-
A Simple and Effective Model for Answering Multi-span Questions
Authors:
Elad Segal,
Avia Efrat,
Mor Shoham,
Amir Globerson,
Jonathan Berant
Abstract:
Models for reading comprehension (RC) commonly restrict their output space to the set of all single contiguous spans from the input, in order to alleviate the learning problem and avoid the need for a model that generates text explicitly. However, forcing an answer to be a single span can be restrictive, and some recent datasets also include multi-span questions, i.e., questions whose answer is a…
▽ More
Models for reading comprehension (RC) commonly restrict their output space to the set of all single contiguous spans from the input, in order to alleviate the learning problem and avoid the need for a model that generates text explicitly. However, forcing an answer to be a single span can be restrictive, and some recent datasets also include multi-span questions, i.e., questions whose answer is a set of non-contiguous spans in the text. Naturally, models that return single spans cannot answer these questions. In this work, we propose a simple architecture for answering multi-span questions by casting the task as a sequence tagging problem, namely, predicting for each input token whether it should be part of the output or not. Our model substantially improves performance on span extraction questions from DROP and Quoref by 9.9 and 5.5 EM points respectively.
△ Less
Submitted 5 October, 2020; v1 submitted 29 September, 2019;
originally announced September 2019.
-
iMAP Beamforming for High Quality High Frame Rate Imaging
Authors:
Tanya Chernyakova,
Dan Cohen,
Meged Shoham,
Yonina C. Eldar
Abstract:
We present a statistical interpretation of beamforming to overcome limitations of standard delay-and-sum (DAS) processing. Both the interference and the signal of interest are viewed as random variables and the distribution of the signal of interest is exploited to maximize the a-posteriori distribution of the aperture signals. In this formulation the beamformer output is a maximum-a-posteriori (M…
▽ More
We present a statistical interpretation of beamforming to overcome limitations of standard delay-and-sum (DAS) processing. Both the interference and the signal of interest are viewed as random variables and the distribution of the signal of interest is exploited to maximize the a-posteriori distribution of the aperture signals. In this formulation the beamformer output is a maximum-a-posteriori (MAP) estimator of the signal of interest. We provide a closed form expression for the MAP beamformer and estimate the unknown distribution parameters from the available aperture data using an empirical Bayes approach. We propose a simple scheme that iterates between estimation of distribution parameters and computation of the MAP estimator of the signal of interest, leading to an iterative MAP (iMAP) beamformer. This results in a significant improvement of the contrast compared to DAS without severe increase in computational complexity or need for fine-tuning of parameters. By implementing iMAP on both simulated and experimental data, we show that only 13 transmissions are required to obtain contrast comparable to DAS with 75 plane-waves. The proposed method is compared to other interference suppression techniques such as coherence factor and scaled Wiener processing and shows improved contrast and better preserved speckle pattern.
△ Less
Submitted 9 June, 2018;
originally announced June 2018.
-
SINGULAB - A Graphical user Interface for the Singularity Analysis of Parallel Robots based on Grassmann-Cayley Algebra
Authors:
Patricia Ben-Horin,
Moshe Shoham,
Stéphane Caro,
Damien Chablat,
Philippe Wenger
Abstract:
This paper presents SinguLab, a graphical user interface for the singularity analysis of parallel robots. The algorithm is based on Grassmann-Cayley algebra. The proposed tool is interactive and introduces the designer to the singularity analysis performed by this method, showing all the stages along the procedure and eventually showing the solution algebraically and graphically, allowing as wel…
▽ More
This paper presents SinguLab, a graphical user interface for the singularity analysis of parallel robots. The algorithm is based on Grassmann-Cayley algebra. The proposed tool is interactive and introduces the designer to the singularity analysis performed by this method, showing all the stages along the procedure and eventually showing the solution algebraically and graphically, allowing as well the singularity verification of different robot poses.
△ Less
Submitted 18 September, 2008;
originally announced September 2008.
