For TTS V1, we need to pre-generate audio files for 7,800 articles across 39 topics. This ticket covers identifying and compiling the article list; the Apps team will review the article list and provide feedback if we need to recompute or if individual items should be substituted.
###Article composition
1,950 seed articles (50 per topic × 39 topics)
5,850 recommended articles (3 MoreLike recommendations per seed article)
###Topics
architecture
visual-arts
comics-and-anime
entertainment
fashion
books
music
performing-arts
sports
films
video-games
Biography
women
business-and-economics
education
food-and-drink
history
military-and-warfare
philosophy-and-religion
politics-and-government
society
transportation
biology
chemistry
internet-culture
geographical
engineering
stem
mathematics
medicine-and-health
physics
technology
africa
asia
central-america
europe
north-america
oceania
south-america
###Seed article selection process (per topic):
# Retrieve all Featured Articles and articles in the 70th–90th percentile of pageviews for the topic
# Filter out articles with quality score < 0.85
# Filter out articles with # of langlinks < 10
# Filter out articles whose lead section contains any of the designated keywords (list to be shared in comments)
# Rank remaining articles by topic score and take the top 50
###Recommended article selection process:
For each seed article, pull the top 3 results from the MoreLike API
###Deliverable
A list of 7,800 articles with the following columns:
- Topic
- Article title
- Article URL
- Type (Seed or Recommendation)