みなさんこんにちは。私はトモキータダです。Dvpua搜集になることをお話しするために、ウイキペディアーと私たちの常識のアーティカルを探すのです。ウイキペディアーは、遠くにいかないアーティカルがある。ウイキペディアーのダウンボードの他に、カバーライン・オディナーリード・インサイクロペディアーのアーティカルを探しています。この一部は、不常識のアーティカルのダウンボードです。不常的なアーティクルを見るのは難しいです。アーティクルは不常的なアーティクルを見るのは難しいです。不常的なアーティクルを見るのは難しいです。まずはGavel is Home.Gavel is Homeはジューメトレックのフィギュアルがインフィニットサフェイスエリアやフィニットボリュームがあります。2. Beard Liberation Float.The Beard Liberation Float is a British Interest Groupwhich campaign in support of Beardsand opposes discrimination against those who wear them.That's Madonna.Madonna is a restaurant located in Iraq.It is designed to resemble the fast food chain McDonaldsboth in appearance and in menu.You may think they are funny articles.The research purpose.The problem we addressed is that the number of unusual articles is small. Our aim is to develop methods for discovering articlesthat can be added to the unusual articles sectionand thereby expand it.Achieving this aim will lead to an improvement in Wikipedia's originalitybecause Wikipedia contains articleson things not covered in ordinary encyclopedias.Also, this goal will lead to automation of adding articles.It is difficult for people to find unusual articlesamong the enormous number of articles on Wikipedia.Therefore, we define an unusual score for an article.There are two main types of information used in the research area of Wikipediacontent information and editor information.Content information is described from a neutral point of viewwhile editor's information is reflectededitors' preferences and knowledge.Also, editor's information is easy to define a scoreby counting the number of editors for an article.So, editor's information is useful to find unusual articles.Here we make a hypothesis based on collaborative filtering.We hypothesis that editors of unusual articlesprefer an elite funny articlessimilar to the unusual articles.This research will verify this hypothesisusing editorial information.The hypothesis is illustrated in this figure.If there are multiple editors who have edited the unusual articles,we can hypothesis that they prefer an elite articlesthat can be added to the unusual articles section.I will explain how to search a candidate's articlesto be added to the unusual articles section.It is the earlier indicated by the red box.We define the article's importance scoreand the editor's importance scoreto evaluate how unusual an article is.The editor's importance score editoris a value that evaluateshow many the editor edits unusual articles.I will explain in detail later.The importance of an article score articleis the sum of the importance scores of editorswho have edited unusual articles.The higher the article's importance score,the more likely it is to be a good articleto add to unusual articles section.Next,I will explain how to evaluateeditors who have edited unusual articles.It is the earlier indicated by the red box.The editor's importance score editorshows how many unusual articles an editor has edited.We divide three methods.Norm-weighted method,Count method,Lation method.We use damped data from English Wikipediaand the editor's damped data.We show table of statisticson the number of editors per article.From this table,we can understandmaximum of all articles is very largeand million in all articles is very small.We conducted an experimentto evaluate score editorand score article automatically.We separated unusual articlesinto training data and test data.For the training data,we calculatedthe article's importance scoreand generated a ranking of articlesthat could be added to the unusual articles section.On the basis of precision at Kand recall at K,we evaluatedthe degree to which there weretest data in the top K cases.We show result of automatic evaluation.The value shown in the red is the highest value.We can understand the relation methodachieve the best performance.We conducted human evaluation.We calculated score article for articles.We extracted 30 articles with high score article.I judges whether they could be addedto the unusual articles section.We show result of human evaluation.We can understand the relation methodachieve the best performance.I will show you some examples ofextracted funny articles.First,intekar.Intekar is an esoteric programming language.Second,golden urges.Golden urges are the simple of McDonald's.Last,measure league cutball.Measure league cutball is an amateurteam sport inspired bythe fictional gameKuditch in the Harry Potter books.Conclusion.The use of editor information based on collaborative filteringis effective for discovering unusual articles.From pages representing preferences,we can find pages with similar preferencesby tracking the editors.It is possible to apply to a group of editorswith specific preferences.Thank you for listening.