tf-idfを用いてWEBページの特徴語を抽出する
tf-idf
まぁ、やってることはこちらでやってることと何一つ変わらない。
ただ、対象テキストがWEBページってだけ。
なので内容については上記リンク先で。
http://ja.wikipedia.org/wiki/Tf-idfを引数にしてテスト
テスト結果
Array ( [ノート] => 70.550270786977 [idf] => 58.242566281241 [トーク] => 39.260016066399 [collapsiblenav] => 37.198972183809 [editwarning] => 37.116381689581 [wgNotice] => 36.981332921722 [tf] => 29.857403204274 [単語] => 27.169254869112 [simplesearch] => 26.690732627211 [runOnloadHook] => 25.982670229996 [false] => 23.308651699028 [vector] => 21.06564049106 [ドキュメント] => 20.848558681228 [アルゴリズム] => 19.467134832372 [出現] => 19.060888352292 [wikipedia] => 18.973873319242 [mediawiki] => 18.968250771126 [ページ] => 17.851484108862 [利用] => 17.851484108862 [ファイル] => 16.863049624051 [言語] => 16.732703103715 [頻度] => 15.989424100918 [true] => 15.620048595254 [表示] => 15.620048595254 [検索] => 15.620048595254 [Tf] => 14.928701602137 [ja] => 14.348497840204 [disablesuggest] => 14.32633618173 [isMSIE] => 13.825560893818 [Wikipedia] => 13.628525681705 [wgWikimediaMobileUrl] => 13.580438435785 [expandablesearch] => 13.497056826846 [wgVectorPreferences] => 13.457836113692 [wgVectorEnabledModules] => 13.457836113692 [footercleanup] => 13.454345708753 [collapsibletabs] => 13.450867444376 [wgCategories] => 13.443947001532 [wgNamespaceIds] => 13.426852568172 [wgSearchNamespaces] => 13.426852568172 [wgMWSuggestMessages] => 13.423468470188 [wgMainPageTitle] => 13.42009578571 [wgFormattedNamespaces] => 13.416734438007 [wgMWSuggestTemplate] => 13.416734438007 [wgDigitTransformTable] => 13.413384351122 [wgSeparatorTransformTable] => 13.413384351122 [wgRestrictionMove] => 13.410045449856 [wgRestrictionEdit] => 13.410045449856 [wgActionPaths] => 13.39024282256 [wgVariantArticlePath] => 13.39024282256 [wgNoticeLocal] => 13.39024282256 [w] => 13.388613081647 [http] => 13.388613081647 [wgBreakFrames] => 13.386980176925 [wgCurRevisionId] => 13.383728141539 [wgEnableWriteAPI] => 13.383728141539 [wgEnableAPI] => 13.377255627033 [wgContentLanguage] => 13.374035012333 [wgUserLanguage] => 13.367624733972 [wgIsArticle] => 13.364434938604 [wgUserGroups] => 13.364434938604 [wgArticleId] => 13.361255285687 [wgCanonicalNamespace] => 13.345506928719 [wgSiteName] => 13.339276378968 [wgUserName] => 13.330002742183 [wgNamespaceNumber] => 13.302686931536 [wgPageName] => 13.27901718745 [MediaWiki] => 12.651079318259 [wgVersion] => 12.611037878597 [wgDBname] => 12.608043864384 [wgAction] => 12.508530003564 [wgScriptExtension] => 12.508530003564 [wgArticlePath] => 12.392402223722 [wgScriptPath] => 12.367415795939 [addMessages] => 12.281874358946 [wgTitle] => 12.254213020858 [wgUrlProtocols] => 12.234472120052 [wgServer] => 12.119894949289 [wgScript] => 12.023751088736 [jawiki] => 11.699255043162 [fixalpha] => 11.612745800252 [org] => 11.157177568039 [者] => 11.157177568039 [stylepath] => 11.052341057641 [失わ] => 10.85799949723 [writeln] => 8.9992694018962 [ヘルプ] => 8.925742054431 [searchTerms] => 8.804875263868 [worldwind] => 8.7156441301401 [テンプレート] => 