This introductory text to statistical machine translation (SMT) provides all of the theories and methods needed to build a statistical machine translator, such as Google Language Tools and Babelfish. In general, statistical techniques allow automatic translation systems to be built quickly for any language-pair using only translated texts and generic software. With increasing globalization, statistical machine translation will be central to communication and commerce. Based on courses and tutorials, and classroom-tested globally, it is ideal for instruction or self-study, for advanced undergraduates and graduate students in computer science and/or computational linguistics, and researchers in natural language processing. The companion website provides open-source corpora and tool-kits.

/pdf/statistical-machine-translation-1yr1jtzcjp.pdf

Statistical Machine Translation

Restarting Automata

The growing context-sensitive languages (GCSL) are characterized by a nondeterministic machine model, the so-called shrinking two-pushdown automaton (sTPDA). Then the deterministic version of this automaton (sDTPDA) is shown to characterize the class of generalized Church-Rosser languages (GCRL). Finally, we prove that each growing context-sensitive language is accepted in polynomial time by some one-way auxiliary pushdown automaton with a logarithmic space bound (OW-auxPDA[log, poly]). As a consequence the class of (generalized) Church-Rosser languages and the class of context-free languages are incomparable under set inclusion, which verifies a conjecture of Mc-Naughton et al [MNO88].

Growing context-sensitive languages and Church-Rosser languages

A hybrid system is described which combines the strength of manual rule-writing and statistical learning, obtaining results superior to both methods if applied separately. The combination of a rule-based system and a statistical one is not parallel but serial: the rule-based system performing partial disambiguation with recall close to 100% is applied first, and a trigram HMM tagger runs on its results. An experiment in Czech tagging has been performed with encouraging results.

Serial Combination of Rules and Statistics: A Case Study in Czech Tagging

This paper explains the principles of dependency analysis by reduction and its correspondence to the notions of dependency and dependency tree. The explanation is illustrated by examples from Czech, a language with a relatively high degree of word-order freedom. The paper sums up the basic features of methods of dependency syntax. The method serves as a basis for the verification (and explanation) of the adequacy of formal and computational models of those methods.

/pdf/modeling-syntax-of-free-word-order-languages-dependency-1jcmv1pb0m.pdf

Modeling syntax of free word-order languages: dependency analysis by reduction

On monotonic automata with a restart operation

The restarting automaton is a special kind of the linear bounded automaton with a read/write window of fixed length. It works in cycles, reducing the length of the actual word in each cycle. We study several versions of restarting automata and we show several hierarchies of language classes recognized by them with respect to the size of the read/write-window.

Lookahead hierarchies of restarting automata

Motivated by natural language analysis we introduce restarting automata with rewriting. They are acceptors on the one hand, and (special) regulated rewriting systems on the other hand. The computation of a restarting automaton proceeds in cycles: in each cycle, a bounded substring of the input word is rewritten by a shorter string, and the computation restarts on the arising shorter word.

On Restarting Automata with Rewriting

We introduce a two-dimensional variant of the restarting automaton with window size three-by-three for processing rectangular pictures. In each rewrite step such an automaton can only replace the symbol in the middle position of its window by a symbol that is smaller with respect to a fixed ordering on the tape alphabet. When restricted to one-dimensional inputs (that is, words) the deterministic variant of these ordered restarting automata only accepts regular languages, while the nondeterministic one can accept some languages that are not even context-free. We then concentrate on the deterministic two-dimensional ordered restarting automaton, showing that it is quite expressive as it can simulate the deterministic sgraffito automaton, and we present some closure and non-closure properties for the class of picture languages accepted by these automata.

František Mráz

Papers

Restarting Automata

On monotonic automata with a restart operation

Lookahead hierarchies of restarting automata

On Restarting Automata with Rewriting

Ordered Restarting Automata for Picture Languages