Common use of Motivation Clause in Contracts

Motivation. While there is a history of strong research on efficiency issues in the IR community, much attention has focused on a few established and often narrow problems and setups motivated by standard IR architectures and systems, leaving many other cases largely unexplored. There have also been a number of emerging changes in modern search systems that call for new directions and approaches. In particular, efficiency researchers need to look at issues related to Multi-Stage Search Systems (MSSs), which are increasingly being deployed with machine-learned models, and examine the end-to-end performance of methods under MSSs. There are also opportunities for further efficiency improvements that require the application of machine learning and data mining techniques, including methods that learn index structures, learn how to optimize queries, or that estimate query distributions and optimize for these. There might also be ways to completely bypass the currently used index structures via neural nets, structures from the Combinatorial Pattern Matching community, or FPGA-based systems. Finally, new emerging IR applications also require attention. In summary, it is time to reconsider some basic assumptions, and to step away (at least partially) from the comfortable world of inverted lists and simple ranking functions that so many (often strong) papers have addressed.