andorei commited on
Commit
c36b655
1 Parent(s): 33834c9

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +30 -0
README.md ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ task_categories:
3
+ - question-answering
4
+ language:
5
+ - en
6
+ ---
7
+
8
+ We present a dataset for graph-based question answering. The dataset consists of <question; candidate answer> pairs. For each candidate, we present a graph that is obtained by finding the shortest path between named entities mentioned in a question and a candidate answer. As a knowledge graph, we adopted Wikidata. Our dataset has the following fields:
9
+
10
+ * `sample_id` - an identifier for <question, candidate answer>;
11
+
12
+ * `question` - question text;
13
+
14
+ * `questionEntity` - comma-separated list of names (textual strings) for Wikidata concepts mentioned in a given question;
15
+
16
+ * `answerEntity` - a textual name of candidate answer (candidate is a concept from Wikidata) for the given question;
17
+
18
+ * `groundTruthAnswerEntity` - a textual name of ground truth answer (answer is a concept from Wikidata) for the given question;
19
+
20
+ * `answerEntityId` - a Wikidata id of candidate answer (see "answerEntity" column). Example: "Q2599";
21
+
22
+ * `questionEntityId` - a comma-separated list of Wikidata ids for concepts mentioned in a given question (list of ids for mentions from "questionEntity" column);
23
+
24
+ * `groundTruthAnswerEntityId` - a Wikidata id of ground truth answer (see "answerEntity" column). Example: "Q148234";
25
+
26
+ * `correct` - either "True" or "False". The field indicates whether a <question, answer candidate> is correct, i.e., candidate answer is a true answer to the given question;
27
+
28
+ * `graph` - a shortest-path graph for a given <question, candidate answer> pair. The graph is obtained by taking the shortest paths from all mentioned concepts ("questionEntityId" column) to a candidate answer("answerEntityId" column) in the knowledge graph of Wikidata. The graph is stored in "node-link" JSON format from NetworkX. You can import the graph using the [node_link_graph](https://networkx.org/documentation/stable/reference/readwrite/generated/networkx.readwrite.json_graph.node_link_graph.html).
29
+
30
+ Please see our [Github](https://github.com/uhh-lt/TextGraphs17-shared-task.git) for baselines, and useful code.