Evolution of the Logos grammar: system design and development methodology

29203

Authors

nan

Abstract

The history of MT shows how the shortcomings of one approach become the catalyst for a new approach. New systems are designed in reaction to old systems' weaknesses. Yet some old systems have survived and are being used increasingly in the "real world" of translation. One of these systems is Logos. While keeping its fundamental approach, Logos has evolved through its own design history so that today's grammar looks quite different from the earliest model. Our paper describes how the Logos system evolved in response to requirements of development, both linguistic and methodological. We describe four "generations" of the Logos English-source system, each situated in its technological context, with the most recent generation allowing for completely separate development of analysis and transfer components. The system has become larger and more complex over the years, but at the same time more modular. Combined with the advances in computer technology, this modularity makes it easier to develop and improve grammars. Linguists can now work independently, both on-site and remotely, on different source and target languages. 1.

Keywords

nan

Full Content

EVOLUTION OF THE LOGOS GRAMMAR: SYSTEM DESIGN AND DEVELOPMENT METHODOLOGY Patricia Schmid and Claudia Gdaniec Logos Corporation 200 Valley Rd., Suite 400 Mt. Arlington, NJ 07856 {pschmid,cgdaniec}@logos-usa.com Abstract The history of MT shows how the shortcomings of one approach become the catalyst for a new approach. New systems are designed in reaction to old systems' weaknesses. Yet some old systems have survived and are being used increasingly in the "real world" of translation. One of these systems is Logos. While keeping its fundamental approach, Logos has evolved through its own design history so that today's grammar looks quite different from the earliest model. Our paper describes how the Logos system evolved in response to requirements of development, both linguistic and methodological. We describe four "generations" of the Logos English-source system, each situated in its technological context, with the most recent generation allowing for completely separate development of analysis and transfer components. The system has become larger and more complex over the years, but at the same time more modular. Combined with the advances in computer technology, this modularity makes it easier to develop and improve grammars. Linguists can now work independently, both on-site and remotely, on different source and target languages. 1. Introduction The history of MT shows how the shortcomings of one approach become the catalyst for a new approach. The MT field expands as new systems are designed in reaction to old systems' weaknesses. Yet even as the field continues to grow and change, there are some old systems which have not only survived, but are being used increasingly and successfully in the "real world" of translation. One of these commercial systems is Logos. Over the years Logos has evolved through its own design history so that today's grammar looks quite different from the earliest model. Logos has never fit exactly into one of the traditional MT classification schemes. At one time, a Logos system was designed for a single language pair, on which grounds Warwick (1987:24) groups Logos among the direct systems, but the present system uses a single source analysis as the basis for the generation of multiple targets.1 Furthermore, unlike a direct system, Logos has always maintained a separation of linguistic data and processing algorithms. Logos also creates a full parse of the sentence, which is characteristic of a transfer system. The SAL representation 1 Currently Logos has an English-source system with German, French, Spanish, and Italian targets, as well as a German-source system with English, French, and Italian targets. 86 into which source text is translated is a type of interlingua, which further complicates classification. In addition, the Logos rulebases have been compared to a neural net because of the way they interact with the linguistic input. (Scott 1992). Thus it cannot be said that Logos has evolved from one MT category into another. It continues to be a hybrid system. Nevertheless, the requirements of linguistic development have over the years led to new stages in the evolution of Logos, both in system design …

Content truncated to 500 words

Labeled Text

The history of MT shows how the shortcomings of one approach become the catalyst for a new approach. New systems are designed in reaction to old systems' weaknesses. Yet some old systems have survived and are being used increasingly in the "real world" of translation. One of these systems is Logos. While keeping its fundamental approach, Logos has evolved through its own design history so that today's grammar looks quite different from the earliest model. Our paper describes how the Logos system evolved in response to requirements of development, both linguistic and methodological. We describe four "generations" of the Logos English-source system, each situated in its technological context, with the most recent generation allowing for completely separate development of analysis and transfer components. The system has become larger and more complex over the years, but at the same time more modular. Combined with the advances in computer technology, this modularity makes it easier to develop and improve grammars. Linguists can now work independently, both on-site and remotely, on different source and target languages. 1.

Publication Details

Paper ID:
29203
Year:
1996
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan