create readme file
Browse files
README.md
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- nl
|
| 4 |
+
tags:
|
| 5 |
+
- text-classification
|
| 6 |
+
- sentiment-analysis
|
| 7 |
+
datasets:
|
| 8 |
+
- train
|
| 9 |
+
- test
|
| 10 |
+
- validation
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
## Dataset overview
|
| 14 |
+
This is a dataset that contains restaurant reviews gathered in 2019 using a webscraping tool in Python. Reviews on restaurant visits and restaurant features were collected for Dutch restaurants.
|
| 15 |
+
The dataset is formatted using the 🤗[DatasetDict](https://huggingface.co/docs/datasets/index) format and contains the following indices:
|
| 16 |
+
- train, 116693 records
|
| 17 |
+
- test, 14587 records
|
| 18 |
+
- validation, 14587 records
|
| 19 |
+
|
| 20 |
+
The dataset holds both information of the restaurant level as well as the review level and contains the following features:
|
| 21 |
+
- [restaurant_ID] > unique restaurant ID
|
| 22 |
+
- [restaurant_review_ID] > unique review ID
|
| 23 |
+
- [michelin_label] > indicator whether this restaurant was awarded one (or more) Michelin stars prior to 2020
|
| 24 |
+
- [score_total] > restaurant level total score
|
| 25 |
+
- [score_food] > restaurant level food score
|
| 26 |
+
- [score_service] > restaurant level service score
|
| 27 |
+
- [score_decor] > restaurant level decor score
|
| 28 |
+
- [fame_reviewer] > label for how often a reviewer has posted a restaurant review
|
| 29 |
+
- [reviewscore_food] > review level food score
|
| 30 |
+
- [reviewscore_service] > review level service score
|
| 31 |
+
- [reviewscore_ambiance] > review level ambiance score
|
| 32 |
+
- [reviewscore_waiting] > review level waiting score
|
| 33 |
+
- [reviewscore_value] > review level value for money score
|
| 34 |
+
- [reviewscore_noise] > review level noise score
|
| 35 |
+
- [review_text] > the full review that was written by the reviewer for this restaurant
|
| 36 |
+
- [review_length] > total length of the review (tokens)
|
| 37 |
+
|
| 38 |
+
## Purpose
|
| 39 |
+
The restaurant reviews submitted by visitor can be used to model the restaurant scores (food, ambiance etc) or used to model Michelin star holders. In [this blog series](https://medium.com/broadhorizon-cmotions/natural-language-processing-for-predictive-purposes-with-r-cb65f009c12b) we used the review texts to predict next Michelin star restaurants, using R.
|