LG AI Research & KAIST at EHRSQL 2024: Self-Training Large Language Models with Pseudo-Labeled Unanswerable Questions for a Reliable Text-to-SQL System on EHRs.
Yongrae JoSeongyun LeeMinju SeoSung Ju HwangMoontae LeePublished in: CoRR (2024)
Keyphrases
- language model
- information retrieval
- language modeling
- n gram
- document level
- document retrieval
- probabilistic model
- training set
- relational databases
- language modelling
- text retrieval
- speech recognition
- retrieval model
- statistical language models
- query expansion
- language models for information retrieval
- vector space model
- context sensitive
- smoothing methods
- test collection
- electronic health records
- text mining
- document ranking
- semi supervised learning
- supervised learning
- relevance model
- training data
- pseudo relevance feedback
- term dependencies
- semi supervised
- document analysis
- passage retrieval
- free text
- query terms
- retrieval systems
- web documents
- unlabeled data
- okapi bm
- improve retrieval effectiveness
- spoken term detection