Papers by Pingxuan Huang
FIBER: Fill-in-the-Blanks as a Challenging Video Understanding Evaluation Framework (2022.acl-long)
Copied to clipboard
Santiago Castro, Ruoyao Wang, Pingxuan Huang, Ian Stewart, Oana Ignat, Nan Liu, Jonathan Stroud, Rada Mihalcea
| Challenge: | Existing video understanding evaluation frameworks that use fill-in-the-blanks do not reflect real-world tasks. |
| Approach: | They propose to use fill-in-the-blanks as a video understanding evaluation framework and introduce a novel dataset that collects multiple perspectives on the same video. |
| Outcome: | The proposed framework does not share the weaknesses of the current state-of-the-art language-informed video understanding tasks, namely: (1) video question answering using multiple-choice questions, where models perform relatively well because they exploit linguistic biases in the task formulation; (2) video captioning, which relies on an open-ended evaluation framework that is often inaccurate because system answers may be perceived as incorrect if they differ in form from the ground truth. |
FGraDA: A Dataset and Benchmark for Fine-Grained Domain Adaptation in Machine Translation (2022.lrec-1)
Copied to clipboard
Wenhao Zhu, Shujian Huang, Tong Pu, Pingxuan Huang, Xu Zhang, Jian Yu, Wei Chen, Yanfeng Wang, Jiajun Chen
| Challenge: | Recent research on domain adaptation neglects diversity in translation within a domain . current research on NMT models considers very broad target domains . |
| Approach: | They propose a fine-grained domain adaptation task for autonomous vehicles, AI education, real-time networks, and smart phone. |
| Outcome: | The proposed task is compared with a dataset of Chinese-English translation tasks for four sub-domains of information technology: autonomous vehicles, AI education, real-time networks, and smart phone. |
In-the-Wild Video Question Answering (2022.coling-1)
Copied to clipboard
| Challenge: | Existing video understanding datasets focus on human interactions with little attention being paid to the “in the wild” settings. |
| Approach: | They propose a video understanding dataset of videos recorded outdoors . they propose identifying visual support for a given question and answer . |
| Outcome: | The proposed dataset examines the ability of models to understand videos, including video question answering, video captioning, and fill-inthe-blank tasks. |