Design and implementation of COSIMA-a smart and speaking e-sales assistant

Abstract

We present a new cooperative user interface for e-shopping in B2C e-commerce. COSIMA is a smart and animated Internet avatar with synchronized dynamic voice output who assists customers through their e-shopping tours and advises them in the spirit of a real salesperson in the old economy. COSIMA benefits from a Preference SQL-based search engine that relies on the Pareto-principle, computing best matching results to the customer's wishes. This enables COSIMA to offer ideal hits or suitable alternatives with one single query, to value these search results in terms of the stated preferences and to generate proper voice output following basic principles of sales psychology. As a sample application we present our meta-search engine for comparison shopping. Since COSIMA constitutes an entire technological framework, it can easily be adjusted to other e-commerce environments, including mobile commerce, by customizing the search engine, the avatar animations and the speech contexts.

Preference SQL — Design, Implementation, Experiences

“…Feedback gathered so far and a growing community of many hundreds of user strongly indicate that this is a promising path to pursue for next generation e/m-commerce systems. Technically, the results in [KFH01] give strong evidence that Pareto accumulation is an indispensable operator when dealing with soft constraints. E.g., predominantly the size of the Pareto-optimal set was between 1 and 20, yielding an easy-to-survey choice of products similar to a traditional sales situation.…”

Querying with Intrinsic Preferences

“…However, saying simply that I prefer a lower price on a book (as we do in Example 1) does not seems possible in Preference SQL. The description of Preference SQL in [17] is so brief that a detailed comparison with our proposal is not possible at this point. [3] introduces the skyline operator and describes several evaluation methods for this operator.…”

“…It encompasses preference logics [23,19,13], preference reasoning [24,22,4], prioritized nonmonotonic reasoning and logic programming [5,6,21] and decision theory [7,8] (the list is by no means exhaustive). However, only a few papers [18,3,11,2,14,17] address the issue of user preferences in the context of database queries. Two different approaches are pursued: qualitative and quantitative.…”

“…Two different approaches are pursued: qualitative and quantitative. In the qualitative approach [18,3,11,17], the preferences between tuples in the answer to a query are specified directly, typically using binary preference relations. Example 1.…”

Agent mediated retailing in the connected local community

“…Rea has been demonstrated acting as a real estate sales advisor. Kießling et al (2001) describe COSIMA, a sales support system presented by a speaking synthetic persona in the form of a young woman. The COSIMA system performs partial match searches from product databases, and generates conversational output based on the completeness of the match found.…”