mono
Model implementation for mono cross-encoder models.
A mono cross-encoder model, such as MonoBERT or MonoT5, maximizes retrieval accuracy by processing the user’s query and the target document simultaneously. Instead of encoding texts separately like DPR or delaying their interaction like ColBERT, a cross-encoder combines the query and document into a single text sequence before passing them through the neural network. This “early interaction” allows every word in the query to deeply contextualize with every word in the document, producing a highly precise relevance score. However, because this architecture requires processing every potential query and document pair together from scratch, it is computationally prohibitive for large databases and is instead used almost exclusively as a second-stage re-ranker to carefully sort a small list of candidate documents already found by faster models.
Originally introduced in Passage Re-ranking with BERT.
Classes
|
Configuration class for mono cross-encoder models. |
|
|
|