Medical Institution Reviews with Sentiment Annotations

Featuring 12,036 user reviews about medical institutions. Each review is tagged with a general sentiment and five aspect-specific sentiments: quality, service, equipment, food, and location.

Use Cases

• Train a multi-task model to predict the general sentiment and the five aspect sentiments (quality, service, equipment, food, location) from the review content.
• Analyze correlations between the general sentiment score and specific aspect scores like quality or service to identify primary drivers of user satisfaction.
• Perform aspect-based sentiment analysis by segmenting review content based on mentions related to equipment or food and comparing against the corresponding sentiment labels.
• Build a text classifier to detect reviews where the location sentiment score significantly deviates from the general sentiment, indicating isolated issues.

Strengths

• Contains 12,036 labeled reviews.
• Provides sentiment annotations for one general and five specific aspects (quality, service, equipment, food, location).
• Data is structured with a review_id and content field for each sample.

Limitations

• The dataset size of 12,036 reviews may be insufficient for training large, complex language models without augmentation.
• No information is provided on label quality, potential class imbalance, or annotation consistency for the six sentiment fields.
• Geographic and linguistic scope is unclear, potentially limiting generalizability.

Provenance

Source

Hugging Face dataset by author 'blinoff'.

Collection Method

User reviews about medical institutions, method of collection unspecified.

Freshness

Last updated on 2022-10-23.