tBERT: Topic Models and BERT Joining Forces for Semantic Similarity Detection
What it is
This is the official code repository for the ACL 2020 paper tBERT, which combines topic models (like LDA) with BERT for semantic similarity tasks. It provides scripts to train and evaluate the model on datasets like MSRP, Quora, and Semeval using TensorFlow 1.x.
The problem it solves
Standard BERT models may miss global document-level or topic-level semantic structures that are useful for similarity detection, which tBERT addresses by fusing topic vectors with BERT embeddings.
What you could use it for
- Researching the integration of topic models with transformer architectures
- Reproducing the results of the ACL 2020 tBERT paper
- Performing semantic similarity detection on short text pairs
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.