Personalized Keyword Search Related...
-
Upload
truongkhue -
Category
Documents
-
view
217 -
download
0
Transcript of Personalized Keyword Search Related...
Introduction
Georgia Koutrika, Alkis Simitsis, Yannis E. Ioannidis: Précis: The
Essence of a Query Answer. ICDE 2006
Kostas Stefanidis, Marina Drosou, Evaggelia Pitoura: PerK:
Personalized Keyword Search in Relational Databases through
Preferences. EDBT 2010
Personalized Keyword Search
Keyword search
Enables users to find information without knowing the database schema
or complicated queries
May return a huge volume of data
Benefits of personalization
Enables returning personalized results to individual users or groups
Enables returning smaller number of results that are also more relevant
to the user(s)
Personalized Keyword Search
Personalization can be implemented in several ways:
Weights assigned to the edges of the database graphs (Koutrika et al.
2006)
Pairwise preferential ordering (Stefanidis et al. 2010)
Q = {western}:
cp1 = ({western}, C. Eastwood ≻ J. Wayne)
MOVIE GENRE
(MID) 0.9
(MID) 1.0
C.
Eastwood
J. Wayne
Sources
arXiv.org (arxiv.org)
Google Scholar (scholar.google.com)
Google (www.google.com)
Keywords: keyword, keyword-search, keyword-based, personalized,
personalization, database, query, search, user profile, preferential,
preferences, ordering
Current Trend: Online Services
Personalization in search engine results
Personalization through browsing history
Search Personalization using Machine Learning. H. Yoganarasimhan. 2014.
http://faculty.washington.edu/hemay/search_personalization.pdf
The Internet of Things
Geo-location used as an external context to offer personalization
Current Trends: Cloud Services
Ranked keyword search for encrypted data
A Novel Approach for Secure and Dynamic Multi-Keyword Ranked
Search Scheme over Encrypted Cloud Data. V. Guguloth, P. Reddy, V.
Pasunuri. 2016. http://ijitech.org/uploads/362451IJIT10110-245.pdf
Enabling Fine-grained Multi-keyword Search Supporting Classified Sub-
dictionaries over Encrypted Cloud Data. H. Li, Y. Yang, T. Luan, X. Liang,
L. Zhou, X. Shen. 2015.
http://www.ieeeprojectsmadurai.com/Basepapers/enabling%20fine.pdf
Preferred Keyword Search over Encrypted Data in Cloud Computing. Z.
Shen, J. Shu, W. Xue. 2013.
https://www.researchgate.net/profile/Zhirong_Shen2/publication/26115
2238_Preferred_keyword_search_over_encrypted_data_in_cloud_comp
uting/links/555c45c108aec5ac2232afc3.pdf
Current Trends: E-Commerce
Keyword search and personalization of results in e-commerce
product searches
A Modern Approach to Integrate Database Queries for Searching E-
Commerce Product. A. Siddiqui, M. Muntjir. 2016.
https://www.researchgate.net/publication/302393360_A_Modern_Appr
oach_to_Integrate_Database_Queries_for_Searching_E-
Commerce_Product
Supporting Keyword Search in Product Database: A Probabilistic
Approach. H. Duan, C. Zhai, J. Cheng, A. Gattani. 2013.
http://www.vldb.org/pvldb/vol6/p1786-duan.pdf
E-Commerce and Product Search
Growing field
Multiple product categories and a huge volume of data
Keyword search simplifies searches and produces accurate results
nVidia 960m
Computers Accessories Displays
Desktops
Laptops
Accessories
Displays
Cables
Bags
Speakers
Microphones
Other
<13’’ displays
13’’-15’’ displays
>15’’ displays
E-Commerce and Product Search- Duan et al. 2013
Vocabulary gap between the specifications of products in the
database and the keywords people use in search queries (H. Duan et al.
2013).
