Documentation Center

Recommendations for dictionary entries

Guidelines to create proper entries in a dictionary.

Recommendations and guidelines

The following guidelines help you create dictionary terms correctly and accurately.

  • One-word entries should be avoided, as they may reduce the overall accuracy in the translation. It is recommended to include only terms consisting of two or more words. This will ensure more accurate rendering of their meaning.
  • Terms should have a high frequency in the context where they are used.
  • Terms should be unambiguous: they must have one clear meaning per context and they must not require additional interpretation.
  • Terms that change for gender or number in the source language must be added as separate entries if fluent terminology is not used.

    For example, Arzt and Ärztin are the German masculine and feminine words for doctor in English. In a German>English dictionary, they should both be entered if fluent terminology is not used.

  • Words that can be spelled in different ways in the source language must be added as separate entries.
  • Terms consisting of common nouns and verbs must be avoided.
  • When a dictionary is applied to a translation, longer terms have priority over shorter terms that may be included in a longer one.

    For example, if a dictionary includes the English terms no credit, no credit check, and no credit check auto loan, the term no credit check auto loan has priority over the other two. Likewise, the term no credit check has priority over no credit.

Reliability guidelines

Start by translating the source content without using dictionaries
Start by translating your content without using dictionaries. This will provide you with a clear view of the initial translation quality.
Analyze the initial translation and identify recurrent issues
After the initial translation, check the output to identify recurrent issues. Most of the times, the issues connected with terminology can be fixed with the use of dictionaries.
You can see below some fo the most common issues connected with terminology.
  • Untranslated words: Words that are still displayed in the source language after machine translation. This affects the meaning of the text because parts of the message are compromised.
  • Mistranslated words: This typically corresponds to cases where a domain-specific term in the source language is translated using a generic word in the target language, or a specific word that belongs to a different domain.
  • Misspelled words: Thought this is not frequent, if you notice that the translation of a specific word is usually misspelled, you may consider adding it to a dictionary. However, consider the following aspects before adding terms to a dictionary:
    • Translation quality is higher if the terms included in a dictionary have at least 2, preferably 3, words.
    • The terms should have a high frequency in the content that will be translated. The higher the frequency, the bigger the impact on the translation quality.
    • The terms should have a clear meaning in the context where they are used. For example, a term with only one meaning and grammatical category is a better choice for a dictionary than a term whose meaning is context-dependent.
Assess the impact of the dictionaries
After translating your content using a dictionary, you should analyze the machine translation output again. Focus on the impact that the terms included in the dictionary have had on the translation. The following situations might occur.
  • A previously wrongly translated term is now predominantly translated correctly. This is the intended effect.
  • A previously wrongly translated term has not changed. Verify if there is a typo or if the term is inflected in the new source.
  • A previously correctly translated term is now translated incorrectly. Probably one of your terms is applied where you did not expect it. Consider removing the term from the dictionary.