Jason Williams
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Jason Williams has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
Akanksha Bindal
Apple, Cupertino, CA, USA
1 shared publicationAlkesh Patel
Apple, Cupertino, CA, USA
1 shared publicationBlaise Thomson
VocalIQ Ltd., Cambridge, UK
1 shared publicationChristopher Klein
Apple, Cupertino, CA, USA
1 shared publication- Hadas Kotek
Apple and MIT, USA
1 shared publication - Jarek Rossignac
School of Interactive Computing, Georgia Institute of Technology, Atlanta, USA
1 shared publication
Latest Publications
- Open Access
- 6 June 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2270-2274
Generating Natural Questions from Images for Multimodal Assistants
Generating natural, diverse, and meaningful questions from images is an essential task for multimodal assistants as it confirms whether they have understood the object and scene in the images properly. The research in visual question answering (VQA) and visual question generation (VQG) is a great step. However, this research does not capture questions that a visually-abled person would ask multimodal assistants. Recently published datasets such as KB-VQA, FVQA, and OK-VQA try to collect questions that look for external knowledge which makes them appropriate for multimodal assistants. However, they still contain many obvious and common-sense questions that humans would not usually ask a digital assistant. In this paper, we provide a new benchmark dataset that contains questions generated by human annotators keeping in mind what they would ask multimodal digital assistants. Large scale annotations for thousands of images are expensive and time-consuming, so we also present an effective way of automatically generating questions from unseen images. In this paper, we present an approach for generating diverse and meaningful questions that consider image content and metadata of image (e.g., location, associated keyword). We evaluate our approach using standard evaluation metrics such as BLEU, METEOR, ROUGE, and CIDEr to show the relevance of generated questions with human-provided questions. We also measure the diversity of generated questions using generative strength and inventiveness metrics. We report new state-of-the-art results on the public and our datasets.
- 6 June 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 2270-2274
Open AccessGenerating Natural Questions from Images for Multimodal Assistants
Generating natural, diverse, and meaningful questions from images is an essential task for multimodal assistants as it confirms whether they have understood the object and scene in the images properly. The research in visual question answering (VQA) and visual question generation (VQG) is a great step. However, this research does not capture questions that a visually-abled person would ask multimodal assistants. Recently published datasets such as KB-VQA, FVQA, and OK-VQA try to collect questions that look for external knowledge which makes them appropriate for multimodal assistants. However, they still contain many obvious and common-sense questions that humans would not usually ask a digital assistant. In this paper, we provide a new benchmark dataset that contains questions generated by human annotators keeping in mind what they would ask multimodal digital assistants. Large scale annotations for thousands of images are expensive and time-consuming, so we also present an effective way of automatically generating questions from unseen images. In this paper, we present an approach for generating diverse and meaningful questions that consider image content and metadata of image (e.g., location, associated keyword). We evaluate our approach using standard evaluation metrics such as BLEU, METEOR, ROUGE, and CIDEr to show the relevance of generated questions with human-provided questions. We also measure the diversity of generated questions using generative strength and inventiveness metrics. We report new state-of-the-art results on the public and our datasets.
- 1 December 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 324-329
The third Dialog State Tracking Challenge
In spoken dialog systems, dialog state tracking refers to the task of correctly inferring the user's goal at a given turn, given all of the dialog history up to that turn. This task is challenging because of speech recognition and language understanding errors, yet good dialog state tracking is crucial to the performance of spoken dialog systems. This paper presents results from the third Dialog State Tracking Challenge, a research community challenge task based on a corpus of annotated logs of human-computer dialogs, with a blind test set evaluation. The main new feature of this challenge is that it studied the ability of trackers to generalize to new entities - i.e. new slots and values not present in the training data. This challenge received 28 entries from 7 research teams. About half the teams substantially exceeded the performance of a competitive rule-based baseline, illustrating not only the merits of statistical methods for dialog state tracking but also the difficulty of the problem.
