#JumpQA Automating ground truth generation.

JumpQA is a projects for generating questions about a corpus. More specifically, it takes a directory of TRECs (in .xml form) and outpus a .csv.

##Requirements To install this project, you will need:

• Maven

• Java 8 SDK

###Maven

1. Download the binary relevant to your machine.
2. Install maven to the directory of your choice.
3. Edit your ~/.bash_profile to include the lines

    export M2_HOME=/path/to/maven
    export PATH=$PATH:$M2_HOME/bin 

###Java 8

1. Download and install the binary relevant to your machine.
2. Edit your ~/.bash_profile to include the line

    export JAVA_HOME=/path/to/java8

##Installation Clone the repository from git. You'll need your IBM ID and password.

git clone https://github.com/cognitive-catalyst/JumpQA

You will be prompted for your IBM ID and password. Once downloaded, run the install script in the main directory.

cd JumpQA/
chmod 755 ./install.sh
./install.sh

##Basic Usage Sample usage of the subprojects within the JumpQA repository. Each subproject contains more thorough documentation and usage instructions.

Before JumpQA can run on a corpus, the corpus must be converted to a .json file. This vastly speeds up how quickly JumpQA can process a corpus. Many of the fields in the TRECs are currently removed. If you wish to keep non-default ones, you will need to edit corpus2json. Eventually, it will be possible to specify which fields to keep in the .properties file.

Once the corpus is in a JSON, JumpQA can process the corpus and generate ground truth.

###Converting Corpus to JSON 0. cd into JumpQa/

1. Edit corpus2json.properties. The file should look something like:

input=sample/
output=sample/output.json

input is a directory holding a list of .xml trecs. Change it to the directory holding your corpus XMLs. output is the output .json file.

2. Run java -jar target/corpus2json-0.1.0.jar. If you have a different version, replace 0.1.0 with the one you are using.

###Using JumpQA 0. cd into JumpQa/

1. CD Edit jumpqa.properties. The file should look something like:

corpus=health-corpus.json
templates=templates.csv
output=health.csv

templates is the templates file to process the TRECs with. trecs is the inputs JSON file. output is the output CSV file. 2. Run java -jar target/jumpqa-0.1.0.jar

##Projects Click on the subproject's name to enter its directory.

###JumpQA This is the main project.

JumpQA takes a corpus in JSON form and creates a set of ground truth. The others are related projects and dependencies.

###Corpus This converts a directory of TRECs to a JSON file that JumpQA can use.

###[Corpus TF-IDF](Corpus TF-IDF/README.md) Calcaulates TFIDF of terms in a corpus for each document.

###NeuralNet Library for neural networks; to be used in JumpQA heuristics. This will eventually be moved to be its own project.

###Random A library which includes an ArrayList which iterates randomly.

##Maintainer Will Beason, wabeason@us.ibm.com