Java code for the baseline solutions for the CPROD1 contest <<<<<<< HEAD
===============
Baseline 1: Baseline 2:
Most actions done with the Baseline class. This program needs as an input a config file in which all actions are defined. java -Xmx4500m -cp "./bin;./lib/*;./lib/mallet" baseline/Baseline <config.file>
The config file is a json file in which you can specify the following parameters (paramters with an (*) are mandatory) parameter name description Default value type train whether to train the model true boolean test whether to test the model true boolean filter whether to filter the results false boolean link whether to link to the catalog false boolean useDictionary whether to use dictionary attributes for the crf true boolean usePatternFeatures whether to use pattern features for the crf false boolean useTokenFeature whether to use token features for the crf false boolean evaluate whether to evaluate the result against the ground truth false boolean tokenWindow token window for the crf 1 integer home_path directory where all data files can be found "" String training_text_file name of the training text file "training-annotated-text.json" String training_annotations_file name of the csv file with traning annotations "training-disambiguated-product-mentions.120725.csv" String test_text_file name of the test text file "evaluation-text.json" String regex_feature_file name of the file with regex features for CRF "regexfeatures_test.json" String dictionary_file name of the dictionary file "dictionary.dat" String annotation_result_file name of the file where the results should be written $model + "_results.csv" String ground_truth_annotations_file name of the file with ground truth annotations "evaluation-disambiguated-product-mentions.120725.csv" String index_file name of the file with index of the product catalog "index.txt" String product_dictionary_file name of file with additionary product terms null String filters list of filters to use empty array one or more of {BLACKLIST, BRANDNAME, FIRST_TOKEN, MENTION_LENGTH, PROCEEDINGWORDS_ALLOWED, PROCEEDINGWORDS_BLACKLIST1, PROCEEDINGWORDS_BLACKLIST2,SPECIAL_CHAR, MILD, STRICT} model Type of model to use BASELINE1 {BASELINE1, BASELINE2, BASELINE3, RULEBASE} model_name Name of the model that corresponds to this config baseline1 String
Other programs: java -cp "./bin;./lib/*;./lib/mallet" baseline/Ensemble <results.file1> <results.file2> <ground.truth.file1> <output.file> Merges two different results files, giving priority to the annotations found in the first file.
java -cp "./bin;./lib/*;./lib/mallet" baseline/PrintAnnotations <text.file> <result.file>
Prints the annotations with the mention phrases included to standard output.
java -cp "./bin;./lib/*;./lib/mallet" baseline/CatalogHandler
args:
-catalog-file <
=======
dc6dd16f034ed38770c31fdf6fa8f092fc742466