作为 Google NLP 中的人的一部分，我如何训练说泰米尔人姓名（泰米尔人种族）

Question

How do I train say Tamil names (Tamil ethnicity) as part of Person in Google NLP

Names like

1. Pazhanivel

2. Kuzhanthaivel

Refer the attached screenshot from NLP-demo

Answer 1

In your description you show a demo run of the Natural Language API. This demo has in its backend a Natural Language model trained with a huge amount of labelled data and you are choosing an entity extraction task.

To use the Natural Language API to perform an entity task you have two options, which you can see if you scroll down a little bit in [1]: Entity Analysis and Custom Entity Extraction.

The first option is what you see in the demo, it takes an entity analysis model and performs the task with the input text. The labels learned by the model cannot be changed so it's not able to recognize the names Pazhanivel and Kuzhanthaivel.

The second option is what you are looking for, a custom entity extraction. To perform a training with this you can check the quickstart [2]. This implies training a custom entity recognition model which is not a simple task, however with Google Cloud Platform you can launch a model training using only the UI and so it becomes manageable. Also in the Google Cloud documentation you have explanations about all the steps to prepare the data [3], training models [4] and so on.

Do not hesitate to ask if you have any question.

[1] https://cloud.google.com/natural-language/#section-5

[2] https://cloud.google.com/natural-language/automl/docs/quickstart

[3] https://cloud.google.com/natural-language/automl/docs/prepare?_ga=2.66840366.-1201206137.1584215642

[4] https://cloud.google.com/natural-language/automl/docs/models