Datasets:
ArXiv:
License:
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,96 @@
|
|
1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
+
annotations_creators:
|
3 |
+
- expert-generated
|
4 |
+
- found
|
5 |
+
configs:
|
6 |
+
- config_name: default
|
7 |
+
data_files:
|
8 |
+
- split: validation
|
9 |
+
path: validation.json
|
10 |
+
- split: test
|
11 |
+
path: test.json
|
12 |
+
dataset_info:
|
13 |
+
features:
|
14 |
+
- name: question_id
|
15 |
+
dtype: string
|
16 |
+
- name: question
|
17 |
+
dtype: string
|
18 |
+
- name: tables
|
19 |
+
dtype: list
|
20 |
+
- name: topic
|
21 |
+
dtype: string
|
22 |
+
- name: python_solution
|
23 |
+
dtype: string
|
24 |
+
- name: ground_truth
|
25 |
+
sequence: float
|
26 |
license: mit
|
27 |
---
|
28 |
+
## Dataset Description
|
29 |
+
|
30 |
+
**KnowledgeMath** is a knowledge-intensive dataset focused on mathematical reasoning within the domain of finance. It requires the model to comprehend specialized financial terminology and to interpret tabular data presented in the questions.
|
31 |
+
**KnowledgeMath** includes **1200 QA examples** across 7 key areas in finance. These examples were collected from financial experts and feature detailed solution annotations in Python format.
|
32 |
+
|
33 |
+
## Dataset Information
|
34 |
+
|
35 |
+
- Paper: https://arxiv.org/abs/2311.09797
|
36 |
+
- Code: https://github.com/yale-nlp/KnowledgeMath
|
37 |
+
- Leaderboard: will be released soon!
|
38 |
+
|
39 |
+
### Data Downloading and Usage
|
40 |
+
|
41 |
+
All the data examples were divided into two subsets: *validation* and *test*.
|
42 |
+
|
43 |
+
- **validation**: 200 examples used for model development, validation, or for those with limited computing resources.
|
44 |
+
- **test**: 1000 examples for standard evaluation. We will not publicly release the annotated solution and answer for the test set.
|
45 |
+
|
46 |
+
You can download this dataset by the following command:
|
47 |
+
|
48 |
+
```python
|
49 |
+
from datasets import load_dataset
|
50 |
+
|
51 |
+
dataset = load_dataset("yale-nlp/KnowledgeMath")
|
52 |
+
```
|
53 |
+
|
54 |
+
Here are some examples of how to access the downloaded dataset:
|
55 |
+
|
56 |
+
```python
|
57 |
+
# print the first example on the validation set
|
58 |
+
print(dataset["validation"][0])
|
59 |
+
|
60 |
+
# print the first example on the test set
|
61 |
+
print(dataset["test"][0])
|
62 |
+
```
|
63 |
+
|
64 |
+
### Data Format
|
65 |
+
|
66 |
+
The dataset is provided in json format and contains the following attributes:
|
67 |
+
|
68 |
+
```json
|
69 |
+
{
|
70 |
+
"question_id": [string] The question id,
|
71 |
+
"question": [string] The question text,
|
72 |
+
"tables": [list] List of Markdown-format tables associated with the question,
|
73 |
+
"python_solution": [string] Python-format and executable solution by financial experts. The code is written in a clear and executable format, with well-named variables and a detailed explanation,
|
74 |
+
"ground_truth": [integer] Executed result of `python solution`, rounded to three decimal places,
|
75 |
+
"topic": [string] The related financial area of the question
|
76 |
+
}
|
77 |
+
```
|
78 |
+
|
79 |
+
### Automated Evaluation
|
80 |
+
|
81 |
+
To automatically evaluate a model on **KnowledgeMath**, please refer to our GitHub repository [here](https://github.com/yale-nlp/KnowledgeMath).
|
82 |
+
|
83 |
+
## Citation
|
84 |
+
|
85 |
+
If you use the **KnowledgeMath** dataset in your work, please kindly cite the paper:
|
86 |
+
|
87 |
+
```
|
88 |
+
@misc{zhao2023knowledgemath,
|
89 |
+
title={KnowledgeMath: Knowledge-Intensive Math Word Problem Solving in Finance Domains},
|
90 |
+
author={Yilun Zhao and Hongjun Liu and Yitao Long and Rui Zhang and Chen Zhao and Arman Cohan},
|
91 |
+
year={2023},
|
92 |
+
eprint={2311.09797},
|
93 |
+
archivePrefix={arXiv},
|
94 |
+
primaryClass={cs.CL}
|
95 |
+
}
|
96 |
+
```
|