應(yīng)用社會網(wǎng)路分析于學(xué)術(shù)社群推薦課件

1、應(yīng)用社會性推薦於學(xué)術(shù)社群 Using Social Recommendation in Academic Community柯皓仁國立臺灣師範(fàn)大學(xué)圖書資訊學(xué)研究所粘怡祥國立交通大學(xué)資訊管理研究所1大綱緒論相關(guān)文獻研究方法系統(tǒng)發(fā)展與實證分析結(jié)論與建議2緒論3研究背景與動機資訊過載(Information Overload)搜尋引擎與推薦系統(tǒng)的出現(xiàn)，成為改善資訊過載問題的兩大利器使用者除本身的主觀喜好之外，其為容受到人際關(guān)係的影響虛擬社群與社會網(wǎng)，成為許多使用者獲得資訊情報的最佳來源本研究探討如何運用社會網(wǎng)路提升資訊推薦的品質(zhì)4研究目的本研究希望透過主題概念萃取與社會網(wǎng)路分析，建構(gòu)資訊推薦系統(tǒng)，藉

2、此達到以下的目標(biāo)：主題概念萃?。狠腿〕鑫募械闹匾P(guān)鍵字用關(guān)鍵字分群的方式，達到主題概念萃取的目的，藉以瞭解使用者所關(guān)注的興趣與議題形成主題社群以向空間模型表示使用者的個別興趣，並結(jié)合使用者社會網(wǎng)路，將相似高且具有相同主題興趣的使用者群聚在一起，以形成主題社群資訊推薦經(jīng)由主題社群的產(chǎn)生，針對使用者個人的主題偏好，進行個人化推薦5相關(guān)研究6社會網(wǎng)路分析社會網(wǎng)路分析(Social Network Analysis)是一種研究社會結(jié)構(gòu)、組織系統(tǒng)、人際關(guān)係、團體互動的概與方法，是在社會計學(xué)基礎(chǔ)上所發(fā)展出的分析方法社會網(wǎng)路分析研究域中，最著名的之一為度分隔理論40最初利用信件傳遞實驗，發(fā)現(xiàn)從寄件者到收件

3、者之間，平均轉(zhuǎn)寄次指互不相干的個人，最多可經(jīng)由五個中介者結(jié)出某種關(guān)係7社會網(wǎng)路示意圖/wiki/Social_network 8社會網(wǎng)路分析(Cont.)在社會網(wǎng)路分析中，個別行動者的量測指標(biāo)主要有以下三項21：Degreenumber of direct connectionsBetweennessrole of broker or gatekeeperCloseness Centralitywho has the shortest path to all others9Clustering AlgorithmPartitioning methodsk-Means Hierarchical

4、methodsAgglomerative Divisive Model-based methodsSelf-Organizing Map 10Clustering Algorithm (續(xù))Partitioning methodsk-Means Hierarchical methodsAgglomerative Divisive Model-based methodsSelf-Organizing Map11Clustering Algorithm (續(xù))Partitioning methodsk-MeansHierarchical methodsAgglomerativeDivisiveMo

5、del-based methodsSelf-Organizing Map12推薦系統(tǒng)推薦系統(tǒng)的目的是從大量資訊中找出使用者最可能感興趣的部份，減少使用者主動搜尋的機會成本目前常應(yīng)用在推薦系統(tǒng)的方法主要有兩種內(nèi)容導(dǎo)向(Content-based)式推薦協(xié)同過濾(Collaborative Filtering)式推薦13資訊檢索向量空間模型The vector model ranks the documents according to their degree of similarity to the query, and retrieve the documents with a degree

6、 of similarity above a thresholdDefineWeight wi,j associated with a pair (ki, dj) is positive and non-binary (t is the total number of index terms)The index terms in the query are also weightedwi,q is the weight associated with the pair ki, q, where wi,q = 0 (t is the total number of index terms)Deg

