Hades-7b
Model Details
Hades-7b is a sophisticated text generation AI model developed by 2digit specifically for the purpose of news analysis. Leveraging advanced natural language processing techniques, Hades-7b is capable of extracting a wide range of information from news articles. Key functionalities of this model include:
- Category Identification: Hades-7b can classify news articles into various predefined categories. This helps in organizing news content and makes it easier for users to find articles related to specific topics of interest.
- Company Name Extraction: The model can identify and extract the names of companies mentioned in news articles. This feature is particularly useful for financial analysis, where tracking mentions of companies in the media can provide insights into market sentiment and potential stock movements.
- Stock-Related Themes: Hades-7b is adept at recognizing themes and topics related to the stock market. This includes identifying news about market trends, investment strategies, regulatory changes, and other stock-related content. By categorizing news articles based on these themes, the model helps analysts and investors stay informed about relevant market developments.
- Keyword Extraction: The model can pinpoint key keywords and phrases within a news article. These keywords summarize the main points of the article, making it easier for users to quickly grasp the content without reading the entire text. This feature enhances the efficiency of news consumption, especially in fast-paced environments where time is of the essence.
Overall, Hades-7b is a powerful tool for anyone involved in news analysis, from financial analysts and journalists to market researchers and investors. By automating the extraction of critical information from news articles, Hades-7b streamlines the process of news analysis and helps users make more informed decisions based on up-to-date information.
License
Use of this model requires company approval. Please contact AI@2digit.io. For more details, please refer to the website below: https://2digit.io/#contactus
Dataset
The model was trained on an internal dataset from 2digit, consisting of 157k dataset.
| task | size | ratio | description |
|---|---|---|---|
| theme | 5,766 | 3.7% | Human-labeled theme stock dataset |
| keyword | 23,234 | 14.8% | Human-labeled main and related keyword datasets |
| category | 24,605 | 15.6% | Human labeling of 10 categories |
| stockname | 103,643 | 65.9% | Human-labeled stockname datasets |
Evaluation
We measured model accuracy through an internal evaluation system.
| task | accuracy | description |
|---|---|---|
| theme | 0.66 | Extract themes and related companies |
| keyword | 0.40 | Extract keywords and keyword type |
| category | 0.88 | News category classification |
| stockname | 0.90 | Extract companies |
- Downloads last month
- 386