Implementing an app with Streamlit for computer vision using OpenAI Vision API. The goal is to compute Sentiment Analysis of newspapers pages from images. Note that it is not an OCR problem, the API does not read the full documents, but only titles and some other highlighted words. The idea is to get a score from -1 to 1, where -1 is extremely negative and 1 extremely positive, for some user-defined categories taking into account only most catchy titles and words.
Click here to test it yourself!
An image is given as input and a downloadable dataframe is returned as output.
| Infos | Values |
|---|---|
| date | 13-11-2023 |
| newspaper | la Repubbiica |
| War & Peace | -0.7 |
| Economic Stability | -0.5 |
| Technological Trends | 0 |
| Social Movements | -0.5 |
| Health & Epidemics | 0 |
| Political Dynamics | -0.8 |
| Cultural Shifts | 0 |
| Global Relations | -0.3 |
| Scientific Progress | 0 |
| Education & Knowledge | 0 |
| Religious Landscape | 0 |
| Migrations & Immigration | 0 |
| Government Policies | -0.8 |
| Natural Disasters | 0 |
| Local Community Events | 0 |