7、ree of similarity of dj with regard to q: The cosine of the angle between the two corresponding vectors14資訊檢索向量空間模型圖示NormalizedTerm-document matrix15資訊檢索向量空間模型圖示16研究方法17語料庫本研究以交通大學(xué)機構(gòu)典藏系統(tǒng) 38所收集的期刊文做為語庫選取標(biāo)題(Title)、摘要(Abstract)、關(guān)鍵字(Keyword)及作者(Author)欄位做為資源.tw 系統(tǒng)雛型展示18前置處理斷詞字(Tokenization)與小寫化(Lowercas

8、ing) 刪除停用字(Stopword Removing) 詞性標(biāo)記(Part-of-speech) 片語化(Chunking) 詞幹還原(Stemming)特徵選擇(Feature Selection)19Some combinatorial characteristics of matrix multiplication on regular two-dimensional arrays are studied. From the studies, the authors are able to design many efficient varieties of the cylindri

9、cal array and the two-layered mesh array for matrix multiplication.some combinatorial characteristics of matrix multiplication on regular two-dimensional arrays are studied from the studies the authors are able to design many efficient varieties of the cylindrical array and the two-layered mesh arra

10、y for matrix multiplicationsome combinatorial characteristics of matrix multiplication on regular two-dimensional arrays are studied from the studies the authors are able to design many efficient varieties of the cylindrical array and the two-layered mesh array for matrix multiplicationcombinatorial

11、 characteristics matrix multiplication regular two-dimensional arrays studied studies authors design efficient varieties cylindrical array two-layered mesh array matrix multiplicationcombinatorial_jj characteristics_nns matrix_nn multiplication_nn regular_jj two-dimensional_jj arrays_nns studied_vbn

12、 studies_nns authors_nns design_vb efficient_jj varieties_nns cylindrical_jj array_nn two-layered_jj mesh_nn array_nn matrix_nn multiplication_nnPOSPhrasenounnounnounverbnounnounverbnounnounnounnouncombinatorial characteristicsmatrix multiplicationregular two-dimensional arraysstudiedstudiesauthorde

13、signefficient varieties cylindrical arraytwo-layered mesh arraymatrix multiplicationPOSPhrasenounnounnounverbnounnounverbnounnounnounnouncombinatori characteristmatrix multiplregular two-dimension arraistudistudiauthordesigneffici varieticylindr arraitwo-lay mesh arraimatrix multipl前置處理(續(xù))20主題關(guān)鍵字分群使

14、用者模型計算語意相關(guān)度建立語意網(wǎng)路圖關(guān)鍵字分群關(guān)鍵字分群標(biāo)記21使用者模型採用TF-IAF (Term Frequency-Inverse Author Frequency)30來衡量使用者與關(guān)鍵字間的關(guān)聯(lián)計算完TF-IAF後，每個使用者皆可以向量的形式來呈現(xiàn)22計算語意相關(guān)度本研究以子為範(fàn)圍，即個關(guān)鍵字在同一子內(nèi)出現(xiàn)才表示其具有語意相關(guān)。透過增加標(biāo)題(Title)及關(guān)鍵字(Keyword)權(quán)重強化這些關(guān)鍵字關(guān)係之代表性23建立語意網(wǎng)圖每個關(guān)鍵字可表示為一個點，點權(quán)重為個別關(guān)鍵字在使用者間TF-IAF的加總，再加上該關(guān)鍵字所有語意相關(guān)度平均關(guān)鍵字間的關(guān)係表示成一個邊，邊權(quán)重即為關(guān)鍵字的語意相關(guān)

15、度運用9的方法進行主題關(guān)鍵字分群24建立語意網(wǎng)路圖25主題關(guān)鍵字分群示意圖926選取重要候選關(guān)鍵字Finding vertices whose weights are larger than the average weight27主題關(guān)鍵字分群(Cont.)k-Nearest Neighbor Approach19考慮圖中的每個點，取與該點最相近的k個點為一組，每組為一個通圖，稱之為候選關(guān)鍵字組產(chǎn)生候選關(guān)鍵字子群以每個候選關(guān)鍵字組為中心，向外還原先前與候選關(guān)鍵字組內(nèi)的點有直接線關(guān)係的邊，形成候選關(guān)鍵字子群，並計算每個子群的權(quán)重，如方程式(3-6)所示。(3-6)28關(guān)鍵字分群Use k-ne

