Change Details

Our article schema contain two model scores - goodfaith and damaging under //Article.Version.Scores//. We add //Article.Version.Scores.Damaging// and //Article.Version.Scores.Goodfaith// in our schema objects if the article belongs to one of the project supported by these models and if the score was available from eventstream or API call. As such, these fields are marked as optional in the schema. Goodfaith and damaging are planned to be deprecated in future from WMF. Instead, there is another model for "revision quality", revert-risk model. For our enterprise data products, we are planning to use [[ https://meta.wikimedia.org/wiki/Machine_learning_models/Proposed/Language-agnostic_revert_risk | revert-risk language-agnostic ]] model. As a part of this ticket, we need to: [] Double-check our customer contract around schema changes. Specifically, about removing goodfaith and damaging, and replacing with revert-risk score. [] Update our docs. Goodfaith and Damaging are going to be deprecated. Will be replaced by revert-risk language-agnostic model score. [] Send message to the customers about this, and gather their feedback.

Our article schema contain two model scores - goodfaith and damaging under //Article.Version.Scores//. We add //Article.Version.Scores.Damaging// and //Article.Version.Scores.Goodfaith// in our schema objects if the article belongs to one of the project supported by these models and if the score was available from eventstream or API call. As such, these fields are marked as optional in the schema. Goodfaith and damaging are planned to be deprecated in future from WMF. Instead, there is another model for "revision quality", revert-risk model. For our enterprise data products, we are planning to use [[ https://meta.wikimedia.org/wiki/Machine_learning_models/Proposed/Language-agnostic_revert_risk | revert-risk language-agnostic ]] model. As a part of this ticket, we need to: [x] Create a draft comms text for Amy [] Double-check our customer contract around schema changes. Specifically, about removing goodfaith and damaging, and replacing with revert-risk score. [] Update our docs. Goodfaith and Damaging are going to be deprecated. Will be replaced by revert-risk language-agnostic model score. [] Send message to the customers about this, and gather their feedback.

Our article schema contain two model scores - goodfaith and damaging under //Article.Version.Scores//. We add //Article.Version.Scores.Damaging// and //Article.Version.Scores.Goodfaith// in our schema objects if the article belongs to one of the project supported by these models and if the score was available from eventstream or API call. As such, these fields are marked as optional in the schema. Goodfaith and damaging are planned to be deprecated in future from WMF. Instead, there is another model for "revision quality", revert-risk model. For our enterprise data products, we are planning to use [[ https://meta.wikimedia.org/wiki/Machine_learning_models/Proposed/Language-agnostic_revert_risk | revert-risk language-agnostic ]] model. As a part of this ticket, we need to: [x] Create a draft comms text for Amy [] Double-check our customer contract around schema changes. Specifically, about removing goodfaith and damaging, and replacing with revert-risk score. [] Update our docs. Goodfaith and Damaging are going to be deprecated. Will be replaced by revert-risk language-agnostic model score. [] Send message to the customers about this, and gather their feedback.