0
Ref 寫法統一<br />標題大小寫<br />置左格式統一<br />流程圖圖形選用、圖形&箭頭說明<br />公式符號說明<br />投影片-老師的建議<br />
[1]<br />修改動機:<br />修改目的(跟本來的技術比較,有何不同)<br />嚴謹的寫出Algo(照著實作,必須做得出來)<br />[2]<br />   Click次數如何搜集?<br />預測有幾種方法?<br />推薦評分值...
解決問題<br />理論上的<br />困難的應用問題<br />為什麼難?<br />貢獻在哪?<br />10/1老師給的建議(cont.)<br />
A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/10/01<br />
UMAP<br />finished *1<br />IUI<br />finished *1, finished but…*2 <br />Related Conference(10/1)<br />
User Modeling(UM, 1986-2007, 11th)<br />California<br />Adaptive Hypermedia and Adaptive Web-Based Systems(AH, 2000-2008, ...
Title :<br />[1]Construction of Ontology-Based User Model for Web Personalization (Cited: 9 times)<br />    H. Zhang, Y. S...
Authors: Hui Zhang, Yu Song, and Han-tao Song<br />Motivation: to provide web information that matches a user’s personal i...
Semantic Web Usage Log Preparation Model(SWULPM)<br />[1]How<br />
Steps:<br />1.S-Log(Semantic-log):representing the semantics of the respective URL(from domain ontology)<br />2.Session an...
         :look up<br />         :union <br />         :ontology<br />[1]How-Imagination of Ontology<br />Global User Ontol...
Each user has a graph:<br />C_Graph(N, u)=<N, A>, N: nodes, A:arcs, u:user<br />arc(s, t)=>label(s, t) = <dst, rst, hst, T...
Duration: 1997-2011<br />Title <br />[1]Personalized News Recommendation Based on Click Behavior (Cited: 2 times)<br />   ...
Authors: Jiahui Liu, Peter Dolan, ElinRonby Pedersen(Google Inc.)<br />Motivation: people was burdened with large online i...
Click behavior<br />advantage<br />no ratings or negative votes<br />after experiment (picture)<br />news interests do cha...
Prediction<br />User’s genuine interests<br />The influence of local news trend<br />Flow<br />predicting user’s genuine n...
[1]How(cont.)<br />: predicting user’s current news interest<br />      : current news trend<br />      : past time user’s...
Recommendation:<br />          (to rank a list of candidate articles)<br />CR(article): content-based recommendation score...
遇到的問題<br />Semantic network與Ontology表達能力;經過學習,建構出符合個別使用者的user model,並依照feedback或觀察使用者的行為,進一步update user model,每個步驟是否真的透徹了解...
A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/09/10<br />
Service Personalization<br />Early research<br />Overview of user-profile-based personalization<br />User Profile<br />Pur...
Early research<br />Personalization<br />
user-profile-based personalization <br />Overview<br />
Purpose<br />To record interest or habit of the user<br />To filter out irrelevant information from the user<br />To ident...
Type<br />Static<br />ex: name, age, country, education level<br />Dynamic<br />short-term <br />long-term<br />User Profi...
Process<br />1.Collecting information about users<br />user identification<br />user information collection<br />explicit<...
User identification<br />Software agents<br />Logins<br />Enhanced proxy servers<br />Cookies<br />Session ids <br />Colle...
User identification(?)<br />
Explicit<br />Providing personal information (My Yahoo![110])<br />Rating (Web pages, Syskill&Webert[68];Movie,           ...
Keyword Profiles<br />Amalthaea[61], Anatagonomy[78], Fab[5], Letizia[43], Syskill&Webert[68], PEA[60]<br />Semantic Netwo...
Extract from documents visited by the user during browsing<br />Web pages<br />Saved by the user<br />Provide by the user<...
To solve the synonymproblem<br />To solve the polysemy problem<br />                                                      ...
Semantic network profiles(cont.)<br />
More abstract topics (not specific words or sets of related words)<br />Concept profiles<br />
User Profile Construction<br />Building keyword profiles<br />Building semantic network profiles<br />Building concept pro...
A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/09/17<br />
GediminasAdomavicius , Alexander Tuzhilin, Using Data Mining Methods  	 to  Build Customer Profiles, Computer, v.34 n.2, p...
Validation operator<br />Similarity-based rule grouping<br />Template-based rule filtering<br />Redundant-rule elimination...
