ShapeWorld - A new test methodology for multimodal language understanding
arXiv:1704.04517
Abstract
We introduce a novel framework for evaluating multimodal deep learning models with respect to their language understanding and generalization abilities. In this approach, artificial data is automatically generated according to the experimenter's specifications. The content of the data, both during training and evaluation, can be controlled in detail, which enables tasks to be created that require true generalization abilities, in particular the combination of previously introduced concepts in novel ways. We demonstrate the potential of our methodology by evaluating various visual question answering models on four different tasks, and show how our framework gives us detailed insights into their capabilities and limitations. By open-sourcing our framework, we hope to stimulate progress in the field of multimodal language understanding.
References in corpus (6)
- Hierarchical Question-Image Co-Attention for Visual Question Answering
- Multimodal Compact Bilinear Pooling for Visual Question Answering and Visual Grounding
- Discovering objects and their relations from entangled scene representations
- RNN Approaches to Text Normalization: A Challenge
- Incorporating Discrete Translation Lexicons into Neural Machine Translation
- Revisiting Visual Question Answering Baselines
Cited by in corpus (16)
- Learning Transferable Visual Models From Natural Language Supervision
- Ground Truth Evaluation of Neural Network Explanations with CLEVR-XAI
- A Benchmark for Systematic Generalization in Grounded Language Understanding
- CLOSURE: Assessing Systematic Generalization of CLEVR Models
- Systematic Generalization: What Is Required and Can It Be Learned?
- Going Beneath the Surface: Evaluating Image Captioning for Grammaticality, Truthfulness and Diversity
- Emergent Communication of Generalizations
- Multimodal Generative Models for Compositional Representation Learning
- KANDINSKYPatterns -- An experimental exploration environment for Pattern Analysis and Machine Intelligence
- SLASH: Embracing Probabilistic Circuits into Neural Answer Set Programming
- Learning to refer informatively by amortizing pragmatic reasoning
- What is needed for simple spatial language capabilities in VQA?
- A new dataset and model for learning to understand navigational instructions
- Open-domain clarification question generation without question examples
- Calibrate your listeners! Robust communication-based training for pragmatic speakers
- TraVLR: Now You See It, Now You Don't! A Bimodal Dataset for Evaluating Visio-Linguistic Reasoning