On a Strictly Convex IBMModel 1 Andrei Simion
نویسندگان
چکیده
IBM Model 1 is a classical alignment model. Of the first generation word-based SMT models, it was the only such model with a concave objective function. For concave optimization problems like IBM Model 1, we have guarantees on the convergence of optimization algorithms such as Expectation Maximization (EM). However, as was pointed out recently, the objective of IBM Model 1 is not strictly concave and there is quite a bit of alignment quality variance within the optimal solution set. In this work we detail a strictly concave version of IBM Model 1 whose EM algorithm is a simple modification of the original EM algorithm of Model 1 and does not require the tuning of a learning rate or the insertion of an l2 penalty. Moreover, by addressing Model 1’s shortcomings, we achieve AER and F-Measure improvements over the classical Model 1 by over 30%.
منابع مشابه
On A Strictly Convex IBM Model 1
IBM Model 1 is a classical alignment model. Of the first generation word-based SMT models, it was the only such model with a concave objective function. For concave optimization problems like IBM Model 1, we have guarantees on the convergence of optimization algorithms such as Expectation Maximization (EM). However, as was pointed out recently, the objective of IBM Model 1 is not strictly conca...
متن کاملSome Experiments with a Convex IBM Model 2
Using a recent convex formulation of IBM Model 2, we propose a new initialization scheme which has some favorable comparisons to the standard method of initializing IBM Model 2 with IBM Model 1. Additionally, we derive the Viterbi alignment for the convex relaxation of IBM Model 2 and show that it leads to better F-Measure scores than those of IBM Model 2.
متن کاملWeighted composition operators between growth spaces on circular and strictly convex domain
Let $Omega_X$ be a bounded, circular and strictly convex domain of a Banach space $X$ and $mathcal{H}(Omega_X)$ denote the space of all holomorphic functions defined on $Omega_X$. The growth space $mathcal{A}^omega(Omega_X)$ is the space of all $finmathcal{H}(Omega_X)$ for which $$|f(x)|leqslant C omega(r_{Omega_X}(x)),quad xin Omega_X,$$ for some constant $C>0$, whenever $r_{Omega_X}$ is the M...
متن کاملA Convex Alternative to IBM Model 2
The IBM translation models have been hugely influential in statistical machine translation; they are the basis of the alignment models used in modern translation systems. Excluding IBM Model 1, the IBM translation models, and practically all variants proposed in the literature, have relied on the optimization of likelihood functions or similar functions that are non-convex, and hence have multi...
متن کاملA Recurrent Neural Network for Solving Strictly Convex Quadratic Programming Problems
In this paper we present an improved neural network to solve strictly convex quadratic programming(QP) problem. The proposed model is derived based on a piecewise equation correspond to optimality condition of convex (QP) problem and has a lower structure complexity respect to the other existing neural network model for solving such problems. In theoretical aspect, stability and global converge...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2015