Building keyword profiles<br />Amalthaea[61]<br />WebMate[13]<br />Alipes[103]<br />User Profile Construction<br />
Building Keyword profiles<br />
Amalthaea’s Ecosystem[61] <br />
 Key V=(W1, W2, W3, …, Wn)<br />(待修改)Amalthaea’s Ecosystem[61](cont.)  <br />Web Pages<br />Stemmer<br />Html2txt filter<b...
WebMate: A personal agent[13]<br />    		 Chen, L., Sycara, K.: A Personal Agent for Browsing and Searching. In:          ...
Definition:<br />	1. Profile set V = { V1, V2,…,VN} <br />      (N domains of interest for each user)<br />	2. Document  D...
Algorithm for multi TF-IDF vector learning:<br />(待修改)WebMate[13](cont.)<br />User marked “I like It”<br />If |V| < N<br /...
Widyantoro, D.H., Yin, J., El Nasr, M., Yang, L., Zacchi, A., Yen, J.: Alipes: A 	Swift Messenger In Cyberspace. In: Proc....
Alipes[103]<br />
Coming soon…<br />Thank you for listening<br />Building semantic network profiles<br />
A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/10/22<br />
Authors<br />Susan Gauch, Jason Chaffee and Alexander Pretschner<br />Motivation<br />It’s impossible to use one approach ...
Reference ontology: Concept, Source<br />Concept<br />To extract top levels of the subject hierarchies (already existing)<...
How<br />
How<br />
How-Mapping(1)<br />
How-Mapping(2)<br />
Now the site can have its content browsed using the personal ontology<br />How-Mapping(3)<br />
Approaches<br />Re-ranking<br />Filtering(X)<br />How-Searching<br />
(1)To extract only html tags(titles, summaries)<br />(2)Classification<br />(3)examine documents which belongs to User’s c...
Thank you very much!!<br />
A Survey on Text Categorization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/11/02<br />
Classification<br />supervised learning<br />pre-defined categories<br />ex. credit of consumer<br />Clustering<br />unsup...
Motivation<br />With the rapid growth of online information, it is difficult and time-consuming to deal with or classify t...
SVM(Support Vector Machine) Vapnik 1995<br />kNN(k-nearest neighbor)<br />NB(Naïve Bayes)<br />LLSF(Linear Least Squares F...
-------以下為中文--------<br />
建立在最小化結構風險理論上<br />將資料根據特徵轉成Rn空間中的向量,每筆資料可視為空間中的一點,並從Rn空間中找到一個n-1維的界線,稱為分類超平面-H。<br />H1, H2為支援超平面(Support hyperplane)<br ...
多類別支援向量機<br />一對多<br />一對一<br />DAG(directed acyclic graph) method<br />Considering all data at once method<br />C&S metho...
SVM<br />
kNN(1)<br />
<ul><li>Based on XML</li></ul>Definition<br /><types>-data types<br /><message>-parameters of a function call<br /><portTy...
<message name="getTermRequest">  <part name="term" type="xs:string"/></message><message name="getTermResponse">  <part nam...
One-way<br />Request-response<br />Solicit-response<br />Notification<br />Operation Types<br />
Elements<br />Envelope (root)<br />Header<br />mustUnderstandattr.<br />actor attr.<br />encoding style attr.<br />Body<br...
A Survey on Text Categorization<br />Std. :Wei-Chen Chang<br />Prof.:Alan Liu<br />2010/11/18<br />
動機:資訊數位化,大量的電子文件分類,需耗費人力,且不客觀也缺乏一致性<br />目的:利用Ontology來協助分類,增加其準確性並節省人力、達到電子文件分類客觀與一致性<br />應用:線上即時新聞自動分類<br />做法:下頁<br />...
系統架構<br />基於Ontology架構之文件分類網路服務研究與建構<br />
Domain Weighted Ontology<br />以Object-Oriented Ontology為基礎,透過專家建構<br />   Domain Ontology(政治、社會、氣象、運動、財經)<br />訓練此Domain O...
分類機制<br />基於Ontology架構之文件分類網路服務研究與建構<br />
五階層式模糊推論機制<br />1.輸入層(Input Layer)<br />2.輸入語意層(Input Linguistic Layer)<br />3.規規層(Rule Layer)<br />4.輸出語意層(Ouput Linguist...
文字前處理網路服務<br />http://140.116.247.14/text_classification/text_classification.asmx?op=autotag<br />新聞分類網路服務<br />http://140...
