In previous posts we have largely discussed the capabilities of our weather insights platform. In this post, we will take a look at our efforts in surfacing these insights in tools that will help data driven climate journalism.
Lead Generator for Breaking News
We applied Generative AI models to the weather patterns discovered by our insights platform to generate 'leads' that can be used by climate news desks. At a high level, generating leads for weather news involves these steps.
Detection
We continuously detect abnormal weather patterns around the world using techniques described in earlier blog posts.
Snippeting
We subsequently convert the detected anomalous weather events into a human friendly message with the appropriate level of detail. We achieve this by using a combination of information retrieval and Generative AI ('RAG') to generate snippets that capture the significance of the event while preserving the links to the backing data.
Join the slack community for a preview of abnormal weather pattern alerts.
Delivery
The leads are available as soon as they are generated. We send push notifications to enterprise chat platforms (currently Slack). From those notifications, it is possible to navigate to details about the event. We believe that addressing immediacy concerns of the news cycle without sacrificing appropriate data context is an important step in quickly publishing trustworthy information.
The leads are also accessible through an interactive webpage.
Storify - Easily generate factual climate content
We have combined Generative AI models with our weather insights to synthesize textual and graphical content that are reader friendly. The content generated is factual and individual pieces of content have links to their backing data.
Bulletin Generation
When there are a series of breaking news snippets, it is useful to be able to generate a summary while preserving references to representative context. Currently, bulletins can be generated on demand at a country level. These pieces of content can be used in tweets, start articles or provided as automated feeds to other news management systems.
Watch this 1-minute demo of a possible breaking news scenario.
Report Generation (Under Development).
We are also working on the ability to automatically generate a ‘state of climate’ synopsis that reads like a ‘quarterly financial’ report for the specified region or city. The report will contain a mix of synthesized text, charts and graphs.
Data Augmenter
We have built a prototype that suggests data snippets to enhance a weather story. We are using a combination of ML and Generative AI to understand existing articles that reference weather events to suggest additional relevant data context.
The goal is to improve climate story that typically describes a weather related story at a given time and place. What might be missing is the data context. For example, how sweltering was it in Rio during the Taylor Swift concert and how unusual was that in early summer? For some kinds of events, we suggest links that contain appropriate visualizations. In others, we directly retrieve a data snippet from our insights platform.
We have added an excerpt from this article. The augmenter suggests data and links that add data context. For example, it surfaces humid heat contexts and provides a link to look at typical weather for the Rio de Janeiro area during the time period.
Take a look at the 1 minute video for another example.
We are planning to suggest these additions by annotating the exact part of the story where the augmentation is appropriate. We are also looking at ways to verify facts/data already mentioned in the story. Watch this space for the improvements.
If you have questions about the tools or are curious to learn more about how we are building them, please do not hesitate to talk to us.
In order to improve both the data quality and the product , we are seeking to expand our team. Ideally, we are looking for a full time scientist co-founder . We are also seeking investment to help us hire and scale our climate communication offerings. Please take a look at our pitch deck to see if there is a fit.