JGloss can use two different parsers for the automatic annotation of Japanese text, the Kanji parser and the ChaSen parser. Select the parser by clicking the respective radio button. The ChaSen parser will only be available if the chasen program is installed (see below).
If the annotate first occurrence option is selected, each word in a document is only annotated the first time it appears. This decreases the RAM usage and the time it takes to display the document. The Guess paragraph breaks option controls how line breaks in the imported document are converted to paragraph breaks. If the option is selected, JGloss tries to determine if a line break in the imported document signifies the end of a paragraph or if it is used for formatting reasons only. In the second case, it will be ignored.
Some documents you can find on the internet already have reading annotations added to kanji words in the form of some hiragana enclosed in brackets after the kanji word. The parsers can generate reading annotation entries for these words. You can select the brackets used in the document for reading annotations with the Brackets used... box. If the document contains no reading annotations, you can select none or simply ignore this setting.
The Kanji parser is built into JGloss. A simple heuristic is used for choosing words to annotate: for a sequence of katakana characters, the whole sequence is treated as one word and looked up. For a sequence of kanji characters followed by hiragana characters, the algorithm first looks for possible inflected forms in the hiragana string and will try to find words that consist of the kanji word and the dictionary form of the inflected forms that appear in the hiragana string. If no match is found, only the kanji part is looked up. If still no match is found in any of the dictionaries, prefixes of the kanji word will be tried and if this leads to a match the process will be repeated with the remainder. A consequence of this method is that hiragana words will never be annotated automatically even if they are in the dictionaries.
The ChaSen parser uses the ChaSen morphological analysis program to decompose Japanese
text in words and to derive the base form of inflected words.
It is slower than the Kanji parser, but will annotate hiragana words as
well as kanji and katakana words. It also does a better job of deinflecting verbs and
adjectives. You can download ChaSen
from
the ChaSen homepage
. On Ubuntu Linux, you can simply install the package chasen
(sudo apt-get install chasen).
After installation, you have to set the path to the chasen executable in the
preferences dialog
. It usually is
/usr/bin/chasen
under Unix or
c:\Program Files\chasen21\chasen.exe
under Windows.
The ChaSen program is used to generate a list of words with their reading and base form from the parsed text. The words will be looked up in the dictionaries, and if an entry is found, an annotation will be generated. If no dictionary entry is found and the word is not inflected, kanji substrings will be tried. A reading annotation with the reading output by ChaSen is also added if the reading returned by ChaSen is different from the first reading found in the dictionaries. Since the ChaSen program uses its own set of dictionaries to detect words, it might not recognize words which are found in the dictionaries used by JGloss but not in the ChaSen dictionaries.