Automated index management for distributed web search

Abstract

Distributed heterogeneous search systems are an emerging phenomenon in Web search, in which independent topic-specific search engines provide search services, and metasearchers distribute user's queries to only the most suitable search engines. Previous research has investigated methods for engine selection and merging of search results (i.e. performance improvements from the user's perspective). We focus instead on performance from the service provider's point of view (e.g, income from queries processed vs. resources used to answer them). We consider a scenario in which individual search engines compete for user queries by choosing which documents (topics) to index. The difficulty here stems from the fact that the utilities of local engine actions should depend on the uncertain actions of competitors. Thus, naive strategies (e.g, blindly indexing lots of popular documents) are ineffective. We model the competition between search engines as a stochastic game, and propose a reinforcement learning approach to managing search index contents. We evaluate our approach using a large log of user queries to 47 real search engines.

Specialisation dynamics in federated web search

“…In particular, we show that our prior experiments [14] constitute a special case of the model here under certain assumptions.…”

“…In [14], we proposed a version of best-response learning using the GAPS RL algorithm in a search engine competition scenario. In GAPS [17], the learner plays a parameterised strategy represented by a non-deterministic finite state automaton (FSA).…”

“…As shown in [14], these properties allow GAPS learners to compete successfully against various opponents with fixed strategies in the Web search game. However, we can expect that in real life multiple search providers will be adapting at the same time.…”

“…The uncertainty about competitors, changes in the environment, and the large number of competing engines make the task difficult. In prior work [14], we proposed a multi-agent reinforcement learning (MARL) approach to competing in federated Web search environments. We used a gradient-based RL algorithm to learn competition strategies from interactions with opponent engines in a simplified ad hoc scenario.…”

“…• We use the previous reinforcement learning approach to deriving competition strategies [14] to analyse the competition These experiments model competition between service providers that simultaneously and independently adapt their topic specialisation to improve individual performance (profits) based on the past experience.…”