Constructing and exploring composite items

Abstract

Nowadays, online shopping has become a daily activity. Web users purchase a variety of items ranging from books to electronics. The large supply of online products calls for sophisticated techniques to help users explore available items. We propose to build composite items which associate a central item with a set of packages, formed by satellite items, and help users explore them. For example, a user shopping for an iPhone (i.e., the central item) with a price budget can be presented with both the iPhone and a package of other items that match well with the iPhone (e.g., {Belkin case, Bose sounddock, Kroo USB cable}) as a composite item, whose total price is within the user's budget. We define and study the problem of effective construction and exploration of large sets of packages associated with a central item, and design and implement efficient algorithms for solving the problem in two stages: summarization, a technique which picks k representative packages for each central item; and visual effect optimization, which helps the user find diverse composite items quickly by minimizing overlap between packages presented to the user in a ranked order. We conduct an extensive set of experiments on Yahoo! Shopping 1 data sets to demonstrate the efficiency and effectiveness of our algorithms.

Composite retrieval of heterogeneous web search

“…In addition, they are experimented mostly on homogeneous environments and their applicability for general heterogeneous web search is not obvious. Within the shopping domain, [5] explored retrieving bundles of items that are subject to budget and compatibility (e.g. an iPhone and all its accessories).…”

“…We introduce a different method, suggested by the observation that established vertical selection methods are prone to noise interference, and inspired by the approach to composite retrieval described in [5]. The basic idea is that we first produce bundles in a central vertical using BOBO and then we attach items from other satellite verticals onto the produced bundles.…”

Shortlisting top-K assignments

“…Set I indicates pairs that must belong to the two assignments, and O includes pairs that cannot participate in them. The root contains the optimal and the second best matchings, A1 and A2, computed by two executions of SSPA 2 . Let e be a pair that is chosen in (i.e., belongs to) A1 but not in A2.…”

“…Although not a work on assignment computation, [2] addresses a problem of some relevance to ours, in the sense that it also needs to search through numerous combinations. Consider the purchase of a mobile device, for which a number of accessories (e.g., speakers, batteries, etc) are available, each at a different price.…”

Heterogeneous Information Access Through Result Composition

“…Other works have focused on building systems that perform composite retrieval under a number of different semantics and targeting specific application domains, such as on-line shopping or travel planning (Basu Roy et al 2010;Xie et al 2010). …”