. Once submitted, the Amorphic CER object will run on the dataset file. Upon completion of the run, the results can be tracked and are available in the “Reload Ivocations” tab or in the “View AI/ML Results” option on the top right corner of dataset files.
In order to view the results of the entities, go to the “Entities” section of AI/ML results.