多屏時代的來臨，使用者需要更貼近生活的網路服務。根據統計，「社群」和「搜尋」是使用者最常用的服務。因此，提供使用者一個好的搜尋服務是一個相當重要的議題。傳統搜尋引擎對於輸入的查詢往往只視為一個簡單任務來進行處理，然而，使用者所下的搜尋詞是由複雜任務驅動，導致搜尋結果不如預期或是需要進行一系列的查詢，這對使用者是一個好的使用者體驗。
在本論文中，我們以複雜任務為基礎，檢索任務目標，幫助使用者尋找提供目標服務的異質實體。我們的模型包含4個步驟，分別為產生任務目標、尋找實體目錄、辨識異質實體與利用使用者意見來推薦實體。首先，我們產生使用者查詢的任務目標，使用Google與Wiki的搜尋結果頁面來尋找每個目標的實體目錄，再利用目錄去尋找實體及辨識與目錄關聯性。最後收集實體的社群網路資料，根據實體在社群網路上使用者的評價，進行排名。舉例來說：使用者輸入的複雜任務叫做「北京旅遊」，子任務目標為「訂購機票」、「預定飯店」及「查詢景點」。我們可以從「訂購機票」的子任務目標找到「航空」的實體目錄，並尋找目錄下的實體 (「國泰航空」、「長榮航空」及「中華航空」)推薦給使用者。我們提出的模型使用多樣性的網路資源，能有效地從使用者複雜任務中推薦符合網路評價的異質實體，提供使用者更好的搜尋服務。

With the times of the Multi-Screen, users pursued the internet services closer to life. The study found that users will use different devices to different services. No matter which devices be used, "Social" and "Search" are the most popular services. Therefore, it is an important issue to provide good services to users. Conventional search engines usually consider a search query corresponding only to a simple task. However, the queries from users are driven by complex tasks. The results are less than expected or repeated query. It’s not a good user experience for the query that had latent goal. We defined those searches as complex task search.
In this work, we based on complex task to retrieval the task goal and help users to find the heterogeneous entity. Our heterogeneous entity recommendation model (HERM) contains four stages. First, topic-event-based complex task model which is generated the subtask goal from complex task. Second, using search engine results and Wikipedia search result page to extract entity category for each subtask goal in entity category. Third, found the heterogeneous entities for each entity category and used SVM to identify the related between entity and entity category. Finally, we collected the opinions from social network resources (i.e. Facebook, Google Plus) and based on opinions to rank entities. For instance, the complex task “travel to Beijing” may involve several subtask goals, including “book flight”, “reserve hotel” and “survey spot”. We can find the entity category “airline” from “book flight”. In next step, we can find the entity (i.e. “Cathay Pacific Airways”, “EVA Airways”, “China Airlines”) from entity category and recommend to users. We proposed that HERM used diverse web resources to opinions form social network to recommend heterogeneous entity for complex task.

[1] Guo, Q. and Agichtein, E. Ready to Buy or Just Browsing? Detecting Web Searcher Goals from Interaction Data. In Proc. of SIGIR, 130-137, 2010.
[2] Jones, R., and Klinkner, K. Beyond the Session Timeout: Automatic Hierarchical Segmentation of Search Topics in Query Logs. In Proc. of CIKM, 699-708, 2008.
[3] Feild, H. and Allan, J. Task-Aware Query Recommendation. In Proc. of SIGIR, 83-92, 2013.
[4] Cui, J., Liu, H., Yan, J., Ji L., Jin R., He, J., Gu, Y., Chen, Z., and Du, X. Multi-view Random Walk Framework for Search Task Discovery from Click-through Log. In Proc. of CIKM, 135-140, 2011.
[5] Boldi, P., Bonchi, F., Castillo, C., Donato, D., Gionis, A., and Vigna, S. The Query-Flow Graph: Model and Applications. In Proc. of CIKM, 609-618, 2008.
[6] Jones, R., and Klinkner, K. Beyond the Session Timeout: Automatic Hierarchical Segmentation of Search Topics in Query Logs. In Proc. of CIKM, 699-708, 2008.
[7] Kellar, M., Watters, C., and Shepherd, M. A Field Study Characterizing Web-based Information-seeking Tasks. JASIST, 58(7), 999-1018, 2007.
[8] Rose, D. E., and Levinson, D. Understanding User Goals in Web Search. In Proc. of WWW, 13-19, 2004.
[9] Yin, X. and Shah, S. Building Taxonomy of Web Search Intents for Name Entity Queries. In Proc. of WWW, 1001-1010, 2010.