8.6324777882718 [ウィキペディア] => 8.415335741836 [enable] => 8.3906442444265 [null] => 8.3209687294533 [候補] => 8.2578720139834 [語] => 8.2001040272595 [みなさ] => 8.0805524658396 [特別] => 7.9141734878465 [Frequency] => 7.7122307469797 [opensearch] => 7.6930177484499 [会話] => 7.5272060006197 [重要] => 7.3255414809827 [nntp] => 7.1821921246839 [編集] => 6.9077552789821 [wiki] => 6.9077552789821 [処理] => 6.8684744714807 [gopher] => 6.8589651148127 [カテゴリ] => 6.6943065408233 [用] => 6.6943065408233 [php] => 6.6943065408233 [更新] => 6.6943065408233 [x] => 6.6943065408233 [if] => 6.6943065408233 [設定] => 6.6943065408233 [個人] => 6.6943065408233 [portal] => 6.6943065408233 [ウィキメディア・コモンズ] => 6.6608952010506 [ツール] => 6.6099570478432 [変種] => 6.5929445391421 [項目] => 6.4730615202138 [wmf] => 6.3683421983642 [メディア] => 6.269988681775 [離れる] => 6.2297217362322 [メイン] => 6.2222862501306 [行っ] => 6.2021855784236 [井戸端] => 6.1610673314942 [クリエイティブ・コモンズ] => 6.1422874368426 [要約] => 6.1057536935101 [telnet] => 6.0469772570382 [template] => 5.9954685524493 [Template] => 5.9934655477751 [果たす] => 5.92567680657 [タブ] => 5.6365912251159 [語句] => 5.6296030764819 [namespace] => 5.6034847533408 [下がり] => 5.5016582905661 [Inverse] => 5.4103668703567 [svn] => 5.3506638469202 [出典] => 5.3144467484779 [一種] => 5.2883670356949 [潜在] => 5.2785147392519 [継承] => 5.2687585643065 [寄付] => 5.212139670307 [mailto] => 5.1939573512238 [バグ] => 5.1585554241729 [抽出] => 5.1243640594246 [コミュニティ・ポータル] => 5.0436751481745 [指標] => 4.954727662158 [namespaces] => 4.9476604949349 [mw] => 4.8730496311437 [usability] => 4.8472417471878 [自然] => 4.8158912173037 [フィルタ] => 4.733003557498 [window] => 4.6764245412744 [役割] => 4.6407973636312 [ライセンス] => 4.5516294190601 [事典] => 4.5421953868267 [ナビゲーション] => 4.528209144852 [二つ] => 4.528209144852 [おまかせ] => 4.4785375350547 [文章] => 4.4628710272155 [中] => 4.4628710272155 [画像] => 4.4628710272155 [的] => 4.4628710272155 [help] => 4.4628710272155 [ログイン] => 4.4628710272155 [場合] => 4.4628710272155 [i] => 4.4628710272155 [category] => 4.4628710272155 [数] => 4.4628710272155 [作成] => 4.4628710272155 [最近] => 4.4628710272155 [リンク] => 4.4628710272155 [関連] => 4.4628710272155 [search] => 4.4628710272155 [image] => 4.4628710272155 [一般] => 4.4628710272155 [他] => 4.4628710272155 [度] => 4.4628710272155 [Category] => 4.4628710272155 [Portal] => 4.4628710272155 [Help] => 4.4628710272155 [l] => 4.4628710272155 [https] => 4.4567501808698 [働き] => 4.4270240006046 [警告] => 4.3583101080566 [irc] => 4.3389671452134 [上げる] => 4.2097554137334 [nbsp] => 4.1135273816389 [分野] => 4.0427013290703 [wikimedia] => 4.0090847182713 [ロード] => 3.9685933569165 [skins] => 3.9606881774094 [テキスト] => 3.9580669439296 [bits] => 3.9528449999484 [出来事] => 3.9450628595063 [アカウント] => 3.9347799925508 [適用] => 3.8397023438485 [計算] => 3.7550192566185 [注] => 3.7550192566185 [バージョン] => 3.6989297899674 [ボックス] => 3.6749765489449 [ftp] => 3.6690768268178 [UTC] => 3.6651629274966 [ポータル] => 3.6287255312942 [containing] => 3.6193533914653 [解析] => 3.545645286028 [var] => 3.4721564706026 [練習] => 3.4705474597969 [参照] => 3.4561816899949 [百科] => 3.4373428282985 [特徴] => 3.4326880487535 [固定] => 3.3726099248102 [全文] => 3.2794223247362 [移動] => 3.1349943408875 [含む] => 3.1315519966965 [日時] => 3.1089009716596 [空間] => 3.0977649991191 [印刷] => 3.0565444129507 [閲覧] => 3.0554822779598 [Polski] => 2.9977342762247 [免責] => 2.9469421093846 [warning] => 2.9469421093846 [履歴] => 2.9095545773129 [Русский] => 2.9004220937497 [しまう] => 2.8735146408297 [主] => 2.8647040111476 [文] => 2.8387285247443 [引用] => 2.8051119139453 [ni] => 2.7887181041697 [最終] => 2.7725887222398 [意味] => 2.7567153730835 [suggest] => 2.7488721956225 [多く] => 2.7333680090865 [フリー] => 2.7333680090865 [報告] => 2.7181005369557 [操作] => 2.6521425691639 [アップ] => 2.5319982573219 [api] => 2.5010360317179 [document] => 2.471003744619 [Document] => 2.471003744619 [特定] => 2.4418471603276 [事項] => 2.41911890925 [状況] => 2.3859667019331 [条件] => 2.3538783873816 [変更] => 2.3126354288475 [Term] => 2.2778924804037 [skin] => 2.2633643798408 [Deutsch] => 2.2314355136078 [版] => 2.2314355136078 [English] => 2.2314355136078 [Fran〓ais] => 2.2314355136078 [年] => 2.2314355136078 [すべて] => 2.2314355136078 [一覧] => 2.2314355136078 [o] => 2.2314355136078 [g] => 2.2314355136078 [案内] => 2.2314355136078 [action] => 2.2314355136078 [お知らせ] => 2.2314355136078 [問い合わせ] => 2.2314355136078 [月] => 2.2314355136078 [日] => 2.2314355136078 [more] => 2.2314355136078 [規約] => 2.2314355136078 [ください] => 2.2314355136078 [プライバシー] => 2.2314355136078 [ポリシー] => 2.2314355136078 [詳細] => 2.2314355136078 [m] => 2.2314355136078 [あなた] => 2.2314355136078 [木] => 2.2314355136078 [下] => 2.2314355136078 [可能] => 2.2314355136078 [追加] => 2.2314355136078 [index] => 2.2314355136078 [view] => 2.2314355136078 [D] => 2.2314355136078 [機能] => 2.2314355136078 [C] => 2.2314355136078 [news] => 2.2314355136078 [talk] => 2.2314355136078 [名前] => 2.2314355136078 [情報] => 2.2314355136078 [":] => -INF [},] => -INF [':'] => -INF [apsibleNavForceNewVersion] => -INF ["-] => -INF [\\:|] => -INF [,] => -INF [=".] => -INF [="",] => -INF [:] => -INF [;:] => -INF [],] => -INF [={"-] => -INF [);] => -INF [gCollapsibleNavBucketTest] => -INF ["",] => -INF [[] => -INF [}\] => -INF ["] => -INF [)] => -INF ["],] => -INF [://] => -INF [|] => -INF [=] => -INF [;] => -INF [?] => -INF [\] => -INF ['')] => -INF [={"] => -INF [!=] => -INF [="";] => -INF [\\:\\/\\/",] => -INF []] => -INF [(] => -INF [",] => -INF ["":] => -INF [/] => -INF ['});] => -INF [="] => -INF [&] => -INF [=["] => -INF [}},] => -INF [','] => -INF ["},] => -INF [&#] => -INF [gCanonicalSpecialPageName] => -INF [{"] => -INF [""],] => -INF [={] => -INF [();] => -INF [=[],] => -INF [=["",] => -INF [.] => -INF [=[] => -INF [({'] => -INF [/$] => -INF [\\:\\/\\/|] => -INF [={},] => -INF [="/] => -INF [_] => -INF [-] => -INF ) -- time:5 sec
- INFの原因はdf値が0(つまり、Yahooでの検索結果が0または取得できなかった場合)なので、スルー
Webページによってこういう特徴あるだろうから、上限の閾値も設けないといけない。
その辺の判定はシビアかもなー
tfとTfを別物として判定してるので、それぞれの値が低くなってる。
大文字小文字区別しなければここら辺は改善できそう。