16、arest neighbor graph approach29主題關(guān)鍵字分群(Cont.)合併候選關(guān)鍵字子群找出互性(Inter-connectivity)最強的個子群將之合併，直到子群間的互相關(guān)(Relative Inter-connectivity)小於門檻值後停止。互相關(guān)度方程式(3-7)所示。(3-7)30合併候選關(guān)鍵字子群31主題關(guān)鍵字分群(Cont.)修正並產(chǎn)生主題關(guān)鍵字分群讓每個子群內(nèi)的關(guān)鍵字個保持在一定的差距內(nèi)子群內(nèi)包含的關(guān)鍵字比平均個數(shù)少，但子群權(quán)重卻大於平均權(quán)重時，將該群保子群經(jīng)修正後仍小於平均權(quán)重，將該群直接刪除子群權(quán)重如方程式(3-8)所示(3-8)32修正並產(chǎn)生主題關(guān)

17、鍵字分群33關(guān)鍵字分群標(biāo)記利用人過出有意義的關(guān)鍵字取權(quán)重最高的關(guān)鍵字做為最後群的標(biāo)記34建立主題社群使用者社會網(wǎng)路使用者分群35使用者社會網(wǎng)路36使用者社會網(wǎng)路(續(xù))37使用者分群將所有使用者向量模型以Nm的矩陣U表示，N代表使用者數(shù)目，m代表所有關(guān)鍵字數(shù)目以矩陣R代表使用者間相關(guān)係數(shù)，乘上以使用者向量模型構(gòu)成的矩陣U ，形成一新的矩陣U代表更新後的使用者向量模型(參數(shù)調(diào)整R的影響程度)38使用者分群(續(xù))以餘弦相似度(Cosine Similarity)計算使用者與個別主題的相似度，當(dāng)使用者與主題間的相似度大於門檻值時，則將其歸類到該主題39推薦模式在社群中的成員都具有相似的主題興趣，但是由

18、於多重主題9的屬性存在，使得使用者可能對多種主題都具有偏好，於是產(chǎn)生個人化推薦與社群推薦兩種推薦模式，茲分述如下：個人化推薦(Collaborative Filtering)依據(jù)內(nèi)容導(dǎo)向方法，對使用者進行論文推薦，即計算社群內(nèi)成員所撰寫的論文與個別成員的相似度，選取相似度最高的n篇論文給予推薦社群推薦(擴展閱讀層面)透過分析社群成員對其他主題的興趣分佈，統(tǒng)計出具有較高偏好比重的主題，推薦項目以與該主題最相關(guān)的n篇論文40系統(tǒng)發(fā)展與實證分析41系統(tǒng)發(fā)展系統(tǒng)架構(gòu)42系統(tǒng)發(fā)展系統(tǒng)介面43系統(tǒng)發(fā)展系統(tǒng)介面(續(xù))44系統(tǒng)發(fā)展系統(tǒng)介面(續(xù))45系統(tǒng)發(fā)展系統(tǒng)介面(Cont.)46系統(tǒng)發(fā)展系統(tǒng)介面(續(xù))47實

19、驗結(jié)果分群結(jié)果評估首先將系統(tǒng)分群的結(jié)果分類，即將相近的群歸屬於同一類依序?qū)€別使用者進行分類之動作採用準(zhǔn)確(Precision)與回現(xiàn)(Recall)兩項指標(biāo)15，來評估分群結(jié)果的好壞48Class labelCluster labelNetwork CommunicationMobile ComputingRouting ProtocolPIM-SMBandwidth RequestsTCPNetwork ManagementArtificial IntelligenceGenetic AlgorithmNetwork MotifBrick Motif Content AnalysisNeu

