Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • c99dbdaa4dba99a7ac4cd8d35cc73416dc7c5f7f
  • master default protected
  • AC
  • Aloui_Dary
  • alexis
  • classifier
  • error_predictor
  • fixhelp
  • ignore_punct
  • johannes
  • libmacaon2
  • maca_common
  • maca_graph_parser
  • maca_trans_frame_parser
  • morpho
  • new_config
  • partial_parser
  • perceptron
  • refactor
  • silvio
  • ssrnn
21 results
Created with Raphaël 2.2.013Apr12111098543230Mar292827262320199852128Feb26222120191698765426Jan2523221815141119Dec141383122Nov21201413107626Oct201626Sep14131229Jul27262421201412107529Jun262220141329May25231817159428Apr27242221201918171312117631Mar272624232018151043123Feb1610226Jan1165414Dec76123Nov211817151486432128Oct272624232120212019181714131211107654329Sep26217Aug6415Jul14131211876530Jun292827Added a new executable to create tape_alphabets.dicCorrected a bug where array values were used but not setEnabled shuffling of training examplesMaking sure we use the same seed for all experimentsbug fixing in maca_corporaadded several tools to manipulate corpora in maca_corpora as well as conll2mcfAdded a way to ensure neural network shape is not changing during decodingMade the shuffling of the training dataset optionnalWe now use less memory by only storing the current batch input of the dnn in memoryCreated a class to represent a fann file in memoryChanged the way the machine change state after an invalid movement.Added more info to debug mode in maca_tm_decoder. We are now able to see movement choices and scores.scf2fann now correctly builds tape_alphabetsFixed an overflowFinished integration of lookup lemma classifierLookup classifier almost workingAdding unknown form & lemmas to dictionnaries when decoding, this helps a lot with lemma prediciton and word embeddingsRemoved useless tabulation on the output of maca_tm_decoderFixed a bug where mcd where created with 1 extra column when read from a file.Fixed a problem where using mcf2scf created a new features dictionnary for each classifiers, even when these features dictionnaries where not emptyFixed a mistake on the getter and added a way to get the current state of the global variablefixed a bug oracle_parser_arc_eagerMerge branch 'master' into ignore_punctadded maca_trans_attach_punct a post parsing program that attaches punctuationAdded a dirty way (global variable in dico.h) to decide if we enable or not the inclusion of unknown tokens to dictionnaries. Also added a function to save the tape_alphabetsCorrected a bug in lemma predicitonAdded a type to classifiers, to make a difference between classifiers that need a learning and the others (lookup, and mono-class)Corrected a bug where we tried to read more columns than mcd hadsuppression of an incongrous printfjson2mcf : update the dependencies in json file to mcf filelemmatizer by rules is now workingfplm is no longer used when decoding (because we can predict the lemma now)Corrected a bug where dico_vec did not detect it finished reading from fileMerge branch 'master' of gitlab.lis-lab.fr:alexis.nasr/macaon2a new function in json2mcf to get the id of the first token of a sentencestill working on rules lemmatizationCreated the tagset for the lemmatizer by rules, the oracle is not finished yetCorrected a bug when searching for a movement in tagsetintroduced update of depenencies in tool json2mcfCorrected a bug where tape_alphabets.dic filename was not set