Open-source Large Language Models are Strong Zero-shot Query Likelihood Models for Document Ranking.
Shengyao ZhuangBing LiuBevan KoopmanGuido ZucconPublished in: EMNLP (Findings) (2023)
Keyphrases
- document ranking
- language model
- relevance model
- probabilistic model
- query expansion
- document retrieval
- language modelling
- probabilistic retrieval models
- language modeling
- statistical language modeling
- federated search
- query terms
- information retrieval
- n gram
- query specific
- smoothing methods
- text retrieval
- pseudo relevance feedback
- retrieval model
- retrieval effectiveness
- expert search
- vector space model
- original query
- translation model
- ranked list
- relevant documents
- retrieved documents
- test collection
- multi modal
- information retrieval systems
- relevance feedback
- image annotation
- ad hoc retrieval
- generative model
- bayesian networks
- search engine