An Evaluation of ChatGPT on Information Extraction task, including Named Entity Recognition (NER), Relation Extraction (RE), Event Extraction (EE) and Aspect-based Sentiment Analysis (ABSA).
In progress
- Language: English
- Domain: General & Biomedical
- Setup: zero-shot & few-shot
[
{
"seq": "SOCCER - JAPAN GET LUCKY WIN , CHINA IN SURPRISE DEFEAT .",
"entities": [
{
"e_name": "JAPAN",
"e_type": "LOC",
"start": 2,
"end": 3,
},
{
"e_name": "CHINA",
"e_type": "PER",
"start": 7,
"end": 8,
}
]
},
...
]
Q: what does the 'aspect' term in Aspect-based Sentiment Analysis task refer to? Explain in one sentence.
A: In Aspect-based Sentiment Analysis task, the term 'aspect' refers to a specific feature, attribute, or aspect of a product or service that a user may express an opinion about.
Q: what does the 'opinion' term in Aspect-based Sentiment Analysis task refer to? Explain in one sentence.
A: In Aspect-based Sentiment Analysis task, the term 'opinion' refers to the sentiment or attitude expressed by a user towards a particular aspect or feature of a product or service.
Q: what does the 'sentiment polarity' term in Aspect-based Sentiment Analysis task refer to? Explain in one sentence.
A: In Aspect-based Sentiment Analysis task, the term 'sentiment polarity' refers to the degree of positivity, negativity or neutrality expressed in the opinion towards a particular aspect or feature of a product or service.
prompts:
According to the following definition:
The term 'aspect' refers to a specific feature, attribute, or aspect of a product or service that a user may express an opinion about.
The term 'opinion' refers to the sentiment or attitude expressed by a user towards a particular aspect or feature of a product or service.
The term 'sentiment polarity' refers to the degree of positivity, negativity or neutrality expressed in the opinion towards a particular aspect or feature of a product or service.
Recognize all aspects terms with their corresponding opinion terms and sentiment polarity in the following reviews in the format of <aspect, sentiment_polarity, opinion>:
Boot time is super fast , around anywhere from 35 seconds to 1 minute .
- D17: 14lap, 14res, 15res (wang)
- D19: 14lap, 14res, 15res, 16res (fan)
- D20a: 14lap, 14res, 15res, 16res (penga)
- D20b: 14lap, 14res, 15res, 16res (pengb)
Recognize all aspect terms in the following review with the format ['aspect_1', 'aspect_2', ...]:
"Great food but the service was dreadful !"
output: ['food', 'service']
Recognize all opinion terms in the following review with the format ['opinion_1', 'opinion_2', ...]:
"Great food but the service was dreadful !"
output: ['Great', 'dreadful']
4.3 Aspect-level Sentiment Classification (ALSC): Predicting the sentiment polarities for every given aspect terms in a sentence.
Recognize the sentiment polarity for aspect term 'food' in the following review with the format ['aspect', 'sentiment']:
"Great food but the service was dreadful !"
output: ['food', 'positive']
4.4 Aspect-oriented Opinion Extraction (AOE): Extracting the paired opinion terms for every given aspect terms in a sentence.
Recognize the opinion term for aspect term 'food' in the following review with the format ['opinion_1', 'opinion_2', ...]:
"Great food but the service was dreadful !"
output: ['Great']
4.5 Aspect Term Extraction and Sentiment Classification (AESC): Extracting the aspect terms as well as the corresponding sentiment polarities simultaneously.
Recognize all aspect terms with their corresponding sentiment polarity in the following review with the format ['aspect', 'sentiment_polarity']:
"Great food but the service was dreadful !"
output: ['food', 'positive']
['service', 'negative']
4.6 Pair Extraction (Pair): Extracting the aspect terms as well as the corresponding opinion terms simultaneously.
Recognize all aspect terms with their corresponding opinion terms in the following review with the format ['aspect', 'opinion']:
"Great food but the service was dreadful !"
output: ['food', 'great']
['service', 'dreadful']
4.7 Triplet Extraction (Triplet): Extracting all aspects terms with their corresponding opinion terms and sentiment polarity simultaneously.
Recognize all aspect terms with their corresponding opinion terms and sentiment polarity in the following review with the format ['aspect', 'sentiment', 'opinion']:
"Great food but the service was dreadful !"
output: ['food', 'positive', 'great']
['service', 'negative', 'dreadful']