20、ral NetworkSPDNNDivide-and-conquer LearningComputer GraphicsContent-based Image RetrievalWatershed SegmentationToboggan ApproachInformation RetrievalSemantic QueryContent ManagementComputer SystemMemory CacheParallel Algorithm Information SecurityEnd-to-end SecurityGraph TheoryInterconnection Networ

21、kSoftware EngineeringReliability Analysis實驗結(jié)果分群結(jié)果評估(續(xù))49實驗結(jié)果分群結(jié)果評估(續(xù))Class label# of authorsNetwork Communication111Artificial Intelligence28Information Retrieval7Computer System6Computer Graphics23Information Security10Graph Theory29Software Engineering4Others17Total23550實驗結(jié)果分群結(jié)果評估(續(xù)) value00.10.20

22、.1Precision0.7071 0.6917 0.6981 0.7107 0.7172 0.7209 0.7209 0.7209 0.7209 0.7209 0.7209 Recall0.6271 0.7606 0.7785 0.7839 0.7817 0.7828 0.7828 0.7828 0.7828 0.7828 0.7828 51實驗結(jié)果推薦結(jié)果評估標(biāo)凖差為0.068，當(dāng)信賴水凖達95%時，信賴區(qū)間為(0.632, 0.897)； Kappa值為0.764，專家同意度為0.95針對專家具有相同意見之推薦結(jié)果，總共有208筆，認為符合使用者需求

23、之推薦有187筆，則推薦之準(zhǔn)確率為187/208=0.899Expert ANoYesTotalExpert BNo21(9.6%)9(4.1%)30(13.7%)Yes2(0.9%)187(85.4%)189(86.3%)Total23(10.5%)196(89.5%)21952作者收錄論文數(shù)量分析論文收錄的篇數(shù)介於1篇到41篇，只收錄1篇文章的作者有129位，佔全部作者的55%；收錄少於5篇的作者有93%53作者收錄論文數(shù)量分析(續(xù))NamePublicationsYu-Chee Tseng (曾煜棋)Jimmy J. M. Tan (譚建民)Lih-Hsing Hsu (徐力行)Yi-B

24、ing Lin (林一平)Ying-Dar Lin (林盈達)Ling-Hwei Chen (陳玲慧)Chuen-Tsai Sun (孫春在)Jang-Ping Sheu (許健平)Hsin-Chia Fu (傅心家)Hao-Ren Ke (柯皓仁)Wei-Pang Yang (楊維邦)Wen-Guey Tzeng (曾文貴)Chien-Chao Tseng (曾建超)Tseng-Kuei Li (李增奎)Wen-Chih Peng (彭文志)Chang-Hsiung Tsai (蔡正雄)Deng-Jyi Chen (陳登吉)Yuan-Cheng Lai (賴源正)41363332261714

25、1311108877666654共同作者分析共同作者數(shù)介於1到6位作者之間，只有單一作者的論文有6篇，佔全部論文數(shù)的3%；共同作者為2到6位間的論文篇數(shù)共有220篇，佔全部的97%55社會網(wǎng)路Yu-Chee Tseng56社會網(wǎng)路量測指標(biāo)分析RankDegreeBetweennessCloseness1234567891011121314151617181920Yu-Chee TsengYi-Bing LinYing-Dar LinJimmy J. M. TanLih-Hsing HsuHsin-Chia FuJang-Ping SheuChien-Chao TsengChuen-Tsai S

26、unHao-Ren KeLing-Hwei ChenWei-Pang YangHsiao-Tien PaoZen-Chung ShihChang-Hsiung TsaiJeu-Yih JengYeong-Yuh XuDeng-Jyi ChenWen-Guey TzengMing-Hour Yang4332292926161514121110887777777Yu-Chee TsengChien-Chao TsengYi-Bing LinMing-Feng ChangYing-Dar LinWen-Chih PengJimmy J. M. TanLih-Hsing HsuChuen-Tsai S

