We investigate why weights from generative models underperform heuristic estimates in phrasebased machine translation. We first propose a simple generative, phrase-based model and verify that its estimates are inferior to those given by surface statistics. The performance gap stems primarily from the addition of a hidden segmentation variable, which increases the capacity for overfitting during...

Recently, ‘determinization in hindsight’ has enjoyed surprising success in on-line probabilistic planning. This technique evaluates the actions available in the current state by using non-probabilistic planning in deterministic approximations of the original domain. Although the approach has proven itself effective in many challenging domains, it is computationally very expensive. In this paper...

We introduce a new formalism called automata over a timed domain which provides an adequate framework for the determinization of timed systems. In this formalism, determinization w.r.t. timed language is always possible at the cost of changing the timed domain. We give a condition for determinizability of automata over a timed domain without changing the timed domain, which allows us to recover...

The complementation problem for nondeterministic word automata has numerous applications in formal verification. In particular, the language-containment problem, to which many verification problems are reduced, involves complementation. For automata on finite words, which correspond to safety properties, complementation involves determinization. The 2 blow-up that is caused by the subset constr...

This paper discusses determinization of conditional term rewriting systems with oriented constructor rules. We present a rule-based transformation system, which transforms a non-deterministic one into a deterministic one, together with examples of the transformation. We prove that the transformation system is simulation sound and simulation complete. We also prove that the transformation system...

L Introduction Large-scale public systems in today’s information control field require. that seveml conditions should be satisfied. Fmt, a phased-in construction of the system must be possible so that the system can be constructed over a long period of time. Second the system must be flexible and be adapted to meet the changes of social needs. All such systems are assumed to constantly grow, ex...

The aim of this paper is to investigate how subset construction performs on specific families of automata. A new upper bound on the number of states of the subset-automaton is established in the case of homogeneous automata. The complexity of the two basic steps of subset construction, i.e. the computation of deterministic transitions and the set equality tests, is examined depending on whether...

In automated synthesis, we transform a specification into a system that is guaranteed to satisfy the specification. In spite of the rich theory developed for temporal synthesis, little of this theory has been reduced to practice. This is in contrast with model-checking theory, which has led to industrial development and use of formal verification tools. We address this problem here by consideri...

In this paper we regard languages and their acceptors – such as deterministic or weighted automata, transducers, or monoids – as functors from input categories that specify the type of the languages and of the machines to categories that specify the type of outputs. Our results are as follows: a) We provide sufficient conditions on the output category so that minimization of the corresponding a...

We place ourselves in a so-called meet-closed description context; that is a context consisting of a finite nonempty entity set E whose elements are described in a complete meet-semilattice D, by means of a descriptor δ. Then we consider multiway quasi-ultrametric dissimilarities on E, a class of multiway dissimilarities that, with their relative k-balls, extend the fundamental in classificatio...