Você está na página 1de 2

Editors Desk

Refining a Search vs. Beginning a New Search


Users rarely begin with the most appropriate term or phrase for their search. More often than not, they do not spend any time in pre-search preparation and even do not bother to make presearch settings of the system. Popular search engines have to be blamed for encouraging this habit of straight away starting search without giving time to think about what is required. They instantly provide some results and as such no proper assessment of results against requirements is made. This kind of searching is okay for hurriedly getting a piece of information or a bit of data or for searching a known item, but certainly not appropriate for scholarly/research pursuit.

Scholarly users need some online or personal help and guidance to steer through their search by refining their search strategy by what they started with. True, users often mindlessly abandon and begin a new search on the same topic where refining a search is called for. It is said that a typical search will have up to 3 terms either ANDed or ORed depending on default configuration of the search engine. More than half of them never get modified beyond this stage mainly due to retrieval of some hits and simple or straight nature of the search. This way, many users partially implement search strategies and miss opportunity to narrow down or broaden their search. Query reformulation or refining search strategy requires some skill, expertise and experience on the part of searcher or some assistance of an intermediary. Refining a search involves (i) having additional information/knowledge about the concepts, (ii) changing the settings and Boolean operators, (iii) dropping or/and adding search terms, (iv) substituting related terms from thesaurus, (v) deciding the desired number of hits, (vi) knowing the postings data for each search term, etc. Users generally go for three broad types of searches - high recall, high precision or a brief search for few relevant references. Often users land up in either too many hits or too few hits.
Tips for narrowing search to increase precision: Generally precision devices associate terms to increase the shades of meaning by coordination (including pre-coordination), linking of terms, role/ relational indicators and term weighing and narrow down the search. i. Explore using Boolean AND operators in the search strategy ii. Choose a more specific narrower term/s (see thesaurus) iii. Explore using subheadings restricting search to a sub-topic iv. Use the features like edit or limit search results (article type/publication type, etc.) v. Combine "Review" with Boolean AND to limit to review of the literature so as to have only those with pre-evaluated, pre-synthesized, etc. vi. Filter by year of publication

M S Sridhar

Page 1

Tips for broadening search to increase recall: Recall devices group terms together into classes of one type or another with control of synonyms, quasi-synonyms and of word forms. Hierarchical grouping and grouping by statistical association are other two recall devices which help to broaden the search. 1. Explore using Boolean OR in search strategy 2. Add more synonymous search terms (examine records retrieved) 3. Explore to include broader term/s (see thesaurus) 4. Explore to include more than one narrower term of a concept (see thesaurus) 5. Use truncation (* or ?) in search terms and search on a root word/s (beware of false hits!) 6. From a relevant article, look for related articles 7. Use citations of relevant articles retrieved to find similar results (snowball technique)

Beginning a new search is indeed a difficult decision. One should explore this option only after considering the reformulation options like narrowing down the search or searching the subset of records (nested search), filtering, clustering, searching with more documents like this one, etc., otherwise one will end up with an array of new searches. Ideally, decision to begin a new search should take place only when the user desires to search on a new specific topic. But in reality a lot of users abandon searches to begin a new search on the same or related topic ignoring the provisions like lateral search and filtering. Yet, lateral search cannot substitute for clustering and faceted search or guided navigation. Some advanced systems have interactive capabilities to help query reformulation. The possibilities of adopting citation pearl growing (similar to snowball technique), bibliographic coupling and co-citation analysis are quite rare though horizontal/vertical browsing for taking to TOC or specific item in the series are common.

It is highly desirable to have a number of hits for each modified search stored and made available against click to see search history so that the user gains confidence that he is successively refining the search in the right direction. In addition to search history, above mentioned filtering, clustering and faceted search/navigation are the value additions an information retrieval system can provide end users to refine their search. Each of these features enable users to narrow down their search to more specific information and also sift and steer though large results retrieved. Obviously they are post-search value additions badly required in these days of enormous recall and bewildering set of results. As far as refining search in J-gate is concerned, users can (i) begin a lateral search on a related topic (ii) narrow down the query by restricting categories, filtering, or adding terms with Boolean AND (ii) increase recall by grouping synonymous terms with Boolean OR. Many of the above value additions/features are in pipe line for the next version of J-Gate. M S Sridhar sridhar@informindia.co.in
M S Sridhar

Page 2

Você também pode gostar