Hi Allison. What you wrote below is not true:

If you are looking to play around with pre-installed data sets, and therefore don’t need to access the spark session, you can get started with the following two lines:
import sparknlp
sparknlp.start()
In my case, I need the SparkSession to load my data from the parquet file, so I’ll add .config(“spark.jars.packages”, “com.johnsnowlabs.nlp:spark-nlp_2.11:2.3.5”) to my SparkSession.builder

sparknlp.start() already returns a Spark session with the packages loaded. So you can just write spark = sparknlp.start() and then it will be same as what session builder does. Actually, Spark NLP already does that on the backend.

For more article about Spark NLP, you may check these

Text Classification in Spark NLP with Bert and Universal Sentence Encoders

https://towardsdatascience.com/named-entity-recognition-ner-with-bert-in-spark-nlp-874df20d1d77