Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -1,4 +1,30 @@
|
|
1 |
---
|
2 |
license: other
|
3 |
---
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: other
|
3 |
---
|
4 |
+
|
5 |
+
## STR-2022: Dataset Description
|
6 |
+
|
7 |
+
The dataset consists of 5500 English sentence pairs that are scored and ranked on a relatedness scale ranging from 0 (least related) to 1 (most related).
|
8 |
+
|
9 |
+
## Loading the Dataset
|
10 |
+
- The sentence pairs, and associated scores, are in the file sem_text_rel_ranked.csv in the root directory. The CSV file can be read using:
|
11 |
+
|
12 |
+
```python
|
13 |
+
import pandas as pd
|
14 |
+
|
15 |
+
str = pd.read_csv('sem_text_rel_ranked.csv')
|
16 |
+
|
17 |
+
row = str.loc[0]
|
18 |
+
sent1, sent2 = row['Text'].split("\n")
|
19 |
+
score = row['Score']
|
20 |
+
```
|
21 |
+
|
22 |
+
- Relevant columns:
|
23 |
+
|
24 |
+
- Text: Sentence pair, separated by the newline character.
|
25 |
+
- Score: The semantic relatedness score between 0 and 1.
|
26 |
+
|
27 |
+
- Additionally:
|
28 |
+
- the SourceID column indicates the source dataset from which the sentence pair was drawn (see Table 2 of our paper)
|
29 |
+
- The SubsetID column indicates the sampling strategy used for the source dataset
|
30 |
+
- and the PairID is a unique identifier for each pair that also indicates its Source and Subset.
|