Where We're Headed With Bento-Box Search

Kevin Beswick, NCSU Libraries

@kbeswick

What is Bento-Box Search?

What isn't Bento-Box Search?

The ultimate, robust discovery solution
A search problem at all?

How Do We View Bento-Box Search?

An opportunity for instruction
A shortcut to known items or other search tools
A recommendation problem

Instead, we think of bento-box search (yes, I know I keep calling it search) as an instructional opportunity. We can show users that are unfamiliar with the library what different search tools we offer and the types of resources we offer. We recognize that many of the search tools we use are tailored to the type of content they search, and attempts to re-create them or abstract into a blended model will lead to loss of data or functionality in a lot of cases. Bento-box search acts as a gateway into these tools, and offers a shortcut for known item searches. Bento box search can also be viewed as a recommendation problem. Given a query and anything else we know about the user, what type of resource do we think could best help them? In a lot of cases, we have less control over search results from specific services, but we do have control over the way we present them to users.

QuickSearch

NCSU's bento-box search solution
Initially created in 2005
Rewritten in 2013 as a Ruby on Rails project

quick_search (or quick_search-core)

https://www.github.com/NCSU-Libraries/quick_search

Open source version of QuickSearch released in 2016
A toolkit for easily creating bento-box applications
Extracted from NCSU's version

Features

Rails Gem Engine - easy to install
Modular & extensible
- Searchers
- Themes
Statistics (local & GA), Best Bets

Installation

Create a new Rails project
Add 'quick_search-core' gem & bundle install
Run "rails g quick_search:install"

Searchers

Plugins for QuickSearch that search an external service
Small amount of code to implement
Automatically hooks into QuickSearch infrastructure for stats collection, etc.

Available Searchers

arXiv
OpenLibrary
Wikipedia
Summon
ContentDM (contributed by Temple University)

Statistics

QuickSearch automatically collects usage data for all searchers

Queries (query text)
Clicks (which box, which part of the box, etc.)
Sessions (on/off campus, mobile/desktop, length)

Statistics Visualization

Updated interface for viewing stats to be more visually appealing and interactive
Undergrad student at NCSU has been working on it, almost ready to merge into QS core
Work happening in the open: https://github.com/rbwest/quick_search/tree/AppVizClean

https://www.youtube.com/watch?v=vFYnujsvRR0

Near-term features

Typeahead - backend support already there, frontend needs to be generalized & migrated from NCSU QS

NCSU Implementations

NCSU QuickSearch

Our original QuickSearch
Now running on open source gem in production since Jan.
Working on extracting/generalizing last features into open source gem

Historical State Search

Special Collections focused bento-box search
Originally developed as an entirely separate application
Migrated to QuickSearch core platform by Nushrat Khan
Launched into production in Mar. 2017

Historical State Search

Planning on releasing as open source as an example project that uses QuickSearch
Also planning to release new searchers for lunr.js, and other services

Next Steps

More library content:
- Events
- Spaces
- People
- Services

Next Steps

Intelligence:
- Show results from more specialized sources depending on subject
- Use collected data for recommendations
- Promote library services
- Context-aware - device type, on/off campus

We're also thinking of how to make bento-box search more intelligent. I mentioned before that I view bento-box search as a recommendation problem. We're interested in how we can show results from more specialized sources depending on what people are searching for. We're experimenting with using the data we've been collecting around sessions, searches and clicks to provide recommendations of library services to users in an effort to promote the fact that libraries are a place for more than just books and articles. We're also thinking about making bento-box search more context aware, and provide more relevant recommendations for someone based on factors that we can detect like whether they are on desktop or mobile, or whether they are on campus or off campus.

What could this look like?

Growing the QuickSearch community

We're interested in working with people that are interested in using QuickSearch
Help installing, customizing, creating new searchers
Goal is to have a large collection of searchers for popular library services, and more examples of applications and themes.

We want to work with you!

kdbeswic@ncsu.edu

@kbeswick

https://www.github.com/ncsu-libraries/quick_search

https://search.lib.ncsu.edu/