File size: 5,092 Bytes
ba48b51
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4e49c6d
 
 
 
 
 
 
 
 
 
 
 
ba48b51
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4e49c6d
ba48b51
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
# Dataset Card for WineSensed

## Table of Contents
- [Table of Contents](#table-of-contents)
- [Dataset Description](#dataset-description)
  - [Dataset Summary](#dataset-summary)
  - [Languages](#languages)
- [Dataset Structure](#dataset-structure)
  - [Data Fields](#data-fields)
- [Additional Information](#additional-information)
  - [Licensing Information](#licensing-information)
  - [Citation Information](#citation-information)

## Dataset Description

- **Homepage:** [WineSensed Dataset](https://https://thoranna.github.io/learning_to_taste/)
- **Repository:**
- **Paper:** [Paper](https://arxiv.org/pdf/2308.16900.pdf)

### Dataset Summary

The dataset encompasses 897k images of wine labels and 824k reviews of wines
curated from the Vivino platform. It has over 350k unique vintages, annotated
with year, region, rating, alcohol percentage, price, and grape composition.
We obtained fine-grained flavor annotations on a subset by conducting a wine-tasting experiment
with 256 participants who were asked to rank wines based on their similarity in flavor,
resulting in more than 5k pairwise flavor distances.

## Hugging Face Datasets

use the following command to load the dataset in Hugging Face Datasets library:

```python
dataset = load_dataset('Dakhoo/L2T-NeurIPS-2023', 'vintages')
dataset = load_dataset('Dakhoo/L2T-NeurIPS-2023', 'napping_participants')
dataset = load_dataset('Dakhoo/L2T-NeurIPS-2023', 'wt_session')
dataset = load_dataset('Dakhoo/L2T-NeurIPS-2023', 'small')
dataset = load_dataset('Dakhoo/L2T-NeurIPS-2023', 'all')
```



## Dataset Structure

### Data Fields

The dataset contains the file metadata.zip, consisting of the files participants.csv, which contains information connecting participants to annotations in the experiment, images_reviews_attributes.csv, which contains reviews, links to images, and wine attributes, and napping.csv, which contains the coordinates of each wine on the napping paper alongside information connecting each coordinate pair to the wine being annotated and the participant who annotated it. The chunk_<chunk num>.zip folders contain the images of the wines in the dataset in .jpg format. 

#### napping.csv contains the following fields:

- session_round_name: session number during the event_name, at most three sessions per event (maps to experiment_round in participants.csv)
- event_name: name of the data collection event (maps to the same attribute in participants.csv)
- experiment_no: which number the napping paper was in the list of papers returned for this session_round_name (maps to experiment_no in participants.csv)
- experiment_id: id the wine being annotated was given in the experiment
- coor1: x-axis coordinate on the napping paper
- coor2: y-axis coordinate on the napping paper
- color: color of the sticker used

#### participants.csv contains the following fields:

- session_round_name: session number during the event_name, at most three sessions per event (maps to experiment_round in napping.csv)
- event_name: name of data-collection event (maps to event_name in napping.csv)
- experiment_no: which number the napping paper was in the list of papers returned for this session_round_name (maps to experiment_no in napping.csv)
- round_id: round number (from 1-3)
- participant_id: id the participant was given in the experiment

#### images_reviews_attributes.csv contains the following fields:

- vintage_id: vintage id of the wine
- image: image link (each .jpg in chunk_<chunk num>.zip can be mapped to a corresponding image link in this column by removing the /p prefix from the link).
- review: user review of the wine
- experiment_id: id the wine got during data collection (each experiment_id can be mapped to the same column in napping.csv)
- year: year the wine was produced
- winery_id: id of the winery that produced the wine
- wine: name of the wine
- alcohol: the wine's alcohol percentage
- country: the country where the wine was produced
- region: the region where the wine was produced
- price: price of the wine in USD (collected 05/2023)
- rating: average rating of the wine (collected 05/2023)
- grape: the wine's grape composition, represented as a comma-separated list ordered in descending sequence of the percentage contribution of each grape variety to the overall blend.

## Dataset Creation

Follow the instructions in this [link](https://thoranna.github.io/learning_to_taste/) to create the dataset.

## Additional Information

### Licensing Information

LICENSE AGREEMENT
=================
 - WineSensed by Thoranna Bender, Simon Søresen, Alireza Kashani, Kristjan Eldjarn, Grethe Hyldig,
   Søren Hauberg, Serge Belongie, Frederik Warburg is licensed under a CC BY-NC-ND 4.0 Licence


### Citation Information

```
 @article{bender2023learning,
  title={Learning to Taste: A Multimodal Wine Dataset},
  author={Bender, Thoranna and S{\o}rensen, Simon M{\o}e and Kashani, Alireza and Hjorleifsson, K Eldjarn and Hyldig, Grethe and Hauberg, S{\o}ren and Belongie, Serge and Warburg, Frederik},
  journal={arXiv preprint arXiv:2308.16900},
  year={2023}
```