Text Analyzer

Concept:

Upload your own document to search JSTOR. Text Analyzer finds the topics discussed in your document and recommends articles and chapters about the same topics from JSTOR.

What we did:

1. During a design sprint exploring ways that JSTOR could support scholars and their preprints, we tested the concept of allowing users to upload their nearly-finished papers to find additional content. Users were exceptionally positive about both the idea and an early functional prototype.
2. Having validated the concept, we designed the tool, incorporating into the design what we'd learned building "the equalizer." For more information on the technology that underlies the tool, see this blog post.
3. We released Text Analyzer as a beta tool on the primary JSTOR website and, as users began to find it valuable, we went through multiple cycles of improvement and promotion.
4. We continiued to improve the algorithm that powers it throughout, adding one major new feature: the ability to submit texts in fifteen different languages and get results in English.

What we built:

Reactions to Our Work:

We collected reactions to Text Analyzer on social media, published media and in (anonymized) email.  We used the late, great Storify to retain these, but when that site went dark, created our own repository:

• February and March 2017
• April 2017
• May 2017
• Summer 2017
• September 2017
• October 2017
• November 2017
• Winter 2017-18
• February and March 2018
• April 2018

What we learned:

1. Searching using a document can help junior researchers who are "keyword thrashing," i.e. looking for the right term in a keyword search.
2. Searching using a document also helps established researchers to find relevant material outside of the specific disciplines or fields they work in.
3. Driving adoption of a new means of search takes partnership with librarians and faculty teaching research methods.