27、unHsin-Chia FuLing-Hwei ChenJang-Ping SheuSunny S.J. LinHao-Ren KeChi-Fu HuangWen-Guey TzengShi-Chun TsaiDeng-Jyi ChenZen-Chung ShihWei-Pang Yang2660.3332180.5002081.3331792.000376.500340.000213.167133.16791.00086.00044.00038.33336.00032.83322.50021.33315.33312.00012.0008.833Yu-Chee TsengChien-Chao

28、TsengMing-Feng ChangChi-Fu HuangHsiao-Lu WuYuan-Ying HsuJung-Hsuan FanYi-Bing LinHang-Wen HwangJang-Ping SheuWen-Chih PengMeng-Ta HsuLin-Yi WuMing-Hour YangChih-Yu LinSze-Yao NiWen-Hwa LiaoShih-Lin WuChih-Shun HsuChi-He Chang0.6780.6780.6780.6770.6770.6770.6770.6770.6770.6770.6770.6770.6770.6770.677

29、0.6770.6770.6770.6770.67757結(jié)論與建議58結(jié)論本研究致於改善資訊推薦的效能，主要的目在於提出結(jié)合主題概念萃取與社會網(wǎng)路分析之資訊推薦系統(tǒng)，以提供符合使用者需求之推薦資訊。經(jīng)由實驗與統(tǒng)計分析的驗證，將本研究的結(jié)果整理如下：主題概念萃?。核?35位作者，226篇論文中，共產(chǎn)生22個主題概念形成主題社群：經(jīng)由實驗發(fā)現(xiàn)，社會網(wǎng)路對提升使用者分群之回現(xiàn)有較佳之效果，代表其能發(fā)掘出更多具有關(guān)聯(lián)性之使用者資訊推薦：資訊推薦之準(zhǔn)確率為0.899，顯見系統(tǒng)之推薦效果，頗能符合使用者需求59後續(xù)建議建主題本體論進主題萃取的過程中，利用階層式分群法以樹結(jié)構(gòu)表示主題分群之結(jié)果，產(chǎn)生主題概階

30、層經(jīng)由使用者主題偏好之關(guān)聯(lián)，建立主題概念之連結(jié)，以形成主題本體論幫助使用者瞭解本身處於何種階層層級，未來可朝哪些研究方向前進使用者評分之應(yīng)用使用者評分可分為明顯性評分與隱含性評分。明顯性評分為使用者依對目標(biāo)物感興趣程度給予主觀評分；隱含性評分的估計通常以使用者的瀏覽行為做依據(jù)經(jīng)由使用者評分可以更精確瞭解使用者偏好所在，使資訊推薦更符合使用者需求60後續(xù)建議(續(xù))社會網(wǎng)路之階層擴展可經(jīng)由建立在共同社會網(wǎng)路中之使用者關(guān)係，進一步探討社會網(wǎng)路之資訊流動及影響。例如使用Floyd-Warshall演算法可找出位於同一社會網(wǎng)路中，兩兩使用者間的最短路徑，則可經(jīng)由節(jié)點的分析，研究其對使用者的影響。機構(gòu)典藏

31、與資料庫系統(tǒng)之加值應(yīng)用以其他方式(如引用、共引)架構(gòu)社會網(wǎng)路61參考文獻(續(xù))A. Iskold, (2007) “The Art, Science and Business of Recommendation Engines.” http:/archives/recommendation_engines.phpA. K. Jain, M. N. Murty, & P. J. Flynn, “Data clustering: A review,” ACM Computing Surveys, vol. 31,pp. 264-323, 1999.B. Krulwich, & C. Burkey,

32、 “The InfoFinder agent: Learning user interests through heuristic phrase extraction,” IEEE Expert: Intelligent Systems and Their Applications, vol. 12, pp. 22-27, 1997.B. Sarwar, G. Karypis, J. Konstan, & J. Riedl, “Analysis of recommendation algorithms for e-commerce,” Proceedings of the 2nd ACM co

