Datasets:
Tasks:
Sentence Similarity
Modalities:
Text
Formats:
json
Sub-tasks:
semantic-similarity-classification
Languages:
English
Size:
100K - 1M
ArXiv:
License:
espejelomar
commited on
Commit
•
bdfd116
1
Parent(s):
d282209
Update README.md
Browse files
README.md
CHANGED
@@ -7,6 +7,7 @@ pretty_name: SPECTER
|
|
7 |
---
|
8 |
|
9 |
# Dataset Card for "ESPECTER"
|
|
|
10 |
## Table of Contents
|
11 |
- [Dataset Description](#dataset-description)
|
12 |
- [Dataset Summary](#dataset-summary)
|
@@ -30,7 +31,9 @@ pretty_name: SPECTER
|
|
30 |
- [Licensing Information](#licensing-information)
|
31 |
- [Citation Information](#citation-information)
|
32 |
- [Contributions](#contributions)
|
|
|
33 |
## Dataset Description
|
|
|
34 |
- **Homepage:** [https://github.com/allenai/specter](https://github.com/allenai/specter)
|
35 |
- **Repository:** [More Information Needed](https://github.com/allenai/specter/blob/master/README.md)
|
36 |
- **Paper:** [More Information Needed](https://arxiv.org/pdf/2004.07180.pdf)
|
@@ -38,24 +41,39 @@ pretty_name: SPECTER
|
|
38 |
- **Size of downloaded dataset files:**
|
39 |
- **Size of the generated dataset:**
|
40 |
- **Total amount of disk used:** 38.3 MB
|
|
|
41 |
### Dataset Summary
|
42 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
### Supported Tasks and Leaderboards
|
|
|
44 |
[More Information Needed](https://github.com/allenai/specter)
|
|
|
45 |
### Languages
|
|
|
46 |
[More Information Needed](https://github.com/allenai/specter)
|
|
|
47 |
## Dataset Structure
|
48 |
-
Specter requires two main files as input to embed the document.
|
|
|
|
|
|
|
|
|
|
|
49 |
metadata.json format:
|
50 |
|
|
|
|
|
51 |
{
|
52 |
'doc_id': {'title': 'representation learning of scientific documents',
|
53 |
'abstract': 'we propose a new model for representing abstracts'},
|
54 |
}
|
55 |
-
|
56 |
-
### Data Instances
|
57 |
-
### Data Splits
|
58 |
-
## Dataset Creation
|
59 |
|
60 |
### Curation Rationale
|
61 |
|
@@ -119,6 +137,7 @@ metadata.json format:
|
|
119 |
year={2020}
|
120 |
}
|
121 |
|
|
|
122 |
SciDocs benchmark
|
123 |
|
124 |
SciDocs evaluation framework consists of a suite of evaluation tasks designed for document-level tasks.
|
@@ -127,11 +146,7 @@ Link to SciDocs:
|
|
127 |
|
128 |
- [https://github.com/allenai/scidocs](https://github.com/allenai/scidocs)
|
129 |
|
130 |
-
```
|
131 |
-
|
132 |
|
133 |
### Contributions
|
134 |
|
135 |
Thanks to [@armancohan](https://github.com/armancohan), [@sergeyf](https://github.com/sergeyf), [@haroldrubio](https://github.com/haroldrubio), [@jinamshah](https://github.com/jinamshah)
|
136 |
-
|
137 |
-
---
|
|
|
7 |
---
|
8 |
|
9 |
# Dataset Card for "ESPECTER"
|
10 |
+
|
11 |
## Table of Contents
|
12 |
- [Dataset Description](#dataset-description)
|
13 |
- [Dataset Summary](#dataset-summary)
|
|
|
31 |
- [Licensing Information](#licensing-information)
|
32 |
- [Citation Information](#citation-information)
|
33 |
- [Contributions](#contributions)
|
34 |
+
|
35 |
## Dataset Description
|
36 |
+
|
37 |
- **Homepage:** [https://github.com/allenai/specter](https://github.com/allenai/specter)
|
38 |
- **Repository:** [More Information Needed](https://github.com/allenai/specter/blob/master/README.md)
|
39 |
- **Paper:** [More Information Needed](https://arxiv.org/pdf/2004.07180.pdf)
|
|
|
41 |
- **Size of downloaded dataset files:**
|
42 |
- **Size of the generated dataset:**
|
43 |
- **Total amount of disk used:** 38.3 MB
|
44 |
+
|
45 |
### Dataset Summary
|
46 |
+
|
47 |
+
SPECTER: Document-level Representation Learning using Citation-informed Transformers.
|
48 |
+
A new method to generate document-level embedding of scientific documents based on
|
49 |
+
pretraining a Transformer language model on a powerful signal of document-level relatedness: the citation graph.
|
50 |
+
Unlike existing pretrained language models, SPECTER can be easily applied to
|
51 |
+
downstream applications without task-specific fine-tuning.
|
52 |
+
|
53 |
### Supported Tasks and Leaderboards
|
54 |
+
|
55 |
[More Information Needed](https://github.com/allenai/specter)
|
56 |
+
|
57 |
### Languages
|
58 |
+
|
59 |
[More Information Needed](https://github.com/allenai/specter)
|
60 |
+
|
61 |
## Dataset Structure
|
62 |
+
Specter requires two main files as input to embed the document.
|
63 |
+
A text file with ids of the documents you want to embed and a json metadata file
|
64 |
+
consisting of the title and abstract information.
|
65 |
+
Sample files are provided in the `data/` directory to get you started.
|
66 |
+
Input data format is according to:
|
67 |
+
|
68 |
metadata.json format:
|
69 |
|
70 |
+
```
|
71 |
+
|
72 |
{
|
73 |
'doc_id': {'title': 'representation learning of scientific documents',
|
74 |
'abstract': 'we propose a new model for representing abstracts'},
|
75 |
}
|
76 |
+
```
|
|
|
|
|
|
|
77 |
|
78 |
### Curation Rationale
|
79 |
|
|
|
137 |
year={2020}
|
138 |
}
|
139 |
|
140 |
+
```
|
141 |
SciDocs benchmark
|
142 |
|
143 |
SciDocs evaluation framework consists of a suite of evaluation tasks designed for document-level tasks.
|
|
|
146 |
|
147 |
- [https://github.com/allenai/scidocs](https://github.com/allenai/scidocs)
|
148 |
|
|
|
|
|
149 |
|
150 |
### Contributions
|
151 |
|
152 |
Thanks to [@armancohan](https://github.com/armancohan), [@sergeyf](https://github.com/sergeyf), [@haroldrubio](https://github.com/haroldrubio), [@jinamshah](https://github.com/jinamshah)
|
|
|
|