KaDnt/Bamanankan

# Bamanankan README Contact: kathrindonandt@yahoo.de To run the programs, the following programs have to be installed: 1. python 3; 2. NLTK (nltk.org); 3. CRFSuite pypi.python.org Running the programs will create the folder "nltk_data" in C:/User directory, where the corpus files in the folder "Corpus" will be automatically copied to. Containing Folders ================== !Note: Corpusreader htmlreader.py is supposed to read files in Daba HTML format (Bambara Reference Corpus). Corpus - Folder which has to contain the corpus files Models - folder empty, but will be used to store models to when CRFTaggers is trained Results - folder empty, results of analyzeContingency.py will be saved to this folder Containing Files ================ analyzeContingency.py - calculates percentage of words tagged i which in reality are j; saves result to a file in Results folder - looks for the words which are responsible for these errors and save each error to a file in Results folder - saves confusionmatrix to Results folder backoffCombi.py - combines taggers according to the backoff-chaining given in NLTK bambara_tagging_htmlreaderALL.py - loads the corpus files and creates a reader needed to work with the sentences/words in the corpus - used by create_reader bamadaba.txt - the Bamadaba dictionary (cormand.huma-num.fr) bamadaba_non_tonal.txt - the Bamadaba dictionary (cormand.huma-num.fr), but without tones confusionmatrix.py - slightly modified version of confusionmatrix.py of NLTK - function was added so that switches (tagger tagged word with tag A instead of the right tag B) can be analyzed create_reader.py - uses bambara_tagging_htmlreaderALL to create a reader (with htmlreaderALL.py) needed to work with the corpus files crf.py - modified crf of NLTK (features added) CrossValidation.py - implementation of a 9-fold crossvalidation disambiguation.py - removes sente …