 to each and every one around the world. So, Alain has already been said about, with regards to machine learning and AI, then I'm continuing with the, a minute. Okay. So I'll be continuing with RS and left wing, which is also a content quality two that was developed by the machine learning team back in 2015. So I'll first start with RS, why it has been deprecated, then move on onto why we are moving most of the models, all the models from RS on to left wing at large. Okay. So first of all, with people who are new to machine learning tools with regards to content quality, improving content quality on a Wikimedia project at large, RS is basically a two, a machine learning two that's helps to improve the content quality on Wikipedia and other Wikimedia projects at large. It helps most of the contributors to be able to detect vandalism, to be able to identify good edits, quality articles and to be able to also manage their workflows at large within the most of the Wikimedia projects. So why RS? It helps most contributors to be able to streamline their workflows in terms of being able to aggregate certain kind of articles that also help them to be able to know the kind of articles to be able to also work on or contribute to. With regards to new editors, it also helps experienced editors to be able to also identify new editors who are contributing to most of the Wikimedia projects that also be able to guide them, provide some form of guidance and some form of assistance to them in terms of they contributing to most of the Wikimedia projects, especially English Wikipedia. And it's also a community support driven, which means that it's open to most of the Wikimedia community as a whole. That also helps people to be able to communicate and work on discussions together, share ideas and know how best to be able to improve the two. Okay. So RS is basically used by the test community managers and also researchers around the world, especially researchers within the Wikimedia ecosystem. It is also a crowdsourcing content for Wikimedia projects such as Wikisnary and Wikidata at large. Then when was RS created? As earlier, I stated that it was created in the year 2015 by the machine learning team. And through year by year it has been able to improve the quality of articles or contents within most of the Wikimedia projects that has been using RS since 2015. Okay. So if there are any questions at the moment, if there isn't any question, I can be able to also move on. So a lot has been said already with regards to model names, models at large, how these models are being built to be able to also assist users to be able to identify articles and contents to be able to also improve upon them. So some of the model names are Vandalism Classifier, which is one of the models that is able to also help contributors who normally do use RS at large to be able to identify contents that involve vandalism and help them to be able to resolve such contents that involve vandalism. It also helps us to be able to classify or be able to identify articles that stand in good faith or contributors who contribute to articles in good faith. And if there are certain minor issues pertaining to their contributions to, we can be able to improve upon their content at large. Then article quality due to that is able to also help us to know the accuracy of contents that are being placed on most of the Wikimedia project, especially English Wikipedia. So with regards to access on RS, although it's been deprecated, but from time to time, I'll also take you through the process of being able to access a left wing at large, how RS is also being moved. The models on RS has been moved to left wing at large. So RS is available online for all and has its own website, which has been moved onto left wing. So at the moment, you can find RS on RS.wikimedia.org. And it can be accessed either through the webpage or any through the media wiki API at large. So why RS? So machine learning is basically a two. It's simply a machine learning, why left wing, sorry. Left wing is basically also a machine learning two that has been designed to improve upon how RS does work, although still under development, there are certain features that needs to be developed in order to be able to also improve upon the contents quality related to most Wikimedia project at large and to be able to also host multiple deployments in different languages and also help different languages to be able to also identify classifications that they do want to work on. So with regards to the differences between RS and left wing, left wing basically is a generic model as it was stated earlier within the first session before my session. So it being a generic model, it basically also helps us, it basically also helps us in terms of article revision as well with regards to RS, but it being a generic model, it's basically hosts multiple models and helps us to be able to expand the production, expand the production rights in terms of generating models, that can be host a lot of models that can be able to also help us in terms of generating scores in different Wikis at large with regards to languages as well. So with left wing, it being able to host models as microservices too, it can be able to also help us to aggregate multiple scores or many scores as compared to RS. And with regards to article revisions too, it can be able to enhance article revisions. So the content quality wise of article revisions improves from time to time with regards to using left wing. So it also helps us in terms of faster deployment and currently RS does has a pre-cache which basically helps us to be able to get updated with regards to revisions that have been made on most recent changes, but with a functionality regards regarding to a left wing, from time to time, that functionality too will be developed in order to be able to also contain that. So why RS has been changed from, why the changes from RS to left wing? So with RS, it will become much more accurate in terms of contributing to Wikimedia project at large, in terms of it accuracy, in terms of using its models to provide article quality, using its models to be able to identify vandalism, using its models to be able to identify coordinated edits and using multiple models within other Wikiproject at large. So left wing is basically more, will be more sustainable, has the potential being more