CMSC 676 Personalized Information Retrieval Meera Patel - PowerPoint PPT Presentation

CMSC 676 Personalized Information Retrieval Meera Patel

Motivation Searched “FB side effects”... FB (in medical terms): Foreign Body

When the user writes a The user expresses the required information in query terms and expects query... semantic content of document What information Retrieval returns? “Documents matching to query terms” , which might not contain the expected result… Then what’s the solution? Personalized Information Retrieval

The Basic idea of Personalization...

“Personalization” Approach: Detecting document genre [1] Based on three aspects: ● Familiarity ● Genre ● Geography

H1: People with low familiarity with a topic will Formulating prefer documents which have a high proportion of concrete terms, and conversely, people with high familiarity with a topic will prefer documents that have a high proportion of abstract terms. Run_score = baseline_score + H2: Documents of a given genre will include Σ w i *metadata_score i terms characteristic of that genre, which can be mathematically modeled so as to differentiate the genre of documents retrieved on a specific topic. where w i = the constant weight of the metadata_score H3: The vocabulary of documents is specific for the geographic area to which they refer: US documents can be distinguished from non-US documents by virtue of vocabulary characteristics.

Another approach of Personalization [2]

Future work Clustering of documents based on semantic contents can make things easier, eg. reranking can be done based on the intersection of user interests and cluster semantics Enhancing the disambiguation of natural language techniques may help in finding more relevant results Adding one more phase in IR simply increases the delay in retrieval, so further work should be focused on faster retrieval

References [1] G. Muresan, C. L. Smith, M. Cole, Lu Liu and N. J. Belkin, "Detecting Document Genre for Personalization of Information Retrieval," Proceedings of the 39th Annual Hawaii International Conference on System Sciences (HICSS'06), Kauia, HI, USA, 2006, pp. 50c-50c. [2] The Personalized Information Retrieval Model Based on User Interest by Songjie Gong [3] Personalized Information Retrieval Approach by Myriam Hadjouni, Mohamed Ramzi Haddad, Hajer Baazaoui, Marie-Aude Aufaure, and Henda Ben Ghezal [4]C. Bouhini, M. Géry and C. Largeron, "Personalized information retrieval models integrating the user's profile," 2016 IEEE Tenth International Conference on Research Challenges in Information Science (RCIS), Grenoble, 2016, pp. 1-9. [5] Towards an architecture for personalized information retrieval implying the user’s profile and votes by Harbaoui Azza , Sahbi Sidhom , Malek Ghenima, Henda Ben Ghezala

Thank you... Any Questions?