33、nference on Electronic commerce, pp. 158-167, 2000.D. Goldberg, D. Nichols, B. M. Oki, & D. Terry, “Using Collaborative Filtering to Weave An Information Tapestry,” Communications of the ACM, vol. 35, pp. 61-70, 1992.D. Koller, & M. Sahami, “Hierarchically classifying documents using very few words,

34、” Proceedings of 14th the International Conference on Machine Learning, pp.170178, 1997.F. Sebastiani, “Machine learning in automated text categorization,” ACM Computing Surveys, vol. 34, pp. 1-47, 2002.G. Karypis, E. H. Han, & V. Kumar, “Chameleon: Hierarchical clustering using dynamic modeling,” C

35、omputer, vol. 32, pp. 68-75, 1999.H. C. Chang, & C. C. Hsu, “Using topic keyword clusters for automatic document clustering,” Transactions on Information and Systems, vol. 88, pp. 1852-1860, 2005.H. Hotta , “User profiling system using social networks for recommendation”, In Proceedings of 8th Inter

36、national Symposium on Advanced Intelligent Systems , 2007.H. Kautz, B. Selman, & F. Park, “Referral Web: Combining social networks and collaborative filtering,” Communications of the ACM, vol. 40 , pp. 63-65, 1997. 62參考文獻(續(xù))H. Sakagami, & T. Kamba, “Learning Personal Preferences on Online Newspaper

37、Articles from User Behaviors,” Computer Networks and ISDN Systems, vol. 29, pp. 1447-1455, 1997.J. B. Schafer, J. Konstan, & J. Riedi, “Recommender systems in e-commerce,” Proceedings of the 1st ACM conference on Electronic commerce, pp. 158-166, 1999. J. MacQueen, “Some methods for classification a

38、nd analysis of multivariate observations,” Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability, vol. 1, pp. 281-297, 1967.J. Makhoul, F. Kubala, R. Schwartz, & R. Weischedel, “Performance measures for information extraction,” Proceedings of DARPA Broadcast News Workshop,

39、 pp. 249-252, 1999.J. Moreno, Who Shall Survive? New York: National Institute of Mental Health, 1934.J. R. Tyler, D. M. Wilkinson, & B. A. Huberman, “Email as spectroscopy: Automated discovery of community structure within organizations,” Communities and technologies, pp. 81-96, 2003.J. Rucker, & M.

40、 J. Polanco, “Siteseer: Personalized navigation for the web,” Communications of the ACM, vol. 40, pp. 73-76, 1997.K. C. Gowda, & G. Krishna, “Agglomerative clustering using the concept of mutual nearest neighbourhood,” Pattern Recognition, vol. 10, pp. 105-112, 1978.K. Faust, “Comparison of methods

41、for positional analysis:Structural and general equivalences,” Social Networks, vol. 10, pp.313-341, 1988.L. C. Freeman, “Centrality in Social Networks: Conceptual clarification,” Social Networks, vol. 1, pp. 215-239, 1979.L. Page, & S. Brin, “The anatomy of a large-scale hypertextual Web search engi

42、ne,” In Proceedings of the seventh international World-Wide Web conference, 1998.63參考文獻(續(xù))L. Garton, C. Haythornthwaite, & B. Wellman, (1997) “Studying Online Social Networks,” http:/jcmc.huji.ac.il/vol3/issue1/garton.htmlM. A. Shah, “ReferralWeb: A resource location system guided by personal relati

43、ons,” Masters thesis, M.I.T., 1997.M. Granovetter, “The strength of weak ties: A network theory revisited,” Sociology Theory, vol. 1, pp. 201-233, 1983.N. Zhong, J. Liu & Y. Yao, “In search of the wisdom web,” Computer, vol. 35, pp. 27-31, 2002.P. Athanasios, Probability, Random Variables and Stocha

44、stic Processes., Second Edition ed. New York: McGraw-Hill, 1984.P. Mika, “Flink: Semantic Web technology for the extraction and analysis of social networks,” Web Semantics: Science, Services and Agents on the World Wide Web, vol 3, pp. 211-223, 2005.P. Pattison, Algebraic models for social networks., Cambridge University Press, 1993.