Main Page

From Livetrix

1 PurpleSearch: Find and Present
2 Notes

[edit] PurpleSearch: Find and Present

PurpleSearch is an academic search-and-present portal based on the Library 2.0 model.

PurpleSearch is currently being developed at the Groningen University Library.

PurpleSearch can search and display enriched information from several digital and print collections such as library catalogs, digital repositories, course management systems, film and audio collections, e-journal collections and remote resources like databases for full-text or abstracts.

It is currently based on the XML interface that is a part of the Metalib package from ExLibris to search and present information from numerous scientific databases (relying on some other APIs for various additional features).

The application is available at http://purplesearch.rug.nl.
Demo accounts are available upon request by writing to: .
You can look at some screenshots for a basic idea of how the application looks.

[edit] Main Features*:

Includes a unique Database Recommender tool.

Combines "live" external search with locally stored data (for fast, relevant and constantly up-to-date information retrieval)

Post-processing of searches, including suggestions for possible alternative queries, alternative database suggestions, spelling suggestions, and more

Advanced result presentation (FRBR-like grouping and resolving to enrich results, for example by suggesting book relations)

On-the-fly lending and location information

Social features: Tagging, Reviewing and Rating

Supports interoperability of standards (like Z39.50, SRU/SRW, OAI-PMH, OpenURL, RSS, APIs) for database searches

Fully integrated harvesting and indexing tool for OAI-PMH Repositories

RSS feeds

Context-sensitive linking (OpenURL) to full-text sources

Full Unicode support for search and presentation
(Greek, Russian, Chinese, Hebrew and others have been observed in the logs and data)

Integration with existing library environments (like LBS from OCLC-PICA and ALEPH from Ex Libris or other library system vendors)

Support to direct links via a link-server to support various proxies

Integrated tiny-URL server with proxy support

Local categorization of information for subject search and display enhancements (BCL, Dewey, LCSH, Mesh etc.)

Management interface with various reports and statistics

Data robustness through automatic correction of some common XML / Unicode encoding errors.

* Some features are still in the development stage

PurpleSearch uses a combination of external "live" metasearch supplemented by local central indexing. Local indices of data material are used where such a solution is practicable, desirable, allowed/available without extra payments. In this fairly unique experiment in combining data, PurpleSearch uses the advantages of both metasearch and local central indexing and can eliminate various disadvantages of both.

Via a search federation system such as Metalib, PurpleSearch can search and index information in many external databases (we have about 200 in active use by PurpleSearch, 100 databeses are used to recommend), backed by various protocols such as Z39.50, SRU/SRW or XML.

[edit] Database recommender tool

The PurpleSearch Federated Search System has a database Recommender tool that can suggest the best databases to search and thus can create a metasearch tailored to a specific query.

PurpleSearch can recommend both online AND off-line resources.

It is currently the default to automatically choose sources based on a query.

The Recommender Tool has also been developed at the Groningen University Library.

To the best of our knowledge, PurpleSearch is the only system in the world that features such a Recommender Tool.

[edit] Details

PurpleSearch was developed in order to overcome the problems associated with database selection during searches. Most federated search engines allow searches in only a limited number of databases simultaneously. This limit changes from system to system and is partly dependent upon the system load at that moment.

Setting a system to search everything does not guarantee efficient searching either. It would resulting in an abundance of results, but since many databases focus on a specific subject area, easily results in way too many irrelevant results. This is already a problem within single databases, let alone hundreds at once. Additionally, for various resources, every search an indirect cost associated to it. Hence, minimizing expenditure by reducing searches in irrelevant databases is desired.

Federated search systems usually address this problem with pre-defined subject-based database groups. Hence, for those users that can already narrow down their search criteria to a specific subject, there is the option of directly searching in tailor-made set of databases catering to that subject. This alternative is provided in the Subject-Guided search in PurpleSearch.

But, this can sometimes be a complication. For example, given a choice between 'Arts and Humanities', 'Linguistics' and 'Psychology' a user may be unsure of which database to choose considering the selection of one database might result in the exclusion of relevant data from other databases. It was for this reason that the Recommender tool was created.

An ideal scenario would be a digital analogue to an informed librarian that responds with

"hmm, 'fertilizer as poison.' Well, definitely try PubMed since that has quite a wide selection of biological and medical articles, AGRICOLA is agricultural so that's likely useful, maybe EMBASE, and you're likely to find something in union catalogues like Picarta or COPAC."

It is unlikely for a non-human system to give an expert opinion and is also unlikely for a human to ever truly know the content of hundreds of online databases. But, it is certainly possible to point the user in a direction which is as accurate as possible.

The current implementation makes database selection transparent for users by default (by using smart word-based statistics). There is also still the option of manual selection, and selection by subject area.

[edit] Enhanced Search results

PurpleSearch enriches the search results with abundant extra information such as cover-images, thesaurus information, table of contents, impact factor information, lending information, book relations, and links to external databases such as, but not limited to, full-text.

[edit] Social features

Rating, reviewing, and tagging

Users that are logged in can add reviews to items found in any database. If another instance of the same record in a different database is visited by a user, this review can also be viewed from the new location. Note that PurpleSearch ensures that both the records are verified to be the same to a 'reasonable' degree. The bookmark tagger allows browsing and searching in a user's own bookmarks, and allows them to add a rating, review, and tags.