- 1 December 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 324-329
The third Dialog State Tracking Challenge
In spoken dialog systems, dialog state tracking refers to the task of correctly inferring the user's goal at a given turn, given all of the dialog history up to that turn. This task is challenging because of speech recognition and language understanding errors, yet good dialog state tracking is crucial to the performance of spoken dialog systems. This paper presents results from the third Dialog State Tracking Challenge, a research community challenge task based on a corpus of annotated logs of human-computer dialogs, with a blind test set evaluation. The main new feature of this challenge is that it studied the ability of trackers to generalize to new entities - i.e. new slots and values not present in the training data. This challenge received 28 entries from 7 research teams. About half the teams substantially exceeded the performance of a competitive rule-based baseline, illustrating not only the merits of statistical methods for dialog state tracking but also the difficulty of the problem.
- 22 September 2008
- Published by International Speech Communication Association
- p. 191-194
Exploiting the ASR n-best by tracking multiple dialog state hypotheses
- 22 September 2008
- Published by International Speech Communication Association
- p. 191-194
Exploiting the ASR n-best by tracking multiple dialog state hypotheses
- 1 October 2007
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Geometry & Applications
- Vol. 17 (05) , 487-503
TIGHTENING: MORPHOLOGICAL SIMPLIFICATION
Given a two- or three-dimensional set S of arbitrary topology and a radius r, we show how to construct an r-tightening of S, which is a set whose boundary has mean curvature with magnitude less than or equal to 1/r and which only differs from S in a morphologically-defined tolerance zone we call the mortar. The mortar consists of the thin or highly curved parts of S and its complement, such as corners, gaps, and small connected components, while the boundary of a tightening consists of components of locally minimal length (in 2D) or area (in 3D) that lie in the mortar. Tightenings are defined independently of shape representation, and it may be possible to find them using a variety of algorithms. We describe how to approximately compute tightenings for two-dimensional sets represented as binary images and for three-dimensional sets represented as triangle meshes using constrained, level-set curvature flow.
- 1 October 2007
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Geometry & Applications
- Vol. 17 (05) , 487-503
TIGHTENING: MORPHOLOGICAL SIMPLIFICATION
Given a two- or three-dimensional set S of arbitrary topology and a radius r, we show how to construct an r-tightening of S, which is a set whose boundary has mean curvature with magnitude less than or equal to 1/r and which only differs from S in a morphologically-defined tolerance zone we call the mortar. The mortar consists of the thin or highly curved parts of S and its complement, such as corners, gaps, and small connected components, while the boundary of a tightening consists of components of locally minimal length (in 2D) or area (in 3D) that lie in the mortar. Tightenings are defined independently of shape representation, and it may be possible to find them using a variety of algorithms. We describe how to approximately compute tightenings for two-dimensional sets represented as binary images and for three-dimensional sets represented as triangle meshes using constrained, level-set curvature flow.
- Open Access
- 12 March 1994
- Published by Association for Computing Machinery (ACM) in ACM SIGCSE Bulletin
- Vol. 26 (1) , 339-343
Understanding the bottom-up SLR parser
This paper describes an application of one of the important abstract concepts taught in a compiler construction course. It demonstrates how the techniques behind the bottom-up SLR parser can be used to perform computer animation. The different phases of the implementation presented are identical to the ones used by the traditional compiler for parsing source codes written in high-level languages. This application can be used either to explain the different phases of the traditional compiler, or as an illustration of the bottom-up SLR parsing techniques applied in a non-traditional environment.
- 12 March 1994
- Published by Association for Computing Machinery (ACM) in ACM SIGCSE Bulletin
- Vol. 26 (1) , 339-343
Open AccessUnderstanding the bottom-up SLR parser
This paper describes an application of one of the important abstract concepts taught in a compiler construction course. It demonstrates how the techniques behind the bottom-up SLR parser can be used to perform computer animation. The different phases of the implementation presented are identical to the ones used by the traditional compiler for parsing source codes written in high-level languages. This application can be used either to explain the different phases of the traditional compiler, or as an illustration of the bottom-up SLR parsing techniques applied in a non-traditional environment.