[10] Broder, A. A Taxonomy of Web Search. In Proc. of SIGIR Forum, 36(2), 3-10, 2002.
[11] Agichtein, E., White, R. W., Dumais, S. T., and Bennett, P. N. In Proc. of SIGIR, 315-324, 2012.
[12] Kotov, A., Bennett, P. N., White, R. W., Dumais, S. T., and Teevan, J. Modeling and Analysis of Cross-Session Search Tasks. In Proc. of SIGIR, 5-14, 2011.
[13] Liu, J. and Belkin, N. J. Personalizing Information Retrieval for Multi-Session Tasks: The Roles of Task Stage and Task Type. In Proc. of SIGIR, 26-33, 2010.
[14] MacKay, B. and Watters, C. Exploring Multi-Session Web Tasks. In Proc. of CHI, 1187-1196, 2008.
[15] Wang, H., Song, Y., Chang, M.-W., He, X., White, R. W., and Chu, W. Learning to Extract Cross-Session Search Tasks. In Proc. of WWW, 1353-1364, 2013.
[16] Lee, U., Liu, Z., and Cho, J. Automatic Identification of User Goals in Web Search. In Proc. of WWW, 391-400, 2005.
[17] Lin, T., Pantel, P., Gamon, M., Kannan, A., and Fuxman, A. Active Objects: Actions for Entity-Centric Search. In Proc. of WWW, 589-598, 2012.
[18] He, X., Gao, M., Kan, M.-Y., Liu Y., Sugiyama K. Predicting the popularity of web 2.0 items based on user comments. In Proc. Of SIGIR, 233-242, 2014.
[19] Spirin, N. V., He J., Develin M., Karahalios K.G., Boucher M., People Search within an Online Social Network: Large Scale Analysis of Facebook Graph Search Query Logs. In Proc. of CIKM, 1009-1018, 2014.
[20] Chang Y., Dong A., Kolari P., Zhang R., Inagaki Y., Diza F., Zha H., Liu Y., Improving Recency Ranking Using Twitter Data. In Proc. Of ACM TIST, article 4, 2013.
[21] Tsai K. Y., Generating Complex Task Names with Sub-Task Goals to Improve Web Search by Utilizing Multiple Web Resources. Master’s thesis, National Cheng Kung University, Tainan, Taiwan, R.O.C., 2013.
[22] Ageev, M., Lagun, D., and Agichtein, E. Improving Search Result Summaries by Using Searcher Behavior Data. In Proc. of SIGIR, 13-22, 2013.
[23] Dror, G., Maarek, Y., Mejer, A., and Szpektor, I. From Query to Question in One Click: Suggesting Synthetic Questions to Searchers. In Proc. of WWW, 391-402, 2013.
[24] Liu, Q., Agichtein, E., Dror, G., Maarek, Y., and Szpektor, I. When Web Search Fails, Searchers Become Askers: Understanding the Transition. In Proc. of SIGIR, 801-810, 2012.
[25] Yan, Q., Wu, L., and Zheng, L. Social Network Based Microblog User Behavior Analysis. Physica A: Statistical Mechanics and its Application, 392(7), 1712-1723, 2013.
[26] Blei, D. M., Ng, A. Y., and Jordan, M. I. Latent Dirichlet Allocation. JMLR, 3, 993-1022, 2003.
[27] Lafferty, J., Mccallum, A., and Pereira, F. Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data. In Proc. of ICML. 282-289, 2001.
[28] Griffiths, T. Gibbs Sampling in the Generative Model of Latent Dirichlet Allocation. Technical Report, Stanford University, 2002.
[29] Hsu C.-W., and Lin C.J., A Comparison of Methods for Multiclass Support Vector Machines. Proc. of IEEE Transactions on Neural Networks, Vol. 13. 415-425, 2001.
[30] Dekel O, Manning C. D., Singer Y., Log-Linear Models for Label Ranking. Technical Report, Hebrew University and Stanford University, 2003
[31] Zheng Y, Zhang L, Ma Z, Xie X, Ma W.Y., Recommending friends and locations based on individual location history. In ACM Transaction on the Web (ACM TWEB), 5(1), 2011.
[32] Zheng V.W, Zheng Y, Xie X, Yang O., Collaborative Location and Activity Recommendations With GPS History Data., WWW,2010
[33] Mukhopadhyay D, Dutta R, Kundu A., Dattagupta R, A Product Recommendation System using Vector Space Model and Association Rule, In Prod ICIT. 279 - 282, 2008