Link:  www.coli.uni-saarland.de/projects/smile/data/data.tar

Data consists of both development(dev) data as well as test data. Each of these directories(dev/test) consists of different script scenarios in separate sub-directory for e.g. preparing coffee is in the sub-directory test/coffee. This data comes from Regneri et al.[1] and OMICS corpus[2]. For details regarding preprocessing steps involved in extracting data please refer to [1],[2],[3] and [4](References below)

Each of the scenario subdirectory consists of 4 files in xml format :

event-ordering.xml
event-paraphrasing.xml
orig-data.xml
segmented.xml

**********************************************
event-ordering.xml : This file has the gold event-pairs for event ordering tasks. Each event is split into predicate and arguments. General format for each event pair is : 
			
<label type=FOLLOWUP/NO_FOLLOWUP>
    <item script="5" slot="5" text=predicate1>
         <ptcp text=argument1/>
         <ptcp text=argument2/>
         <ptcp text=argument3/>
   </item>
   <item script="1" slot="6" text=predicate2>
         <ptcp text=argument3/>
         <ptcp text="imp_protagonist"/>
    </item>
</label>

Label can be of two types : FOLLOWUP(first event follows the second) or NO_FOLLOWUP(first event does not follow the second event). 

**********************************************

event-paraphrasing.xml : This file has the gold event-pairs for event paraphrasing task. Each event is split into predicate and arguments. General format for each event pair is : 
			
<label type=NO_PARAPHRASE/PARAPHRASE>
    <item script="5" slot="5" text=predicate1>
         <ptcp text=argument1/>
         <ptcp text=argument2/>
         <ptcp text=argument3/>
   </item>
   <item script="1" slot="6" text=predicate2>
         <ptcp text=argument3/>
         <ptcp text="imp_protagonist"/>
    </item>
</label>

Label can be of two types : PARAPHRASE(two events are paraphrases) or NO_PARAPHRASE(two events are not paraphrases). 

**********************************************

orig-data.xml : This file consists of the original scripts for the scenario. In order words it is collection of ESDs for that scenario.
General format is :

<script id="1">
  <item slot="1" text="Find bus stop" />
  <item slot="2" text="Wait at bus stop" />
  <item slot="3" text="flag bus" />
                |
                |
    SEQUENCE OF EVENTS IN THIS SCENARIO
                |
                |
                |
  <item slot="8" text="watch route" />
  <item slot="9" text="use alert bell to signal stop" />
  <item slot="10" text="disembark when bus stops" />
</script>



**********************************************

segmented.xml : This file contains the training data for the scenario. It consists of sequence of events involved but the events are splited into predicate and arguments.
General format is  :

<script id="1">
  <item slot="1" text=predicate1>
    <ptcp text="imp_protagonist"></ptcp>   //In case no argument is present we represent this as implied protagonist but this is not used while composing the event in the model
  </item>
  <item slot="2" text=predicate2>
    <ptcp text=argument1/>
    <ptcp text=argument2/>
  </item>
  <item slot="3" text=predicate3>
    <ptcp text=argument1/>
    <ptcp text=argument2/>
    <ptcp text=argument3/>
    <ptcp text=argument4/>
  </item>
  <item slot="4" text=predicate4>
   <ptcp text=argument1/>
  </item>
</script>


**********************************************

REFERENCES:

[1] Michaela Regneri, Alexander Koller, and Manfred Pinkal. 2010. Learning script knowledge with web experiments. In Proceedings of ACL.
[2] Rakesh Gupta and Mykel J. Kochenderfer. 2004. Common sense data acquisition for indoor mobile robots. In Proceedings of AAAI.
[3] Lea Frermann, Ivan Titov, and Manfred Pinkal. A hierarchical bayesian model for unsupervised induction of script knowledge. In EACL, Gothenberg, Sweden, 2014.
[4] Ashutosh Modi and Ivan Titov. 2014. Inducing Neural Models of Script Knowledge. CoNLL 2014.