The InfoCycle^TM: Information Feedback for
Interactive Multilingual Document Production

The Information Feedback Loop, frequently referred to as the Information Cycle (Infocycle^TM) represents a continuous process involving never-ending input, processing, storage, retrieval, manipulation, heuristics, variable output, and data exchange. Nevertheless, for illustrative purposes, we can mark a starting point for any given piece of content by defining an initial process function: monolingual text authoring.

Functions: Content units are subject to a number of functions over the course of their life within an information management system:

1. Content Units (CUs) are authored using some sort of authoring tool .
2. Authoring tools ideally mark up texts as they are created, or markup can be added later with automatic markup utilities.
3. Some texts will be created using single-source authoring tools together with controlled language.
4. Texts in multilingual environments can be subject to human translation or the creation of parallel, multilingual documents written originally in the various target languages (multilingual technical writing).
5. Translated texts may be created using Translation Memory programs, which store matched Translation Units (TUs) for later reuse in similar documents. TUs can also be extracted from existing parallel translations using text alignment tools.
6. Some texts, particularly those created in controlled language environments, are translated using Machine Translation (MT).
7. All texts are stored in various types of corpora, from which information can be retrieved using the appropriate retrieval strategies (taxonomies, ontologies, and topic maps).
8. Information is constantly subject to interchange and handoff within the complex system of functions.

Resources: Functions rely on various language resources throughout the course of the information cycle. The lexical resources processed throughout the feedback loop and stored in the various repositories making up the "information warehouse" are comprised of content units that can be characterized by different text chunk sizes or degrees of granularity, as reflected in the Content Cline . Resources can include:

• Monolingual and multilingual text corpora, including parallel (aligned translated) texts and comparable texts in similar subject areas and text types
• Single-source authoring corpora containing both monolingual and multilingual text components
• Translation memory databases containing translation units
• Terminology databases and data banks
• MT and other NLP (natural language processing) lexica
• Controlled language vocabularies
• Taxonomies, ontologies, and topic maps

Standards: Numerous standards have been created for the purpose of governing the interchange of data within the feedback loop, to provide internal markup for documents, and to ensure proper handoff of data from one application to another. Romary provides a brief overview of some of these standards, and more detailed information is available in a chart illustrating Standards for Content Creation and Globalization.

Numerous industry experts have used the notion of an information cycle or some sort of cyclic document and content handling model, linking the functions described in the model to various information warehouses from which intellectual capital can be leveraged to produce new documents. Click the following thumbnails to view several similar models.

Traditional Authoring & Librarianship

Traditional, linear hardcopy authoring environments are characterized by a logical beginning (the creation of the document) and an end (the delivery of the document to the end user). Opportunities for feedback between creators and users are limited. If present, the translation phase can be interposed at any stage following initial document authoring.

Authoring in the Context of Information Management

This University of St. Cloud (MN) image describes the document production chain as one involving dynamic interaction among all stakeholders and at all levels during the document life cycle. In this model, users and access providers (frequently librarians and information brokers) can provide feedback to authors and publishers in a way that was not possible in the era of strictly hardcopy publishing. Again, translation can be introduced at almost any stage in this process. In enterprise-oriented situations, feedback comes ideally from all levels of the system, e.g., designers, engineers, programmers, authors, translators and localizers, sales, testers, and end users

The InfoCycle in Translation Environments

In 1998, Berlitz/équipe's Rose Lockwood described the "InfoCycle" in a Language International article and in a presentation at Kent State University. Key to this representation is the presence of "Information Repositories" that serve as resources for the various functions that are performed throughout the cycle. Note the significant role she gives to terminology management.

Action-Oriented Information Feedback and Mitre Corp.

This information cycle from Mitre Corp , ca. 2000, goes beyond the document production cycle to diagram a field of action involving the interoperability of different kinds of information in multilingual environments. The goal here is to provide information for coordinated action as well as the further communication.

The InfoCycle^TM for Content Localization

Before its purchase by SDL, Alpnet trademarked the Infocycle^TM coinage, although, as evidenced by Rose Lockwood's earlier use of the term, it was not necessarily original to Alpnet. This illustration from around 2000 introduces the notion of "content creation" and "content management and publishing" based on multilingual knowledge warehousing, concepts that came into vogue around that time. Consequently, the maturation of the cycle metaphor is accompanied by the notion of content as opposed to texts per se, and the notion of a "document" has definitely moved from hardcopy texts to Web content and various modalities involving electronic text generation.

A Generic Reference Model for Web site Globalization

i18n.ca (Pierre Cadieux) applies the cyclic image to Globalization and Web content mangement systems in a detailed article on the Generic Reference Model. Again, language resources and language resource maintenance occupy a central position. This image reflects the growing emphasis on the Globalization aspect of the GILT paradigm (Globalization, Internationalization, Localization, and Translation), underscoring the need for total enterprise solutions in order to fully exploit the intellectual capital residing in an enterprise's language resources.

Text Content Units Viewed as a Cline

Content units can be viewed at a variety of levels along a cline gradient, ranging in relative size from individual words and terms to collocations and phrases, co-occurrences, controlled language units, translation memory units (called TUs or translation segments), example-based machine translation units, programming strings, and full chunks of standard text or single-source authoring components. They may exist as monolingual units with roughly parallel equivalents in other languages or truly parallel texts derived from, or intended for the creation of, aligned texts in two or more languages. Content units can be generated in several ways:

• Individual terminology or content units can be "manually" added to a knowledge repository
• Matched bilingual or multilingual units can be created during the generation of new original text units and their translation
• Matched content units can be created automatically using text mining algorithms and system heuristics.

The various levels represented on the cline relate to functions and resources in the Information Management Cycle.

Standards for Language Technology

Laurent Romary's presentation of language standards positions them with respect to other important information handling and retrieval standards, such as the Dublin core, Eagles/Isle & TEI standards, etc. As noted above, see Standards for Content Creation and Globalization for more detailed information on language industry standards.

Additional References:

Wright, Sue Ellen. Trends in Language Engineering

Wright, Sue Ellen. Convergence Technologies

Return to top