Abstract:As one of the fundamental video tasks in computer vision, Open-Vocabulary Action Recognition (OVAR) recently gains increasing attention, with the development of vision-language pre-trainings. To enable generalization of arbitrary classes, existing methods treat class labels as text descriptions, then formulate OVAR as evaluating embedding similarity between visual samples and textual classes. However, one crucial issue is completely ignored: the class descriptions given by users may be noisy, e.g., misspellings and typos, limiting the real-world practicality of vanilla OVAR. To fill the research gap, this paper pioneers to evaluate existing methods by simulating multi-level noises of various types, and reveals their poor robustness. To tackle the noisy OVAR task, we further propose one novel DENOISER framework, covering two parts: generation and discrimination. Concretely, the generative part denoises noisy class-text names via one decoding process, i.e., propose text candidates, then utilize inter-modal and intra-modal information to vote for the best. At the discriminative part, we use vanilla OVAR models to assign visual samples to class-text names, thus obtaining more semantics. For optimization, we alternately iterate between generative and discriminative parts for progressive refinements. The denoised text classes help OVAR models classify visual samples more accurately; in return, classified visual samples help better denoising. On three datasets, we carry out extensive experiments to show our superior robustness, and thorough ablations to dissect the effectiveness of each component.

Denoising Table-Text Retrieval for Open-Domain Question Answering

Enhancing Open-Domain Table Question Answering via Syntax- and Structure-aware Dense Retrieval

KET-QA: A Dataset for Knowledge Enhanced Table Question Answering

Evidentiality-aware Retrieval for Overcoming Abstractiveness in Open-Domain Question Answering

End-to-End Training of Neural Retrievers for Open-Domain Question Answering

Answerability in Retrieval-Augmented Open-Domain Question Answering

Retrieval Augmented Via Execution Guidance in Open-domain Table QA

Denoise while Aggregating: Collaborative Learning in Open-Domain Question Answering

Denoising Distantly Supervised Open-Domain Question Answering

SS-BERT: A Semantic Information Selecting Approach for Open-Domain Question Answering

An Information Retrieval-Based Approach to Table-Based Question Answering.

Novel Knowledge-Based System with Relation Detection and Textual Evidence for Question Answering Research.

Dense Hierarchical Retrieval for Open-Domain Question Answering

DENOISER: Rethinking the Robustness for Open-Vocabulary Action Recognition

TACR: A Table-alignment-based Cell-selection and Reasoning Model for Hybrid Question-Answering

Synthetic Target Domain Supervision for Open Retrieval QA

Dual Reader-Parser on Hybrid Textual and Tabular Evidence for Open Domain Question Answering

CABINET: Content Relevance based Noise Reduction for Table Question Answering

L2R-QA: An Open-Domain Question Answering Framework

TACR: A Table Alignment-based Cell Selection Method for HybridQA.

A Survey on Table Question Answering: Recent Advances