透過專家定義好的Ontology,來完成分類工作,不需要花很長的訓練時間。<br />學習到透過訓練資料,如何給予Ontology中的概念與關係權重值。<br />Ontology如何轉成Graph,並導出三個模糊變數<br />學到網路服務的...
短期<br />如何建構適合的Ontology<br />Ontology-based personalized search and browsing 的Reference<br />網路服務、網頁瀏覽、搜尋個人化相關的中文論文<br />模...
Upcoming SlideShare
Loading in...5
×

維辰Survey101

705

Published on

Just test

Published in: Technology
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
705
On Slideshare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
5
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Transcript of "維辰Survey101"

  1. 1.
  2. 2. Ref 寫法統一<br />標題大小寫<br />置左格式統一<br />流程圖圖形選用、圖形&箭頭說明<br />公式符號說明<br />投影片-老師的建議<br />
  3. 3. [1]<br />修改動機:<br />修改目的(跟本來的技術比較,有何不同)<br />嚴謹的寫出Algo(照著實作,必須做得出來)<br />[2]<br /> Click次數如何搜集?<br />預測有幾種方法?<br />推薦評分值的範圍?相乘後的結果分析?<br />10/1老師給的建議<br />
  4. 4. 解決問題<br />理論上的<br />困難的應用問題<br />為什麼難?<br />貢獻在哪?<br />10/1老師給的建議(cont.)<br />
  5. 5. A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/10/01<br />
  6. 6. UMAP<br />finished *1<br />IUI<br />finished *1, finished but…*2 <br />Related Conference(10/1)<br />
  7. 7. User Modeling(UM, 1986-2007, 11th)<br />California<br />Adaptive Hypermedia and Adaptive Web-Based Systems(AH, 2000-2008, 5th)<br />Italy<br />UM+AH = UMAP(2009, 17th)<br />UMAP(Adaption, Personalization)<br />
  8. 8. Title :<br />[1]Construction of Ontology-Based User Model for Web Personalization (Cited: 9 times)<br /> H. Zhang, Y. Song, and H.T. Song, “Construction of ontology-based user model for web personalization,” Proceeding of the 11th international conference User Modeling 2007, pp. 67–76.<br />UM2007<br />
  9. 9. Authors: Hui Zhang, Yu Song, and Han-tao Song<br />Motivation: to provide web information that matches a user’s personal interests<br />Purpose:<br />Application: personalized web browsing and search<br />[1]Construction of Ontology-Based User Model for Web Personalization<br />
  10. 10. Semantic Web Usage Log Preparation Model(SWULPM)<br />[1]How<br />
  11. 11. Steps:<br />1.S-Log(Semantic-log):representing the semantics of the respective URL(from domain ontology)<br />2.Session analysis algorithm<br /> outcome : semantic session include thematic categories<br />3. IS = user’s new session=outcome<br /> B(IS):user ontology(beginning of the visit is empty)<br /> S(IS):structure of the site(automatically built)<br />4.O = B(IS) U O (O:global user’s ontology)<br />[1]How(cont.)<br />
  12. 12. :look up<br /> :union <br /> :ontology<br />[1]How-Imagination of Ontology<br />Global User Ontology<br />
  13. 13. Each user has a graph:<br />C_Graph(N, u)=<N, A>, N: nodes, A:arcs, u:user<br />arc(s, t)=>label(s, t) = <dst, rst, hst, Tst> <br />dst: semantic independence coefficient<br />rst: semantic relevance coefficient<br />hst: hit coefficient<br />Tst: time coefficient<br />s,t : concept<br />[1]Pre-defined<br />
  14. 14. Duration: 1997-2011<br />Title <br />[1]Personalized News Recommendation Based on Click Behavior (Cited: 2 times)<br /> J. Liu, P. Dolan, and E.R. Pedersen, “Personalized news recommendation based on click behavior,” Proceeding of the 14th international conference on Intelligent User Interfaces, 2010, pp. 31–40.<br />IUI 2010<br />
  15. 15. Authors: Jiahui Liu, Peter Dolan, ElinRonby Pedersen(Google Inc.)<br />Motivation: people was burdened with large online information<br />Purpose: to help users find the information that are interesting to read<br />Application: Google News<br />[1]Personalized News Recommendation Based on Click Behavior<br />
  16. 16. Click behavior<br />advantage<br />no ratings or negative votes<br />after experiment (picture)<br />news interests do change over time<br />click distributions reflect the news trend<br />different news trends in different locations<br />news interests ↔ news trend in location (a certain extent) <br />[1]How<br />
  17. 17. Prediction<br />User’s genuine interests<br />The influence of local news trend<br />Flow<br />predicting user’s genuine news interest from a specific time period t<br />combining predictions of past time periods<br />predicting user’s current news interest<br />recommendation <br />[1]How(cont.)<br />
  18. 18. [1]How(cont.)<br />: predicting user’s current news interest<br /> : current news trend<br /> : past time user’s news interest<br />Nt : all user’s clicks times in t time period<br />G : the number of virtual clicks(smoothing factor)<br />
  19. 19. Recommendation:<br /> (to rank a list of candidate articles)<br />CR(article): content-based recommendation score<br /> CF(article): collaborative filtering recommendation score<br />[1]How(cont.)<br />
  20. 20. 遇到的問題<br />Semantic network與Ontology表達能力;經過學習,建構出符合個別使用者的user model,並依照feedback或觀察使用者的行為,進一步update user model,每個步驟是否真的透徹了解?<br />每個學習法(類神經網路、貝氏分類、最接近鄰居、決策樹)的差異為何?什麼情形選用哪種?<br />未來進度<br />UM 07,AH08,UMAP 09-10<br />IUI 09-10<br />目前進度<br />
  21. 21.