Probabilistic model to find and rank the most relevant items for generic queries
new computer
asus rog i7 win10 nvidia geforce 960m
laptop with dedicated graphics card
Additional Reading
Thesis: Beyond Keyword Search: Representations and Models for
Personalization. K. El-Arini. 2013.
http://www.cs.cmu.edu/~./kbe/kbe_thesis.pdf
Additional Sources
sciencedirect.com
Leading information solution for researchers, teachers, students, health careprofessionals and information professionals . It is the world's largest electroniccollection of science, technology, and medicine full-text and bibliographicinformation. It offers a wide variety of features and content.
acm.org
The world's largest educational and scientific computing society, deliversresources that advance computing as a science and a profession. ACM providesthe computing field's premier Digital Library and serves its members and thecomputing profession with leading-edge publications, conferences, and careerresources.
Ieeexplore.ieee.org:
Provides web access to more than four-million full-text documents from some ofthe world's most highly cited publications in electrical engineering, computerscience and electronics.
Current Trends: Overview
Social Media ( Twitter, Facebook, LinkedIn, etc)
Items and user context data are highly dynamic: Real-time scoring in personalized keyword search.
Implicit feedback techniques based on user's interactions to automatically capture user’s interests.
Personalized keyword search and semantic approach.
User information needs are typically rather complicated.
Personalized keyword search on XML documents.
Current Trends: Social Media
Real-time scoring in personalized keyword search.
F. Borisyuk, K. Kenthapadi, D. Stain, B. Zhao: "CaSMoS: A Framework for Learning CandidateSelection Models over Structured Queries and Documents " . 2016
Propose a machine learned candidate selection framework to prune irrelevantdocuments.
LinkedIn: Web Platform to find people, jobs, companies, groups and other professional content.
Results in real time, must offer a high degree of data freshness, and must respond with low latency.
Search engines are known to be good at retrieving a small set of relevant documents matching agiven query out of a huge document set.
Retrieval process in two stages: 1.First invoke a computationally inexpensive model to select asmall candidate. 2. Perform a more computationally expensive second pass scoring and ranking onthe obtained candidate set.
Current Trends: Social Media
Real-time scoring in personalized keyword search.
Y. Cheng, Y. Xie, Z. Chen, A. Agrawal, A. Choudhary: "JobMiner: a real-time
system for mining job-related patterns from social media " .Proceedings of the 19th
ACM SIGKDD international conference on Knowledge discovery and data mining,
2013.
LinkedIn: Web Platform to find people, jobs, companies, groups and other
professional content.
JobMiner: providing a better understanding of the dynamic processes in real
time in this kind of platforms and in that way obtaining a more accurate
identification of important entities in the job market.
Current Trends: Implicit feedback
techniques
An explicit request of information to the user implies to burden the user, and
to rely on the user’s willingness to specify the required information.
Sometimes this scenario is not realistic.
Several techniques have been proposed to automatically capture the user’s
interests: They are based on the collection and the analysis of: browser
history, query history, click-trough data, desktop information, document
display time, bookmarks, and several other kinds of signals.
Current Trends: Implicit feedback
techniques
G. Pasi: "Implicit feedback through user-system interactions for defining usermodels in personalized search", 2014.
Implicit feedback techniques to collect user interests and preferences via user-systeminteractions are presented.
k. Gao, X. Dong: "A Context-awareness Based Dynamic Personalized HierarchicalOntology Modeling Approach", 2016.
Developing a dynamic updated ontology model based on capturing the implicitcontext information and implementing the personalized scalability demands of contextmodel for different users.
E. Nunez, J. Cueva, O. Sanjuan, V- Garcia, "Implicit feedback techniqueson recommender systems applied to electronic books", 2012.
Current Trends: Semantic approach
Personalized keyword search and semantic approach.
J.Li, D. Arja,V. Ha-Thuc,S. Sinha: "How to Get Them a Dream Job? Entity-
Aware Features for Personalized Job Search Ranking",2016.
Traditional information retrieval systems, and particularly Web search engines,
have focused on keyword matching.