sustainable in terms of going through the process of being able to update multiple models that have been built year by year, but it being using microservices will give it elasticity and will be much more sustainable in terms of using left wing. So with regards to accuracy, for left wing, Stady has already indicated that left wing was able to identify 80% of quality revisions that were being made, while RS was able to currently identify 70% of high quality revisions as compared to 80% of low quality revisions for RS at large. So in terms of scalability, RS can also be used in a number of ways in which revisions being done on left wing can also be used in a number of ways in which revisions being done on RS are not able to meet. Since Wikipedia keeps growing and technology keeps evolving, within the technological infrastructure of Wikimedia still needs to grow and there needs to be improvements in terms of technological infrastructure at large. That's why we've decided to be able to move on to left wing to be able to improve upon content quality within most of the Wikimedia projects. So as stated with sustainability, it also helps us to be able to update new data in order to be able to remain accurate. That was the job that RS was basically doing. So as time went on to coming up with a left wing, that would make it much more simpler since it being in a human in a loop system whereby both humans and us that can be able to blend together in terms of supervised machine learning and also active learning. Combined blending these two together to be able to be able to rely on data, to rely less on data in order to be able to improve upon contents based on this kind of blend between active learning and supervised machine learning, it will make it much more easier. It wouldn't only solely rely on new data to be able to be updated all the time, which is much more costive too and takes a lot of time to be able to deploy such solutions. So my call for this particular session is to be able to raise the expectations of Wikipedia content quality and also to be able to improve upon content quality so far. And we also do look forward for you guys being able to also follow the update and process procedures that have been implemented in order to be able to also help and contribute in terms of improving content quality within most of the Wikimedia wikis that we do find ourselves in. And as time goes on, as models are being generated in different languages from a good faith model to other different classifiers and models too can be able to also help us in terms of knowing how best we can be able to also improve upon our contents on most of our Wikimedia wikis. Now to be able to find most of the resources that I currently use for this particular session, you can be able to access the RS website to be able to go through the whole procedure then understand how this to serve a great purpose within the Wikimedia ecosystem aside that you can also, this is the RS page on Wikitech, same for left-wing, then if you want to try or have a few of how the RS is currently working, then you can be able to also try the special page on recent changes on English Wikipedia or if you are already an administrator on any of the wikis on English Wikipedia so you can just feel free then, try to go through the process of downloading and installing Hago, then know how best to be able to use these tools to be able to improve upon content quality on most of our Wikimedia wikis. Okay, thank you. If there are any questions, you can feel free and ask. Okay, so with regards to the models that I was talking about, currently RS was holding these kind of models that have been moved to left-wing and the machine learning team is currently also looking forward to be able to add more models in terms of left-wing being able to hold a number of large model-based in terms of having multiple models in different languages that can be able to also help improve upon most of the Wikimedia wikis at large. So we have the article quality model type, we have the draft model type, the damage-gen model type, which as we can see already, some of them are already in different language wikis, then we have the good faith model type that we can see in different language wikis at large. Then we have a model type for reverted, articles that can be reverted, having the reverted button to be able to revert contents that contributors do feel that does not meet the criteria needs to be reverted or improved upon. Same for article topic, then draft topic at large. Thank you very much. Okay, so with a last face, so with regards to having RS on most of these models, special wiki pages, we can be able to find that these models are being implemented at the back of most of our Wikimedia wikis at large and based on the model types that we do also find, we can be able to identify that as we are able to take any of the, select any of the models that we do want to find revisions related to these kind of models, then we can be able to identify models that are related to such, and we can be able to also find revisions that are related to such models. And based on these, it can be able to also help us to review or be able to go through some of the contents that are being updated or that are being published within our special revisions or revisions that we do find in our wikis. So as this too is currently being deprecated, I do know that from once the whole two has been, well once the whole left wing two has been deployed, it would make it much more easier for people to be able to also use the left wing two at large. Okay, so a template like this, this template can be reviewed based on the contributions that the person made, we can be able to also review such contribution then know if the contribution is of good faith, if it's not on good faith, if it's been identified manually as the vandalism contents, then we can be able to also select the vandalism option to be able to also rate the contents or the updates or the revision that has been made as the vandalism. If we want to roll it back based on your user rights that you do already have, you can be able to also roll it back, same for all the user rights that you also do have using RS, okay. So I would like to end here and have more time for yourselves to prepare for other sessions if there aren't any other questions at large. Thank you all very much for joining this session.