STUFF

Disruption Department

More of a prototype than an experiment, trying to optimize a previously developed system.

The goal is to produce and publish more and better content in a shorter time. The key factor to reach this, is the automatzied tanslation of text content, meta data of articles and images. The system uses DeepL's API for translations.

The strategy to improve the content, is to summarize peer-reviewd research papers pubished on PLOS. This should ensure a high quality of source of the content.

Depending on the user and internet connection, the process of creating a article, translated to 29 langues, takes 7 to 15 minutes.

Summary
URLthedisruptiondepartment.org
TypeNews site, multingual
Domain historyExpired domain; mixed history
TimeAugust - November 2023
StatusTerminated: observing, no further development, no new content.
Learnings
  • Search performance is below expectations. Maybe because pure summaries without adding new information aren't considered valuable information by search engines. Another possibility is the partially spammy history of the domain or low search volume for scientific niche topics.
  • Markdown (.md) format can be messed up when translated to certain languages by DeepL. Therefore, HTML input seems to be a better choice for lists and tables.
  • Further improvements are needed. The time needed to create content is still too long to use the system for a long-term experiment. The whole process (selecting a topic, creating text content and metadata with LLM, selecting and uploading images with metadata and translations) shouldn't take longer than 5 minutes.