Sefika commited on
Commit
168397c
·
verified ·
1 Parent(s): 977c190

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +25 -9
README.md CHANGED
@@ -1,11 +1,12 @@
1
  ---
2
  license: mit
 
3
  task_categories:
4
  - text-classification
5
  - text-generation
6
  language:
7
  - en
8
- pretty_name: Relation Extraction for Inverse Relations (FewRel & TekGen)
9
  size_categories:
10
  - 1K<n<10K
11
  tags:
@@ -15,6 +16,21 @@ tags:
15
  - inverse-relations
16
  ---
17
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
  # Relation Extraction for Inverse Relations
19
 
20
  This dataset is designed to evaluate the robustness of Relation Extraction (RE) models, with a specific focus on **Inverse Relations** and entity substitution. It is derived from the **FewRel** and **TekGen** benchmarks (from Text2Bench).
@@ -23,8 +39,9 @@ This dataset is designed to evaluate the robustness of Relation Extraction (RE)
23
 
24
  The primary goal of this dataset is to test whether models can recognize a relationship regardless of the direction or the surface form of the entities. It includes:
25
  1. **Inverse Mapping**: Every sample includes both the forward relation (e.g., *Mother*) and its logical inverse (e.g., *Child*).
26
- 2. **Artificial Substitutions**: A version of the dataset where real-world entities are replaced with synthetic names (e.g., replacing "Denmark" with "Emilyfort") to test if models rely on memorized entity knowledge rather than linguistic context.
27
-
 
28
  ### Relation Examples
29
  | Forward Relation | Inverse Relation |
30
  | :--- | :--- |
@@ -37,18 +54,17 @@ The primary goal of this dataset is to test whether models can recognize a relat
37
 
38
  The dataset is provided in JSON format and split into four main distributions:
39
 
40
- * **Original FewRel/TekGen**: The ground-truth sentences and relations.
41
- * **Artificial FewRel/TekGen**: Sentences where entities have been substituted with synthetic placeholders.
42
- * **Relation Metadata**: Supplemental files providing human-readable definitions for Wikidata P-IDs and TekGen relation strings.
43
 
44
  ### Key Fields
45
 
46
  * `tokens`: The input text (tokenized or raw string) in FewRel.
47
- * `sent`: The input sentence (raw string) in Tekgen
48
  * `head` / `tail`: The subject and object entities.
49
  * `head_to_tail`: The ID/Name of the forward relation.
50
  * `tail_to_head`: The ID/Name of the inverse relation.
51
- * `artificial_data`: A mapping showing which `Original` entities were changed to which `Artifical` placeholders.
52
 
53
  ## Usage
54
 
@@ -59,7 +75,7 @@ You can load this dataset for tasks such as:
59
  ### Repository
60
 
61
  For scripts and more technical details on how the artificial data was generated, visit the official repository:
62
- 👉 [sefeoglu/inverserelations](https://github.com/sefeoglu/inverserelations)
63
 
64
  ## Citation
65
 
 
1
  ---
2
  license: mit
3
+
4
  task_categories:
5
  - text-classification
6
  - text-generation
7
  language:
8
  - en
9
+ pretty_name: Relation Extraction for Inverse Relations FewRel
10
  size_categories:
11
  - 1K<n<10K
12
  tags:
 
16
  - inverse-relations
17
  ---
18
 
19
+ <script type="application/ld+json">
20
+ {
21
+ "@context": {
22
+ "@language": "en",
23
+ "@vocab": "https://schema.org/",
24
+ "cr": "http://mlcommons.org/croissant/",
25
+ "sc": "https://schema.org/"
26
+ },
27
+ "@type": "sc:Dataset",
28
+ "name": "InverseRelationsBenchmark",
29
+ "description": "Benchmark based on FewRel with original entity mentions and synthetic substitutions for robustness testing.",
30
+ "conformsTo": "http://mlcommons.org/croissant/1.0"
31
+ }
32
+ </script>
33
+
34
  # Relation Extraction for Inverse Relations
35
 
36
  This dataset is designed to evaluate the robustness of Relation Extraction (RE) models, with a specific focus on **Inverse Relations** and entity substitution. It is derived from the **FewRel** and **TekGen** benchmarks (from Text2Bench).
 
39
 
40
  The primary goal of this dataset is to test whether models can recognize a relationship regardless of the direction or the surface form of the entities. It includes:
41
  1. **Inverse Mapping**: Every sample includes both the forward relation (e.g., *Mother*) and its logical inverse (e.g., *Child*).
42
+ -- *Original data* : original_fewrel_inverse.json
43
+ 3. **Synthetics Substitutions**: A version of the dataset where real-world entities are replaced with synthetic names (e.g., replacing "Denmark" with "Emilyfort") to test if models rely on memorized entity knowledge rather than linguistic context.
44
+ -- *synthetic_fewrel_inverse.json*
45
  ### Relation Examples
46
  | Forward Relation | Inverse Relation |
47
  | :--- | :--- |
 
54
 
55
  The dataset is provided in JSON format and split into four main distributions:
56
 
57
+ * **Original FewRel**: The ground-truth sentences and relations. (*original_fewrel_inverse.json*)
58
+ * **Synthetic FewRel**: Sentences where entities have been substituted with synthetic placeholders. (*synthetic_fewrel_inverse.json*)
59
+ * **Relation Metadata**: Supplemental files providing human-readable definitions for Wikidata P-IDs (*fewrel_inverse_relations.json*)
60
 
61
  ### Key Fields
62
 
63
  * `tokens`: The input text (tokenized or raw string) in FewRel.
 
64
  * `head` / `tail`: The subject and object entities.
65
  * `head_to_tail`: The ID/Name of the forward relation.
66
  * `tail_to_head`: The ID/Name of the inverse relation.
67
+ * `artificial_data`: A mapping showing which `Original` entities were changed to which `Synthetic` placeholders.
68
 
69
  ## Usage
70
 
 
75
  ### Repository
76
 
77
  For scripts and more technical details on how the artificial data was generated, visit the official repository:
78
+ [sefeoglu/inverserelations](https://github.com/sefeoglu/inverserelations)
79
 
80
  ## Citation
81