  22. 22. A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/09/10<br />
  23. 23. Service Personalization<br />Early research<br />Overview of user-profile-based personalization<br />User Profile<br />Purpose<br />Type<br />Process of user-profile-based personalization<br />Outline<br />S.Gauch, M.Speretta, A. Chandramouli, and A. Micarelli, “User Profiles for Personalized Information Access, ” The Adaptive Web, LNCS 4321, pp.54-89<br />
  24. 24. Early research<br />Personalization<br />
  25. 25. user-profile-based personalization <br />Overview<br />
  26. 26. Purpose<br />To record interest or habit of the user<br />To filter out irrelevant information from the user<br />To identify additional information of likely interest for the user<br />User Profile<br />
  27. 27. Type<br />Static<br />ex: name, age, country, education level<br />Dynamic<br />short-term <br />long-term<br />User Profile<br />
  28. 28. Process<br />1.Collecting information about users<br />user identification<br />user information collection<br />explicit<br />implicit<br />2.User Profile Representations<br />3.User Profile Construction<br />User Profile<br />
  29. 29. User identification<br />Software agents<br />Logins<br />Enhanced proxy servers<br />Cookies<br />Session ids <br />Collecting information about users<br />
  30. 30. User identification(?)<br />
  31. 31. Explicit<br />Providing personal information (My Yahoo![110])<br />Rating (Web pages, Syskill&Webert[68];Movie, NetFlix[62];Consumer, ePinions[24])<br />Implicit <br />Browsing history (OBIWAN [71])<br />Browsing activity ([71], Trajkova[99], Barrett[6])<br />All user activity (Seruku[83], Surfsaver[94]…)<br />Search (Miserach[87], Liu[45])<br />User information collection <br />
  32. 32. Keyword Profiles<br />Amalthaea[61], Anatagonomy[78], Fab[5], Letizia[43], Syskill&Webert[68], PEA[60]<br />Semantic Network Profiles<br />Minio[56], SiteIF[92], InfoWeb[28], WIFS[53], AltaVista[3], ifWeb[4], Gasparetti[25,26]<br />Concept Profiles <br />Bloedorn[8], Sensus ontology[31,38], Yahoo!directory[42,111], OBIWAN[72]<br />User Profile Representations<br />
  33. 33. Extract from documents visited by the user during browsing<br />Web pages<br />Saved by the user<br />Provide by the user<br />Keyword Profiles<br />
  34. 34. To solve the synonymproblem<br />To solve the polysemy problem<br /> :planet<br /> :satellites <br />Semantic network profiles<br />
  35. 35. Semantic network profiles(cont.)<br />
  36. 36. More abstract topics (not specific words or sets of related words)<br />Concept profiles<br />
  37. 37. User Profile Construction<br />Building keyword profiles<br />Building semantic network profiles<br />Building concept profiles<br />Thank you for attendance!<br />Coming soon…<br />
  38. 38.
  39. 39. A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/09/17<br />
  40. 40. GediminasAdomavicius , Alexander Tuzhilin, Using Data Mining Methods to Build Customer Profiles, Computer, v.34 n.2, p.74-82, February 2001 (Journal)<br />Building Customer Profiles by data mining methods<br />
  41. 41. Validation operator<br />Similarity-based rule grouping<br />Template-based rule filtering<br />Redundant-rule elimination<br />Profile-building process<br />
  42. 42.
