LSA or BERTtransformers? Which one is better to use for real-time semantic Similairty and semantic clustering of short sentence?

Question

Which one technique is more applicable for real-time measurement of semantic similarity and semantic clustering? For example, classifying students' answers during the online session into different clusters based on their similarities. Sentences could be from any domain.

Answer 1

if you have a large amount of dataset,you can choose bert.

But,I recommend lsa,which is more simple and effective at small amount of dataset.