The ABCs of Morphology: False Positives in Terminology Management

To continue Nimdzi’s series of Finger Food posts on terminology management (including this one and this one), let’s talk about morphology. The aim of morphology includes the definition of a word as a language object and a description of its internal structure.

Morphology is all about how words are formed: their roots, compounds, declensions, conjugations, and relations with other words. To make the word analysis, morphology looks into the structure of words and their parts (e.g., stems, prefixes, suffixes).

Source: Queen's University

Some languages are more "highly inflected," meaning the word form may change depending on grammatical case, gender, and number. Examples of highly inflected languages include Slavic languages, Latin and Romance languages and certain Germanic languages. In translation, morphology becomes crucial for terminology control. A simple word form that can be used as a noun, a verb, and part of a compound noun in English can translate into multiple word forms in other languages. 

Take, for example, the English word “database.” You store data in a database, export from a database, and you have multiple databases. When you translate these word forms (database/databases) into an inflected language such as Russian, you may have up to 12 word forms in a single text.

When you run a terminology check (with a tool comparing a translation with nominative glossary entries), you may get false positives in 10 of these cases. This escalates into wasted hours of running through terminology reports full of false positive errors that could have been avoided.

As the strategy of morphological control is different for different languages, some language technology providers argue that the morphology-related functionality would be better supported in Translation Management Systems and CAT-tools. But others treat morphology with due respect. For example, they may use specifically developed morphological engines. There are also tools like Term Morphology Editor which helps during the preparation of termbases for efficient term recognition.

Some examples of dealing with morphology in terminology management:

• In Kaleidoskope’s quickTerm the search for terminology can include stemming and decomposition: users can search in several languages simultaneously and benefit from the morphological stemming of checkTerm.
• Lingo24's termfinder has statistical forms for terms.
• memoQ's QTerm features prefix-based term matching.  
• When Terminotix’s LogiTerm searches a term, it searches for many different forms (plural/singular, gender, tense, etc).
• Tilde uses a more sophisticated approach of dividing languages into groups and supporting the morphology of each language in a distinct way.

All these methods help control terminology more efficiently. When you run a terminology check, you won’t get as many false positives as you would in tools where morphology is not supported. This helps save time and effort on terminology maintenance.