Decades of research and development in Natural Language Processing and Automated Linguistic Analysis have resulted in real-world tools that enable reuse of approved phrases and sentences—without adding to writer or editor workload. Forward-thinking global companies in a variety of industries have discovered that this capability has far-reaching ramifications not just in terms of quality, but in terms of cost and time-to-global-market as well.

The Problem: How many ways can you say “the end date must be after the start date”? In one large, sophisticated software system, we found 51 variants of that simple statement in the UI messages that appeared on various screens. In the documentation for a complex mechanical system, we found an amazing 129 variants of an even simpler sentence: “turn the switch to the run position.” We’ve analyzed dozens of large document sets, translation memories, websites, and knowledgebases and found that in any given corpus, anywhere from 10% to 15% of the content is expressed an average of 3.5 different ways.

The Impact: So what—who cares? Variety is the spice of life, isn’’t it?  Maybe … but that spice can be quite expensive if you’re writing to a global audience.  In a corpus of a million sentences, if you consistently used a single, approved sentence instead of 3.5 variations on the theme, you could avoid translating 250,000 sentences. At an average cost of $2 per sentence. Per language. And, reduce your time-to-global-market by 10% to 15% or more while you’re at it.

Even if you’re not translating, there are substantial benefits to be had. Chances are your audience consists of a significant number of non-native speakers whose vocabularies and knowledge of English sentence construction isn’t as extensive as the average professional writer. Or, if your product goes to average American consumers—80% of whom read at a tenth grade level or less. Can you spell “litigation risk avoidance”?

The Solution: Natural Language Processing and Linguistic Analysis research and development has been in vogue practically since the introduction of the first computer.  The concept of having a machine analyze text like a human being has been appealing, but very difficult to achieve, and even more difficult to apply in a practical, real-world environment. But things started to change at the turn of the century in the NLP Lab at the German Research Institute for Artificial Intelligence (DFKI – Saarbrucken). A powerful Natural Language Processor was integrated with sophisticated Linguistic Analysis algorithms, and presented in a common word processing environment using a spellcheck-like User Interface. The system was spun off and productized in 2002, and further refined and enhanced over the past six years.

This session will describe and demonstrate how linguistic meaning-based matching enables effective phrase- and sentence-level reuse, including:

• Terminology and phrase harvesting and validation
• Reusable sentence harvesting (micro-clustering) and validation
• The writer/editor user-experience
• Auto-generated metrics and reporting

The session will conclude with examples of real-world applications and results, and an interactive question and answer free-for-all.