Each hunspell dictionary comes in two files. The .dic file which is the list of words, and the .aff file which is a list of rules and other options. You will normally only need to add to the .dic file, but you may need to refer to the .aff file for several things. We maintain some additions to the dictionary files which we keep in a separate file ending in .dic_delta. Keeping the original unchanged allows us to more easily pull updated versions from the dictionary maintainers. Chromium actually reads an optimized format which uses the suffix .bdic. This file is generated from the .aff, .dic, and .dic_delta files using the convert_dict tool in the Chromium project. All of these files are checked into deps/third_party/hunspell_dictionaries/ in Subversion, and checked out into src/third_party/hunspell_dictionaries when syncing locally.
The .dic and .aff files are normally in the encoding most appropriate for that language, which is unfortunately not usually UTF-8. You will need to open the files in an editor with that encoding enabled. Most of the Western European languages use Latin1 which should be easy. For other ones, search in the .aff file for the line that begins with "SET" to see which character set it uses.
The .dic_delta files are always in UTF-8! This is confusing, but it allows us to additional flexibility to add foreign words. The .bdic files are always UTF-8 internally, and the convert_dict tool converts things appropriately when it runs.
Our .dic_delta files do not generally have affix rules. These rules tell Hunspell, for example, how to convert a word into its plural or possessive forms. The rest of this document explains how to do this.
The easiest way to get started is to find a word in the .dic file that is like the word you're looking at in the .dic_delta file, and copy the rules.
You can also help by reviewing word lists of words we've identified as popular but not in the dictionary. Please contact brettw AT chromium.org if you are interested.
Affix rules optionally follow each word in the dictionary file. These are separated from the word using a slash. Each file uses one of two formats. It will generally be obvious which one is which. Affix rules tell you what prefixes and suffixes (affixes) can apply to the word. I will describe as much as you need to know. There are additional details at http://sourceforge.net/docman/display_doc.php?docid=29374&group_id=143754
Explicit affix rules
Explicit affix rules look like a string of letters and numbers after each word. This is the most common type. For example, here are some lines from the French file:
Each one of the letters indicates a rule that can possibly apply. You can see that aberré has no rules. This is OK. The other words have a sequence of rules: each rule is identified by a unique letter. Case matters so "F" is different than "f".
Each rule is in the .aff file for that language. The rules come in two flavors: SFX for suffixes, and PFX for prefixes. Each line begins with PFX/SFX and then the rule letter identifier (the ones that follow the word in the dictionary file:
PFX <rule_letter_identifier> <combineable_flag> <number_of_rule_lines_that_follow>
You can normally ignore the combinable flag, it is Y or N depending on whether it can be combined with other rules. Then there are some number of lines (indicated by the <number_of_rule_lines_that_follow>) that list different possibilities for how this rule applies in different situations. It looks like this:
PFX <rule_letter_identifier> <number_of_letters_to_delete> <what_to_add> <when_to_add_it>
SFX B Y 3
SFX B 0 able [^aeiou]
SFX B 0 able ee
SFX B e able [^aeiou]e
If "B" is one of the letters following a word, then this is one of the rules that can apply. There are three possibilities that can happen (because there are three lines). Only one will apply:
- able is added to the end when the end of the word is "not" (indicated by "^") one of the letters in the set (indicated by "[ ]") of letters a, e, i, o, and u. For example, question → questionable
- able is added to the end when the end of the word is "ee". For example, agree → agreeable.
- able is added to the end when the end of the word is not a vowel ("[^aeiou]") followed by an "e". The letter "e" is stripped (the column before able). For example, excite → excitable.
PFX rules are the same, but apply at the beginning of the word instead for prefixes.
Numbered affix rules
Some dictionaries use numbers instead of a list of affix rules. Each number is a unique combination of rules, which is convenient since there are usually not very many unique combinations of rules that can apply. This just means an extra step. For example, the English dictionary has:
You can see that abeyant has no rules, so it is never conjugated. The other rules are 6 and 7. To see what these are, look up in the .aff file. The "AF" lines you which set of rules correspond to each number. For example, the en-US.aff file has a bunch of lines like this:
AF mn # 1
AF 1n # 2
AF pt # 3
AF p # 4
AF ct # 5
AF M # 6
AF MS # 7
AF DGLRS # 8
You can then see that 6 corresponds to the rule sequence "M" and 7 corresponds to the rule sequence "MS". You would look these sequences of rules up just like the explicit affix rules discussed above. So you look up in the PFX and SFX lines and find the rule for "M" is:
SFX M Y 1
SFX M 0 's .
Which means to make it possessve. For example, Abey → Abbey's.