In this tutorial, we’ll use a model from the Hugging Faсe hub to predict text sentiment.

​
Connect a database

We start by connecting a demo database using the CREATE DATABASE statement.

CREATE DATABASE example_db
WITH ENGINE = "postgres",
PARAMETERS = {
    "user": "demo_user",
    "password": "demo_password",
    "host": "samples.mindsdb.com",
    "port": "5432",
    "database": "demo",
    "schema": "demo_data"
    };

Let’s preview the user_comments table.

SELECT *
FROM example_db.user_comments;

​
Create a Hugging Face model

Our Hugging Face integration automatically manages downloading and deploying of pre-trained transformers from Hugging Face’s hub. For example, we can download a transformer which has been trained to classify the sentiment of text.

CREATE MODEL sentiment_classifier
PREDICT sentiment
USING
   engine='huggingface',
   model_name= 'cardiffnlp/twitter-roberta-base-sentiment',
   task='text-classification',
   input_column = 'comment',
   labels=['negative','neutral','positive'];

To create a model in MindsDB, we use the CREATE MODEL statement. Next, we define the target column using the PREDICT clause. Finally, we specify all required parameters in the USING clause.

Once the above query is executed, we can check the status of the creation process:

DESCRIBE sentiment_classifier;

​
Make predictions

Once the status is complete, the behavior is the same as with any other AI table you can query it and provide input data in the WHERE clause, like this:

SELECT * FROM sentiment_classifier
WHERE comment='It is really easy to do NLP with MindsDB';

The above query should predict the comment as ‘positive’.

We can also make batch predictions by joining the input data table with the model, like this:

SELECT input.comment, model.sentiment
FROM example_db.user_comments AS input
JOIN sentiment_classifier AS model;