  43. 43. Building keyword profiles<br />Amalthaea[61]<br />WebMate[13]<br />Alipes[103]<br />User Profile Construction<br />
  44. 44. Building Keyword profiles<br />
  45. 45. Amalthaea’s Ecosystem[61] <br />
  46. 46. Key V=(W1, W2, W3, …, Wn)<br />(待修改)Amalthaea’s Ecosystem[61](cont.) <br />Web Pages<br />Stemmer<br />Html2txt filter<br />Removal(commonly used)<br />Html2url filter<br />Hc x TF x IDF<br />Moukas, A.: Amalthaea: Information Discovery And Filtering Using A Multi-agent Evolving Ecosystem. In: Applied Artificial Intelligence 11(5) (1997) 437-457 (Journal, Publisher : Taylor & Francis)<br />
  47. 47. WebMate: A personal agent[13]<br /> Chen, L., Sycara, K.: A Personal Agent for Browsing and Searching. In: Proceedings of the 2nd International Conference on Autonomous Agents, Minneapolis/St. Paul, May 9-13, (1998) 132-139<br />
  48. 48. Definition:<br /> 1. Profile set V = { V1, V2,…,VN} <br /> (N domains of interest for each user)<br /> 2. Document Di -> Vector Vi, i={1,…N}<br /> Vi={ e1,e2,…,eM}, <br />ej =TF(wj, Di) x IDF(wj), j={1,…,M} <br />WebMate[13](cont.)<br />
  49. 49. Algorithm for multi TF-IDF vector learning:<br />(待修改)WebMate[13](cont.)<br />User marked “I like It”<br />If |V| < N<br />Add in set V <br />T<br />Parse HTML page<br />F<br />Compare every two vectors by (a)<br />Extract TF-IDF vector<br />Combine Vp, Vq with most similarity<br />Vp = Vp + Vq<br />Sort<br />(a)<br />
  50. 50. Widyantoro, D.H., Yin, J., El Nasr, M., Yang, L., Zacchi, A., Yen, J.: Alipes: A Swift Messenger In Cyberspace. In: Proc. 1999 AAAI Spring Symposium Workshop on Intelligent Agents in Cyberspace, Stanford, March 22-24 (1999)62-67<br />Alipes[103]<br />Control<br />
  51. 51. Alipes[103]<br />
  52. 52. Coming soon…<br />Thank you for listening<br />Building semantic network profiles<br />
  53. 53. A Survey on Service Personalization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/10/22<br />
  54. 54. Authors<br />Susan Gauch, Jason Chaffee and Alexander Pretschner<br />Motivation<br />It’s impossible to use one approach to browsing or searching for every user according to preference.<br />Purpose<br /> Personalized web browsing and search<br />Application<br />Web sites<br />Ontology-based personalized search and browsing (Cited: 194 times)<br />
  55. 55. Reference ontology: Concept, Source<br />Concept<br />To extract top levels of the subject hierarchies (already existing)<br />Source<br />associated web pages from Yahoo, Magellan, Lycos, and the Open Directory Project<br />How-Browsing<br />
  56. 56. How<br />
  57. 57. How<br />
  58. 58. How-Mapping(1)<br />
  59. 59. How-Mapping(2)<br />
  60. 60. Now the site can have its content browsed using the personal ontology<br />How-Mapping(3)<br />
  61. 61. Approaches<br />Re-ranking<br />Filtering(X)<br />How-Searching<br />
  62. 62. (1)To extract only html tags(titles, summaries)<br />(2)Classification<br />(3)examine documents which belongs to User’s concepts<br />(4)…<br />How-Re-ranking<br />
  63. 63. Thank you very much!!<br />
  64. 64. A Survey on Text Categorization<br />學生:張維辰<br />指導教授:劉立頌<br />時間:2010/11/02<br />
  65. 65. Classification<br />supervised learning<br />pre-defined categories<br />ex. credit of consumer<br />Clustering<br />unsupervised learning<br />unknown categories<br />ex. similarity of consumer<br />Preliminary<br />
