Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,63 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
task_categories:
|
4 |
+
- visual-question-answering
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
- zu
|
8 |
+
- id
|
9 |
+
- it
|
10 |
+
- de
|
11 |
+
- th
|
12 |
+
- ar
|
13 |
+
- ko
|
14 |
+
- zh
|
15 |
+
- hi
|
16 |
+
- ru
|
17 |
+
tags:
|
18 |
+
- multilingual
|
19 |
+
- OCR
|
20 |
+
- Plot
|
21 |
+
size_categories:
|
22 |
+
- n<1K
|
23 |
+
---
|
24 |
+
# SMPQA (Synthetic Multilingual Plot QA)
|
25 |
+
|
26 |
+
<!-- Provide a quick summary of the dataset. -->
|
27 |
+
|
28 |
+
The SMPQA evaluation dataset proposed in [Centurio: On Drivers of Multilingual Ability of Large Vision-Language Model](gregor-ge.github.io/Centurio/).
|
29 |
+
|
30 |
+
SMPQA is composed of synthetic bar plots and pie charts (generated using word lists of different languages) together with questions about those plots.
|
31 |
+
The datasets aims at providing an initial way of evaluating multilingual OCR capabilities of models in arbritrary languages.
|
32 |
+
|
33 |
+
There are two sub-tasks:
|
34 |
+
1. *Grounding* text labels from the question to the image to answer yes/no questions ("Is the bar with label X the tallest?")
|
35 |
+
2. *Reading* labels from the plot based on the question ("What is the label of the red bar?")
|
36 |
+
|
37 |
+
For more details, check out our paper.
|
38 |
+
|
39 |
+
## Dataset Details
|
40 |
+
|
41 |
+
The data is structured as follows for every language (11 languages right now; found in `smpqa.zip`):
|
42 |
+
|
43 |
+
- `bar_annotations_$lang.json` and `pie_annotations_$lang.json` contain the questions and answers for all images.
|
44 |
+
We provide unpacked examples for English for easy viewing.
|
45 |
+
There are 8 *grounding* and 5 *reading* questions per plot.
|
46 |
+
- Images `bar_plot_$i.png` and `pie_plot_$i.png` ranging from 0 to 49 (= 100 plots in total per language). We provide two unpacked English example images.
|
47 |
+
|
48 |
+
|
49 |
+
## Extending to New Languages
|
50 |
+
|
51 |
+
SMPQA is easy to expand to new languages.
|
52 |
+
We provide the code and the plot definitions (i.e., colors, size, etc.) used to generate the existing plots.
|
53 |
+
This way, you can create plots and questions that are identical and thus comparable to the existing plots in new languages.
|
54 |
+
|
55 |
+
Currently, we use word lists from [this source](https://github.com/frekwencja/most-common-words-multilingual) but other sources can also work.
|
56 |
+
|
57 |
+
|
58 |
+
## Citation
|
59 |
+
|
60 |
+
<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->
|
61 |
+
|
62 |
+
**BibTeX:**
|
63 |
+
Coming soon.
|