Rishiraj Saha Roy (IIT Kharagpur), Monojit Choudhury and Kalika...
1
Structural Aspects Mean length of queries is steadily increasing Queries are bags of (possibly multiword) units gprs config | nokia n93 | telstra australia Analogous to natural language, query units can be said to be of two types – content (harry potter, united nations) and function (download, general information) units Like NL, queries have a kernel and peripheral lexicon Similarities with assumed protolanguage structure – Short sequences of symbols Ad hoc sentences Simple or no grammar Large number of symbols (?) Introduction Web users communicate their information need to a search engine through queries. Queries have a structure far simpler than NL, but more complex than the commonly assumed bag-of-words model. In fact, Web search queries define a new and fast evolving language of its own, whose dynamics is governed by the behavior of the search engine towards the user and that of the user towards the engine. Are Web Search Queries a self- organizing system of language? Is it an evolving protolanguage? Contact: [email protected], [email protected], [email protected] Rishiraj Saha Roy (IIT Kharagpur), Monojit Choudhury and Kalika Bali (Microsoft Research Lab India) Dynamics of Web Search Queries Search engines evolve at two levels – algorithms and models Algorithmic evolution: More sophisticated algorithms for ranking & retrieval; analogous to biological evolution Model evolution: information that the search system learns from user interactions through machine learning and uses to present better results in the future; similar to cultural transmission of NLs Functional Aspects Hockett’s Design Features Homogeneity and Symmetry Image source: http://shishikwena.blogspot.in/ 2010 2006 2001 Image source: http://www.pep-web.org/
Transcript of Rishiraj Saha Roy (IIT Kharagpur), Monojit Choudhury and Kalika...
Structural Aspects
Mean length of queries is
steadily increasing
Queries are bags of (possibly
multiword) units
gprs config | nokia n93 | telstra australia
Analogous to natural language, query units
can be said to be of two types – content (harry potter, united
nations) and function (download, general information) units
Like NL, queries have a kernel and peripheral lexicon
Similarities with assumed protolanguage structure –
Short sequences of symbols
Ad hoc sentences
Simple or no grammar
Large number of symbols (?)
Introduction Web users communicate their information need to a search engine through queries. Queries have a structure far simpler than NL, but more complex than the commonly assumed bag-of-words model. In fact, Web search queries define a new and fast evolving language of its own, whose dynamics is governed by the behavior of the search engine towards the user and that of the user towards the engine.
Are Web Search Queries a self-organizing system of language? Is it an evolving protolanguage?
Contact: [email protected], [email protected], [email protected]
Rishiraj Saha Roy (IIT Kharagpur), Monojit Choudhury and Kalika Bali (Microsoft Research Lab India)
Dynamics of Web Search Queries
Search engines evolve at two levels – algorithms and models
Algorithmic evolution: More sophisticated algorithms for
ranking & retrieval; analogous to biological evolution
Model evolution: information that the search system learns
from user interactions through machine learning and uses to
present better results in the future; similar to cultural
transmission of NLs
Functional Aspects
Hockett’s Design Features
Homogeneity and Symmetry
Image source: http://shishikwena.blogspot.in/
2010
2006
2001
Image source: http://www.pep-web.org/