LinkedIn: Web Platform to find people, jobs, companies, groups and other
professional content.
It is important to understand the structure of the information.
Current Trends: Semantic approach
Personalized keyword search and semantic approach.
M. Tvarožek , M. Bieliková: "Personalized Exploratory Search in the Semantic Web",
2010.
Effective access to information on the Web is being hampered by informationoverload, unavailability of information, navigation issues and user diversity.
Goal: view generation with incremental graph visualization to enable end-
user grade exploration of semantic web content.
Current Trends: Personalized Keyword
Search on XML documents
B. Kurinchi Meenakshi 1 , Dr.C.Nalini, "Managing XML Retrieval throughPersonalization Using Search Engine". International Journal of Innovative Researchin Computer and Communication Engineering, 2015.
Improving the search of XML documents according to the user requirement andpreference.
X. Liu , L. Chen, Ch. Wan, D. Liu, N. Xiong, "Exploiting structures in keyword queriesfor effective XML search ", 2015.
Focused on the query side of XML keyword search.
G. Li, C. Li, A. Li, J. Feng, A. Li, L. Zhou: "SAIL: Structure-aware indexing for effectiveand progressive top-k keyword search over XML documents", 2011.
It develops a novel method, called SAIL, for efficient XML keyword search.
Additional Reading Personalization & Search Engine Rankings. Search Engine Land.
http://searchengineland.com/guide/seo/personalization-search-engine-rankings
Advantages and Disadvantages of Personalized Search, RAJIB KUMAR,
http://www.techncom.net/2012/03/advantages-disadvantages-personalized-
search/
scholar.google.com
Personalized Task Recommendation in Crowsourcing
Current Trends: Personalized search
Personalization & Search Engine Rankings.
Search result was the same years ago # The search result of today are
personalized.
Locality
CountryPersonal History
Social Connections
Country
Someone in the US searching for “football” will get results about American football; someone in the UK will get results about the type of football that Americans would call soccer.
Locality : Search engines tailor results to match the city.
Personal History : Used by SEO (Google, Bing)
What has someone been searching for and clicking on from their search results?
What sites do they regularly visit?
Have they “Liked” a site using Facebook, shared it via Twitter ?
Personalized History therefore takes personalization to an individual level.
Social Connections
It is based on someone’s friends behaviour on Social network i.e.; What does your friends think about
a website ?
Search engines view those connections as a user’s personal set of advisors (As it happens offline).
Elements of Personalization & How to Perform Better in PersonalizedSearch .
In this articles, more detailed used by SE to personalized a search are added.
Location : where is the searcher
Device : What type of devices or OS used
Browser: Search history, potentially what you’ve clicked on in the
results.
Email calendar: In the case you are using Gmail, Google
calendar, Google will use the data for potential search result
Google+
Visit history
Bookmarks :
Logged-in visitors
If we logged-out the search the results of a search will appears differently still they are many
ways of personalization. A fresh and unused window will show a different result than those of
the browser that we use everyday.
Performing better personalized search :
Get potential searchers to know and love your brand before the query
Be visible in all the relevant locations for your business
Get those keyword targets dialed in
Share content on Google+ and connect with your potential customers
Be multi-device friendly and usable
Advantages and Disadvantages of Personalized Search.
The starting point of personalized search was the launch of the
‘Search plus Your World’ program by Google.
The Goal was to enable people with who you interact within
your community could have a direct influence on the search
results shown to you.
As a result nowadays, it is impossible to ignore personalization in
search engine query
Advantages of Personalized Search.
Personalized make search result centered to the users needs and context of use.
It had increase the creation of high quality, interactive content for the benefit of the users .
Disadvantages of Personalized Search.
Personalized restrict the users scope of search and force to operate certain limitations, therefore
limit the users to discover new result outside his circle.
In the case of Google’s “Search Plus Your World”, personalization of the user’s search does not
include other social media sites as Facebook, Twitter.