In a recent post, Miles Efron proposed a distinction between different kinds of information retrieval: “macro IR” that concerns with generic tasks such as searching the web, and “micro IR” that represents more focused interaction. My sense is that one key distinction between the two is the degree to which the system represents the context of the search, and therefore is able to act on the results. Miles’ examples–finding restaurants, books, music, people–have a transactional quality about them. The system has a sufficient representation of the task to both structure the query in an appropriate manner (e.g., Yelp! metadata about restaurants) and to act on the selected result (e.g., offer to make a reservation). Macro IR, on the other hand, lacks a strong contextual representation, and leaves it to the user to act on the retrieved information.
To me, then, the principal difference between the two is not the choice of ranking algorithms or index structure (though they may differ) but rather the system’s ability to manage transitions from and to the task that motivated the information need and the search itself. Traditional conceptions of information retrieval tasks start with an information need and end with one or more documents, leaving it an exercise to the reader to both articulate the information need and to act on the results. As a generalization that enables algorithmic research, that’s a reasonable application of reductionism.
But some aspects of the real world are left out to the user’s detriment. The kinds of systems that Miles characterizes as “micro IR” make steps in the direction of helping the user to both articulate the information need and to do something useful with the results. But these examples are not necessarily limited to buying goods and services based on strong product metadata. The Blueprint example that Miles mentions uses some additional heuristics on top of full-text search to express the query (using the local context of the code for which help is needed) and to select and format results such that relevant code fragments (rather than the documents that contain them) are returned.
Another class of applications that possess some (but not all) of the characteristics of micro IR are tools such as the Remembrance Agent that derive queries implicitly from a user’s actions in some other task. Remembrance Agent did this for document creation by offering suggestions of related documents based on full text search using recently-entered text as a query. The user’s goal in that case is writing, but the system understands that writing often involves finding references, and tries to help by finding related information. We implemented this approach in XLibris by using freeform digital ink marks that readers created on documents for the purposes of making sense of their reading to identify related documents based on the the annotated passages. We found that such queries were more effective at discovering related documents than document-level relevance feedback.
Recent work on personalized search that uses a profile derived from users’ personal data to affect the search results shown to them is a weaker form of contextualization because it doesn’t represent the specific task that motivated the search, but rather some statistical average of contexts for a given user. The better a system can understand what aspects of context are relevant to the given information need, the more effective it can be at managing the transition to search. Transitions from search results back to the motivating task require additional contextual information. One approach might be to provide built-in transitions from search results to one of several probable motivating tasks (e.g., generate a reference to a found document, compare found documents, etc.). While these enhancements by themselves may not transform a macro IR interaction into micro IR, they may nudge the user experience in that direction.
These examples illustrate that incorporating context of people’s work into information seeking episodes can improve the effectiveness and efficiency of information seeking, and can allow the user to focus on the broader task without getting bogged down as much in the mechanics of information seeking.