Independent Transfer Using Graph Unification

نویسندگان

  • Lauri Carlson
  • Maria Vilkuna
چکیده

We present a MT system that applies graph unification in transfer from English to Finnish. The work described below is an outgrowth of a multilingual MT project initiated by the IBM in 1987 with the aim of studying multilingual translation using a common English language parser. The transfer system presented here is independent of the parsing and generation modules. Any source language parser can be used whose output can be expressed in a directed graph form. The transfer system is responsible for generating target language phrase structure. Target language word order and morphology are left to the generation modules. The transfer system is lexically based. Transfer rules, presented in the form of bilingual graphs, are declarative statements of symmetric transfer relationships between words, phrases or constructions in the two intertranslatable languages. Transfer is structure driven in that the transfer algorithm traverses the source language graph, nondeterministically trying to apply the relevant transfer rules in the lexicon. Each successful transfer yields a bilingual graph, whose target hmguage half is extracted and subjected to linearization and morphological generation. The main f(ycus of attention in our project is the development of the lexicon subsystem. The lexicon system consisk~ of separate transfer and monolingual lexicons and a common lexicon of language independent definitions.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Quasi-Destructive Graph Unification

Graph unification is the most expensive part of unification-based grammar parsing. It often takes over 90% of the total parsing time of a sentence. We focus on two speed-up elements in the design of unification algorithms: 1) elimination of excessive copying by only copying successful unifications, 2) Finding unification failures as soon as possible. We have developed a scheme to attain these t...

متن کامل

Unifying local and non-local signal processing with graph CNNs

This paper deals with the unification of local and non-local signal processing on graphs within a single convolutional neural network (CNN) framework. Building upon recent works on graph CNNs, we propose to use convolutional layers that take as inputs two variables, a signal and a graph, allowing the network to adapt to changes in the graph structure. In this article, we explain how this framew...

متن کامل

Rank 2 Types for Term Graph Rewriting Extended

We define a notion of type assignment with polymorphic intersection types of rank 2 for a term graph rewriting language that expresses sharing and cycles. We show that type assignment is decidable through defining, using the extended notion of unification from [5], a notion of principal pair which generalizes ml’s principal type property.

متن کامل

Progress Report: Term Dags Using Stobjs

We explore in this paper the use of efficient data structures to implement operations on first-order terms, that can be formally verified. Specifically, we present the status of our work on defining and verifying a unification algorithm acting on terms represented as directed acyclic graphs (dags). This implementation is done using single threaded objects (stobjs) to store a dag representing th...

متن کامل

On ‎c‎omputing the general Narumi-Katayama index of some ‎graphs

‎The Narumi-Katayama index was the first topological index defined‎ ‎by the product of some graph theoretical quantities‎. ‎Let $G$ be a ‎simple graph with vertex set $V = {v_1,ldots‎, ‎v_n }$ and $d(v)$ be‎ ‎the degree of vertex $v$ in the graph $G$‎. ‎The Narumi-Katayama ‎index is defined as $NK(G) = prod_{vin V}d(v)$‎. ‎In this paper,‎ ‎the Narumi-Katayama index is generalized using a $n$-ve...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1990