  66. 66. Motivation<br />With the rapid growth of online information, it is difficult and time-consuming to deal with or classify the information by hand.<br />Purpose<br />To manage and use information easily<br />Application<br />Filter(personal portal site, email)<br />Portal site<br />Semantic identifier<br />Image classification<br />multimedia document classification<br />Text Categorization(TC)<br />
  67. 67. SVM(Support Vector Machine) Vapnik 1995<br />kNN(k-nearest neighbor)<br />NB(Naïve Bayes)<br />LLSF(Linear Least Squares Fit)<br />NNet(Neural network)<br />Approaches of TC<br />
  68. 68. -------以下為中文--------<br />
  69. 69. 建立在最小化結構風險理論上<br />將資料根據特徵轉成Rn空間中的向量,每筆資料可視為空間中的一點,並從Rn空間中找到一個n-1維的界線,稱為分類超平面-H。<br />H1, H2為支援超平面(Support hyperplane)<br />SVM(1)<br />
  70. 70. 多類別支援向量機<br />一對多<br />一對一<br />DAG(directed acyclic graph) method<br />Considering all data at once method<br />C&S method<br />SVM(2)<br />
  71. 71. SVM<br />
  72. 72. kNN(1)<br />
  73. 73. <ul><li>Based on XML</li></ul>Definition<br /><types>-data types<br /><message>-parameters of a function call<br /><portType>-function library<br /><binding>-message format and protocol<br />WSDL(Web Service Description Language)<br />
  74. 74. <message name="getTermRequest">  <part name="term" type="xs:string"/></message><message name="getTermResponse">  <part name="value" type="xs:string"/></message><portType name="glossaryTerms">  <operation name="getTerm">    <input message="getTermRequest"/>    <output message="getTermResponse"/>  </operation></portType><br />Example<br />
  75. 75. One-way<br />Request-response<br />Solicit-response<br />Notification<br />Operation Types<br />
  76. 76. Elements<br />Envelope (root)<br />Header<br />mustUnderstandattr.<br />actor attr.<br />encoding style attr.<br />Body<br />Fault<br />Namespace<br />Soap Envelope<br />Soap Encoding<br />SOAP(Simple Object Access Protocol)<br />
  77. 77. A Survey on Text Categorization<br />Std. :Wei-Chen Chang<br />Prof.:Alan Liu<br />2010/11/18<br />
  78. 78. 動機:資訊數位化,大量的電子文件分類,需耗費人力,且不客觀也缺乏一致性<br />目的:利用Ontology來協助分類,增加其準確性並節省人力、達到電子文件分類客觀與一致性<br />應用:線上即時新聞自動分類<br />做法:下頁<br />鐘明強, “基於Ontology架構之文件分類網路服務研究與建構”, 成功大學資訊工程所, 2004<br />基於Ontology架構之文件分類網路服務研究與建構<br />
  79. 79. 系統架構<br />基於Ontology架構之文件分類網路服務研究與建構<br />
  80. 80. Domain Weighted Ontology<br />以Object-Oriented Ontology為基礎,透過專家建構<br /> Domain Ontology(政治、社會、氣象、運動、財經)<br />訓練此Domain Ontology成為Domain Weighted Ontology<br />概念<br />概念之間的關係<br />基於Ontology架構之文件分類網路服務研究與建構<br />
  81. 81. 分類機制<br />基於Ontology架構之文件分類網路服務研究與建構<br />
  82. 82. 五階層式模糊推論機制<br />1.輸入層(Input Layer)<br />2.輸入語意層(Input Linguistic Layer)<br />3.規規層(Rule Layer)<br />4.輸出語意層(Ouput Linguistic Layer)<br />5.輸出層(Output Layer)<br />模糊推論<br />
  83. 83. 文字前處理網路服務<br />http://140.116.247.14/text_classification/text_classification.asmx?op=autotag<br />新聞分類網路服務<br />http://140.116.247.14/fuzzyclassification/service1.asmx?op=SRS<br />網路服務<br />
  84. 84. 透過專家定義好的Ontology,來完成分類工作,不需要花很長的訓練時間。<br />學習到透過訓練資料,如何給予Ontology中的概念與關係權重值。<br />Ontology如何轉成Graph,並導出三個模糊變數<br />學到網路服務的基本觀念。<br />結論與吸收<br />
  85. 85. 短期<br />如何建構適合的Ontology<br />Ontology-based personalized search and browsing 的Reference<br />網路服務、網頁瀏覽、搜尋個人化相關的中文論文<br />模糊理論<br />洪正鑫, ”應用個人本體論於個人化網路服務選擇之研究”<br />中期<br />UMAP、IUI<br />未來目標<br />
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×