Change Details

===Brief summary The Research team is working on improving the Machine Learning based tools to support Wikipedia Patrollers. As part of this effort we are developing a new model to detect revisions that require patrollers' attention (T314385). The current model is based on implicit users’ feedback (revisions that have been reverted), and gives recommendations on which revisions are likely to be reverted. To improve this model and to make it easier to use its output, we want to build a web app that allows users to 1) rate the quality of the recommendation and 2) directly revert edits based on these recommendations. Inspired by the [[https://speedpatrolling.toolforge.org/ | SpeedPatrolling Tool, ]] we want to build an app that allows users to give explicit feedback on our recommendations as well as allow them to directly revert revisions when needed. The app should be able to connect with our API to pull the recommendations and show them to the users. Also it should save the users feedback allowing to retrain/finetune our existing model. ===Skills required - Intermediate JavaScript - HTML & CSS - Familiarity with Flask - Design/UX skills welcome but not required - Experience with Sklearn welcome but not required ===Mentor(s) @MunizaA @diego ===Microtasks https://public.paws.wmcloud.org/User:Diego_(WMF)/MicroTaskOutReachyRoun25-Patrolling.ipynb

===Brief summary The Research team is working on improving the Machine Learning based tools to support Wikipedia Patrollers. As part of this effort we are developing a new model to detect revisions that require patrollers' attention (T314385). The current model is based on implicit users’ feedback (revisions that have been reverted), and gives recommendations on which revisions are likely to be reverted. To improve this model and to make it easier to use its output, we want to build a web app that allows users to 1) rate the quality of the recommendation and 2) directly revert edits based on these recommendations. Inspired by the [[https://speedpatrolling.toolforge.org/ | SpeedPatrolling Tool, ]] we want to build an app that allows users to give explicit feedback on our recommendations as well as allow them to directly revert revisions when needed. The app should be able to connect with our API to pull the recommendations and show them to the users. Also it should save the users feedback allowing to retrain/finetune our existing model. ===Skills required - Intermediate JavaScript - HTML & CSS - Familiarity with Flask - Design/UX skills welcome but not required - Experience with Sklearn welcome but not required ===Mentor(s) @MunizaA @diego ===Microtasks TBA

===Brief summary The Research team is working on improving the Machine Learning based tools to support Wikipedia Patrollers. As part of this effort we are developing a new model to detect revisions that require patrollers' attention (T314385). The current model is based on implicit users’ feedback (revisions that have been reverted), and gives recommendations on which revisions are likely to be reverted. To improve this model and to make it easier to use its output, we want to build a web app that allows users to 1) rate the quality of the recommendation and 2) directly revert edits based on these recommendations. Inspired by the [[https://speedpatrolling.toolforge.org/ | SpeedPatrolling Tool, ]] we want to build an app that allows users to give explicit feedback on our recommendations as well as allow them to directly revert revisions when needed. The app should be able to connect with our API to pull the recommendations and show them to the users. Also it should save the users feedback allowing to retrain/finetune our existing model. ===Skills required - Intermediate JavaScript - HTML & CSS - Familiarity with Flask - Design/UX skills welcome but not required - Experience with Sklearn welcome but not required ===Mentor(s) @MunizaA @diego ===Microtasks https://public.paws.wmcloud.org/User:Diego_(WMF)/MicroTaskOutReachyRoun25-Patrolling.ipynbTBA