Wyoung1 commited on
Commit
d5197b7
·
verified ·
1 Parent(s): b33a1af

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +82 -3
README.md CHANGED
@@ -1,3 +1,82 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ ---
4
+ # M-ABSA
5
+
6
+ This repo contains the data and code for our paper ****M-ABSA: A Multilingual Dataset for Aspect-Based Sentiment Analysis****.
7
+
8
+ [![arXiv](https://img.shields.io/badge/arXiv-2502.11824-b31b1b.svg)](https://arxiv.org/abs/2502.11824)
9
+
10
+
11
+ # Data Description:
12
+
13
+ All datasets are stored in the data/ folder:
14
+
15
+ - All dataset contains 7 domains.
16
+
17
+ ```
18
+ domains = ["coursera", "hotel", "laptop", "restaurant", "phone", "sight", "food"]
19
+ ```
20
+ - Each dataset contains 21 languages.
21
+ ```
22
+ langs = ["ar", "da", "de", "en", "es", "fr", "hi", "hr", "id", "ja", "ko", "nl", "pt", "ru", "sk", "sv", "sw", "th", "tr", "vi", "zh"]
23
+ ```
24
+ - Each dataset is divided into training, validation, and test sets. Each sentence is separated by __"####"__, with the first part being the sentence and the second part being the corresponding triplet. Here is an example, where the triplet includes __[aspect, category, sentiment]__.
25
+
26
+ ```
27
+ This coffee brews up a nice medium roast with exotic floral and berry notes .####[['coffee', 'food quality', 'positive', 'nice']]
28
+ ```
29
+
30
+ ## Requirements
31
+
32
+ We recommend you to install the specified version of the following packages:
33
+
34
+ - transformers==4.0.0
35
+ - sentencepiece==0.1.91
36
+ - pytorch_lightning==0.8.1
37
+
38
+ ## Quick Start for the Baseline
39
+
40
+ - Set up the environment as described in the above section.
41
+ - Download the pre-trained mT5-base model from [https://huggingface.co/google/mt5-base](https://huggingface.co/google/mt5-base) and place it under the folder mT5-base/ .
42
+ - Run command bash run.sh, which train the model on source language under UABSA/TASD task.
43
+ - Run command bash evaluate.sh, which test the model on target language under UABSA/TASD task.
44
+
45
+ ****Detailed Usage****
46
+
47
+ We conduct experiments on two ABSA subtasks with M-ABSA dataset in the paper, you can change the parameters in run.sh to try them:
48
+
49
+ ```
50
+ python main.py --task tasd \
51
+ --dataset hotel \
52
+ --model_name_or_path mt5-base \
53
+ --paradigm extraction \
54
+ --n_gpu 0 \
55
+ --do_train \
56
+ --do_direct_eval \
57
+ --train_batch_size 16 \
58
+ --gradient_accumulation_steps 2 \
59
+ --eval_batch_size 16 \
60
+ --learning_rate 3e-4 \
61
+ --num_train_epochs 5
62
+ ```
63
+
64
+ - $task refers to one of the ABSA task in [uabsa, tasd]
65
+ - $dataset refers to one of the seven datasets in [food, restaurant, coursera, laptop, sight, phone, hotel]
66
+
67
+
68
+ ## Citation
69
+
70
+ If the code or dataset is used in your research, please star our repo and cite our paper as follows:
71
+ ```
72
+ @misc{wu2025mabsa,
73
+ title={M-ABSA: A Multilingual Dataset for Aspect-Based Sentiment Analysis},
74
+ author={Chengyan Wu and Bolei Ma and Yihong Liu and Zheyu Zhang and Ningyuan Deng and Yanshu Li and Baolan Chen and Yi Zhang and Barbara Plank and Yun Xue},
75
+ year={2025},
76
+ eprint={2502.11824},
77
+ archivePrefix={arXiv},
78
+ primaryClass={cs.CL},
79
+ url={https://arxiv.org/abs/2502.11824},
80
+ }
81
+ ```
82
+