Why a PDF is not enough
Publishing a dictionary as a PDF is quick, but readers cannot search across entries, find a word from an inflected form or follow relations. Search engines cannot show thousands of entries inside one large file either. When every entry has its own page, the dictionary is found entry by entry.
Preparing the data
In Word files, parts such as the headword, part of speech, sense and example are usually set apart with bold, italics or specific markers. These regular patterns are the basis for parsing entries into fields.
- Extracting formatting and marker patterns
- Mapping Excel columns to entry fields
- Scanning and correcting OCR output for printed dictionaries
- Collecting abbreviation and source lists
Trial on a sample section
Before the full import, a sample section (for example one letter) is imported; we review the result together and refine the rules, so the same error is not repeated across thousands of entries.
Review and clean-up
- Finding duplicate and incomplete entries
- Merging inconsistent abbreviations and source spellings
- Extracting person, source and date information from the text
- Turning "see also", synonym and antonym references into links
Standard formats
So that the dictionary does not live only on one website, it matters that the data can also be exported in standard formats. TEI Lex-0 is a TEI-based XML format developed for dictionary data; SKOS is used for term and concept lists, and CSV and JSON for general data exchange.
After publishing
The imported dictionary keeps living in an editor panel: new entries are added, reader feedback comes in, and searches with no result become hints for new entries.