Research on Personalized Information Retrieval Based on User p s New Interest Detection

An im por tant characteristic of next generat ion sear ch engine is perso nalizat ion. Perso nalized infor matio n retrieval (PIR) f ocuses on users. It capt ures usersp int er est in dif ferent kinds (ex plicit , implicit int erest and interest of sim ilar users). T hese inf orm at ion of users are int egrat ed and used t o improv e the result o f inf ormat ion ret riev al sy st em. Perso nalized inf ormat ion ret riev al can g rasp t he usersp retr ieval int ent io n and f ind personalized result s. T he aut ho rs propose t he new int erest det ect io n t ask, which identif ies t he queries cont aining usersp new ret rieval int erest by t he chang e of ret rieval object. Sim ult aneo usly, by using and impr oving t he T ex t T iling algo rithm, the retrieval sy st em is enabled to auto mat ically choo se the appropr iat e dynamic t hreshold and det ect the chang e of usersp interest. T he ret riev al inf ormat ion and labeled answ ers of user s are used t o establish t he ex perimental dat aset. T he evaluat ion mat rix includes f alse alar m rate, m iss alarm rat e, and cost o f det ect ion. In t he experiment of personalized infor mation r et rieval syst em, the improved T ex tT iling algo rithm improves t he new int erest det ect io n sy st em by 16. 4%. Whatps more, t he new int er est det ect ion t ask improves t he perf orm ance of t he perso nalized inf orm at ion ret rieval sy st em is by 3. 8%. T he ex perim ent show s t hat mining usersp interest w it h t his method can decrease the false inf ormat ion in usersp mo dels and improv e the result of pr ecision of usersp int er est det ect ion. Key words personalized inf orm at ion ret rieval; user sp new int er est detection; T ext T iling algorit hm; dynam ic t hreshold; sear ch interest chang e 个性化信息检索可以根据用户的检索兴趣返回个性化的检索结果. 提出了用户新兴趣发现子任 务, 根据用户检索对象的变化识别包含新检索兴趣的查询. 同时, 引入 …