Unable to execute the pipeline

[durgeshnandini]

I am trying to run the pipeline but I am getting the following error as shown in the attachment. Please suggest a solution.

[ktymoshenko]

Hello,

please have a look at the "Running experiments on WikiQA" section of the documentation (https://github.com/iKernels/ANLPIR-2018/tree/master/partIV/qapipeline/RelationalTextRanking#running-experiments-on-wikiqa). From your screenshot it looks that you are missing some parameters.

Here is an example of how to launch the experiment_launcher script, which creates a sh-file with the commands to perform structure extraction, training of an svm with the classification PTK kernel, and prediction on the WikiQA corpus with examples represented as shallow-pos structures:

python scripts/experiment_launchers/experiment_launcher.py -l ${corpus_name} -o scripts/generated_scripts -c CH -p "-t 5 -F 3 -C T -m 1000" -e it.unitn.nlpir.experiment.fqa.CHExperiment -suf T -s it.unitn.nlpir.system.core.ClassTextPairConversion -ate " -skipAllSame" -ade " -skipAllSame"

The above will generate a sh-file which contains the commands to run the pipeline end-to-end. The meanings of the parameters are as follows:

• -l ${corpus_name}: is the id of the corpus configuration file (paths to corpus questions/answer files etc). When you specify this parameter, the system uses the scripts/experiment_launchers/corpus_settings/${corpus_name}.settings configuration file; for example if it is "-l wikiqa", then the system will use the scripts/experiment_launchers/corpus_settings/wikiqa.settings file (https://github.com/iKernels/ANLPIR-2018/blob/master/partIV/qapipeline/RelationalTextRanking/scripts/experiment_launchers/corpus_settings/wikiqa.settings). If you are doing some experiment on a new corpus, you have to create this file yourself with paths to the relevant files; Note that train_n, dev_n, test_n specify the maximum amount of candidate answer sentences to be used per question for training, development and testing;
• -o scripts/generated_scripts -> -o specifies the folder where the resulting sh-file with all the commands should be stored
• -c (CH) is the id of the experiment, will be used as a part of the name of the folder with the experimental data; all the files generated by the pipeline for you specific experiment will be placed into "data/examples/${corpus_name}CH"
• -p "-t 5 -F 3 -C T -m 1000" -> parameters for svm_learn for training
• -e it.unitn.n)lpir.experiment.fqa.CHExperiment -> which structure do you want to build?
• -suf T : suffix to add to the prediction files (in this example the name of the prediction files will be svm_T.pred)
• -s it.unitn.nlpir.system.core.ClassTextPairConversion -> do you want to generate the data for the Reranking or Classification kernel? (in the example it is classification)
• -ate " -skipAllSame" -ade " -skipAllSame" -> skip all questions which have only positive and only negative answers when training and testing (questions with only negative answers are skipped by default when training; If you want to keep them, add the -atr " -keepAllNegatives" parameter.)