ViLMA: A Zero-Shot Benchmark for Linguistic and Temporal Grounding in Video-Language Models.
Ilker KesenAndrea PedrottiMustafa DoganMichele CafagnaEmre Can AcikgozLetitia ParcalabescuIacer CalixtoAnette FrankAlbert GattAykut ErdemErkut ErdemPublished in: ICLR (2024)
Keyphrases
- language model
- language modeling
- temporal information
- document retrieval
- n gram
- speech recognition
- video data
- video sequences
- information retrieval
- language modelling
- probabilistic model
- test collection
- statistical language models
- query expansion
- retrieval model
- language models for information retrieval
- video content
- video frames
- vector space model
- statistical language modeling
- document ranking
- video retrieval
- natural language
- multimedia
- ad hoc information retrieval
- smoothing methods
- language model for information retrieval
- query terms
- relevance model
- cross lingual
- context sensitive
- pseudo relevance feedback
- text retrieval
- key frames