Commits

Author Commit Message Labels Comments Date
lars
[svn] Added a script which sorts images into subdirectories.
lars
[svn] Added a script which sorts image files into subdirectories so as to lessen the filesystem load.
lars
[svn] Cleaned up the log output slightly by separating the kana log from the pinyin log.
lars
[svn] Added the ability to dump and load unigram models from flat files.
lars
[svn] Added output of the number of rows and the branching factor of each step of database creation. Also cleaned up the explainFailedQuery script a bit.
lars
[svn] Greatly simplified the reading model code by using the naieve reading pool provided by kanjidic.
lars
[svn] Added a simple reading pool property for kanjidic entries.
lars
[svn] Added javascript form focus, so that each time the page is loaded, the cursor goes into the search box immediately.
lars
[svn] Added tests for the new kana.canonicalForms method.
lars
[svn] Two significant changes. Firstly, included okurigana readings as valid readings for a kanji, including the suffix, e.g. たべる as a reading for 食. Secondly, added code to automatically consider all devoicings for kana lookup, thus handling rendaku (sequential voicing).
lars
[svn] Fixed the pluralization of stimulus (trivial change).
lars
[svn] Fixed a bug with queries left over from the pinyin X update, where some code hadn't been updated to work with the new table structure.
lars
[svn] Added the ability for the hanzi transliteration to be many-to-many, and added X for occasions where the mapping to hanzi arrived at one which was too rare. Also refactored some of the similar kanji transliteration code to improve the implementation.
lars
[svn] Removed the unused ngram module.
lars
[svn] Removed the duplicate frequency distribution class, and added tests for the UnigramModel class.
lars
[svn] Fixed a lingering path problem, and updated the test case for hanziTable.
lars
[svn] Updated data file paths, after sorting the data directory into more sensible subdirectories to manage the file explosion.
lars
[svn] Added module for accessing and pruning similarity caches.accessing and pruning similarity caches.accessing and pruning similarity caches.accessing and pruning similarity caches.
lars
[svn] Fixed a bug in the pinyin table where some strings containing unicode didn't weren't properly marked.
lars
[svn] Added a sheet of Chinese words to look up for the demo.
lars
[svn] Finished the evaluation half of the presentation, including some additional diagrams.
lars
[svn] Added configuration test for libgd, and also moved some settings from individual modules in the pixel class to the settings.py module.
lars
[svn] Added a pixel/ module to the project containing pixel similarity code.
lars
[svn] Fixed bug 81, where on Internet Explorer, the search wouldn't work correctly without the get action "search" enabled.
lars
[svn] Went through each of the problem cases provided by Tanaka-sensei, and verified each. The remaining ones were not bugs, but problems with the mapping table.
lars
[svn] Added all the potential bugs as test cases, and have started to eliminate them one by one, with the help of the explainFailedQuery script. Also fixed test cases for the kanji reading model.
lars
[svn] Updated the script to explain search failure to add explanation in the case where no transliteration exists.
lars
[svn] Changed the reading model to include any additional readings from EDICT for single-kanji words which aren't already included in Kanjidic.
lars
[svn] Added two tables for kana errors.
lars
[svn] Added a basic draft of the rinko evaluation presentation.
  1. Prev
  2. Next