Financial News Sentiment is a dataset of approximately 2000 manually validated sentiment labels for Canadian news articles. The dataset also includes a topic column with values such as acquisition, quarterly financial release, and dividend, which was generated automatically using a zero-shot classification model. The dataset was created by Jean-Baptiste and last updated on December 29, —2022.
Use Cases
- Training sentiment classifiers for financial news based on the manually validated sentiment labels.
- Analyzing the distribution of corporate news topics like acquisitions and quarterly releases based on the provided topic column.
- Benchmarking zero-shot classification models against a manually reviewed sentiment ground truth.
- Studying the relationship between news topics and sentiment polarity within Canadian financial media.
Strengths
- Contains approximately 2000 data points with manually validated sentiment labels.
- Includes a topic column with 10 distinct categories relevant to corporate finance.
- Provides a clear distinction between manually reviewed sentiment and automatically generated topic labels.
Limitations
- The topic labels were generated automatically and were not reviewed manually, which may affect their accuracy.
- Row count, column details, and file formats are unknown, limiting suitability assessment.
- Last updated 2022-12-29 03:14:44; freshness should be verified.
Provenance
- Source
- Jean-Baptiste via Hugging Face.
- Collection Method
- Sentiment labels were manually validated; topics were generated automatically using a zero-shot classification model.
- Time Range
- null
- Freshness
- Last updated 2022-12-29 03:14:44.
- Geography
- The dataset is based on Canadian news articles.