日本データベース学会の皆様， （重複して受け取られた場合にはご容赦ください） 京都大学の加藤と申します． 曖昧なクエリや不明快なクエリに秘められた検索意図を扱うタスクのご案内で す。奮ってご参加ください． ################################################### CALL FOR TASK PARTICIPATION NTCIR-10 INTENT-2 Task: English/Chinese/Japanese Subtopic Mining subtasks and Chinese/Japanese Document Ranking subtasks http://research.microsoft.com/en-us/people/tesakai/intent2.aspx Topics released: May 2012 Evaluation results released: Jan 2013 Final NTCIR-10 workshop: June 18-21 2013 ################################################### ### OVERVIEW ### The NTCIR INTENT task addresses the problem of handling ambiguous and/or underspecified queries in Web search. Does "office" mean a Microsoft product or a workplace? (ambiguity) Does "harry potter" mean the books, the films or the character? (lack of specificity) Unlike traditional information retrieval where only relevance plays the central role, INTENT considers multiple possible intents for a given query and aims to balance relevance and diversity. The first round of INTENT at NTCIR-9 was concluded successfully, thanks to great research contributions from 17 research groups from seven countries/regions. At INTENT-1, it has been found that some successful systems were relevance-oriented, while others were diversity-oriented. NTCIR-10 will host the second round of INTENT (INTENT-2) to further clarify and hopefully solve the challenges of handling multiple intents. ### TASKS ### SUBTOPIC MINING subtask: Given a query, systems must return a ranked list of "subtopic strings" that represent diverse intents. In addition to Chinese and Japanese, this time we are running an English Subtopic Mining subtask as well. For this, we will syncronise with the TREC 2012 diversity task and share the topic set. DOCUMENT RANKNIG subtask: Given a query, systems must return a selectively diversified web search results. Unlike the TREC diversity task, we take the importance of intents and per-intent graded relevance into account when evaluating the participating systems. This time, we will include some "one item search" queries in the topic set, which probably do not require diversification. Can systems selectively and effectively diversify? In addition to topics, the organisers will provide query suggestion data and baseline web search results to enhance repeatability and fairness, and to encourage newcomers to participate. ### IMPORTANT DATES ### May 2012 topics, query suggestions and baseline runs released July 2012 all submissions due Aug-Dec 2012 identify intents from subtopics -> per-intent relevance assessments Jan 2013 evaluation results released March 2013? draft papers due May 2013? camera ready papers due June 18-21 2013 NTCIR-10 ### HOW TO PARTICIPATE ### Task registration page will be available soon on the NTCIR-10 website (see below). Participants who plan to index the document corpora need to obtain the SogouT data for Chinese, ClueWeb09-JA for Japanese, and ClueWeb09 for English. How to obtain the data is described in the INTENT-1 homepage (see below). We encourage new participants to have a look at the INTENT-1 overview paper to familiarise themselves with the task: http://research.nii.ac.jp/ntcir/workshop/OnlineProceedings9/NTCIR/01-NTCIR9-... Please note: task participants must write a paper for the NTCIR-10 meeting (unrefereed) and attend the meeting at NII, Tokyo in June 2013. ### ORGANISERS ### Tetsuya Sakai, MSRA Min Zhang, Tsinghua University Ruihua Song, MSRA Yiqun Liu, Tsinghua University Zhicheng Dou, MSRA Makoto Kato, Kyoto University Mayu Iwata, Osaka University Takehiro Yamamoto, Kyoto University ## QUESTIONS? PLEASE CONTACT ## ntcadm-intent at nii dot ac dot jp ## LINKS ## INTENT-2 homepage: http://research.microsoft.com/en-us/people/tesakai/intent2.aspx INTENT-1 homepage: http://www.thuir.org/intent/ntcir9/ NTCIR-10: http://research.nii.ac.jp/ntcir/ntcir-10/index.html NTCIR online proceedings: http://research.nii.ac.jp/ntcir/publication1-en.html