Searching for Better Game Recommendations with Jev
Turkish version: Medium I run likethisgame.com . Every game has a page, and that page lists other games someone who liked it might also like. A language model writes those recommendations and I store them. When I looked at the database, I found a problem. Black Myth: Wukong was on one in every ten pages that had recommendations. It had been recommended for a music game, for a sports game, for a…
The present game recommendation system listed similar games based on commonalities and user ratings. However, this caused common popular games to be recommended on unrelated pages. To address this, Jev, a decision model from TypeSafe, was employed. Jev is fed a game and its candidates, and returns probabilities indicating the likelihood of fans of the original game also liking the candidates.
By using Jev, the new system ranks games before the other model generates recommendations, resulting in more accurate recommendations and reducing the cost.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
