| Technology |
|
The Vector Method | | | The "OneBox" Model | | | Social Methods |
|
The "OneBox" Model
The OneBox model was designed to deliver a range of specialized results through a single search box. When the user submits a search query, the OneBox engine passes the terms to a number of third party subsystems, each one focusing on a particular information niche, such as a particular database or file system, for example. It then displays the results from the various subsystems alongside its own search results. This means that even for a query as specific as a product code, the user need only type into one search box. Behind the scenes the search engine will pass the query to a subsystem provided by the manufacturer, and then deliver the resulting information itself. In theory, this saves the user time and effort but there are important disadvantages to this approach that should be considered.
Information Overload
OneBox is actually a somewhat unsophisticated approach that aims to circumvent some of the problems associated with keyword search. Since the engine has no way of knowing the conceptual origin of the search query, it systematically processes the terms through every possible information context available to it in the hope that one of them will correspond to the context of the search query. Consequently, this approach risks overloading the underlying repositories with queries going to each individual subsystem when only one conceptual query would have been necessary. This represents a significant obstacle to scalability in the context of a global enterprise solution.
Inaccurate Results
Additionally, each of the third party subsystems use a different scoring algorithm to determine the relevance of results. This means that in identifying the result that is most relevant to the query, the search engine receives conflicting reports from each of the subsystems. Unfortunately, for this reason OneBox cannot be relied upon to yield the most accurate result every time.
Autonomy's Approach
In contrast, Autonomy uses conceptual retrieval technology to identify the correct result based on the context supplied by the user. Due to its unique meaning-based approach, the IDOL Server is able to deliver comprehensive search over each of the underlying repositories with better accuracy and reliability than can be achieved through their own intrinsic search capabilities. It searches throughout the entire index to identify the search terms in the desired context, without impeding the performance of the underlying repositories whatsoever. If the user has a specific reason to process the query through a third party subsystem, Autonomy is able to offer Automatic Query Federation (AQF) technology. This module federates the query to a system it has identified as the correct specialist, based on its conceptual understanding of the query. For example, if a user enters a shipment tracking number, IDOL recognizes from the form of the query that the shipping company will be most likely to return the correct result, and only federates the query to that system. In this way, unlike the OneBox approach, IDOL avoids bombarding all of the systems with the same query, and simply selects the most likely one to return the specialist result.
The Evolution of SearchAutomatic Query Federation (AQF)| Technology |
|
The Vector Method | | | The "OneBox" Model | | | Social Methods |
|
