Today ProQuest is pushing out a big update to Summon, our discovery layer. Below is a summary of the changes (with some slight commentary from me). You can read the detailed release notes on ProQuest’s site. In the new year, KARMS and I will be looking over the new options available for recommending items, and we’ll be reaching out to folks in R&I and UX to help us improve our users’ search experience. If you have any questions, let me know!
Coming December 15, 2016
ProQuest is making the following changes to improve performance and usability with Summon 2.0:
We have added new Library of Congress Subject Headings to query expansions so that “variants” (as specified by the LCSH) are now used to expand queries. For example, a search for “bionanotechnology” is now expanded to include results for “nanobiotechnology” since “bionanotechnology” is a variant of “nanobiotechnology”. Or “science of language” is now expanded to include results for “linguistics” since “science of language” is a variant of “linguistics.”
Another new feature included in this release is the “context-sensitive” expansion feature. LCSH uses the notion of “qualifying terms” indicated in parentheses. For example, “optimization (linguistics)” is defined as a variant term for the heading “optimality theory” with the qualifying term “linguistics”. This means the term “optimization” used in the context of “linguistics” is a variant for the heading “optimality theory”. The new feature looks for both the variant term “optimization” and the qualifying term “linguistics”, and it does the expansion only if both terms are present in the query.
- optimization >> no expansion
- optimization engineering >> no expansion
- optimization linguistics >> search expanded to include results for “optimality theory”
- linguistics optimization >> search expanded to include results for “optimality theory”
Database Recommender Improvements
Any databases you have rights to will be made available for recommendation by default. (Previously only databases that were included in the Summon index were available for recommendation.)
Topic Explorer Relevance Improvements
- Better handling of close matches for longer phrases, to eliminate false matches; for example: Previously the topic “stress in the workplace” would trigger a match for a search for either “_______ in the workplace” or “stress in the ______”, so that “lunch in the workplace” or “stress in the joints” would result in the matching topic “stress in the workplace”; phrase matching (having 3 words in order) would trigger the match despite the disconnect. Now we have balanced phrase matching with better term matching, so that the above examples no longer return the topic, but the following do: stress in the workplace, stress and workplace, stress from the workplace, and stress in a workplace
I’m especially excited about this fix, and I’ll be doing a lot of testing to see how well it works. Earlier this year I published a lengthy analysis of the effectiveness of Topic Explorer, and one of the biggest problems I encountered was the “Mad Lib” search query - where you can put any subject into a phrase and return the same Topic Explorer result (for instance, “women in the workplace” returned “stress in the workplace.”) Brent and the Summon team worked really hard on this fix, and it was an especially tricky problem to solve. I’m really excited that ProQuest took these issues seriously and worked so hard on a fix.
We will be replacing our current reporting system with Oracle Business Intelligence (OBI). In addition to the launch of the initial dashboard, which includes reporting equivalent to our current set of reports, this release lays the groundwork for expanded reporting of events throughout the Summon interface.
For us, we’re hoping this means more accurate Summon search statistics. Ive been maintaining our own software to get accurate and useful stats out of Summon, but I hope this new analytics tool will make my software unnecessary.
Relevance Improvements: Subject Term Boost
We are publishing a new relevance algorithm with Subject Terms having a higher influence in relevance. Previously subject terms were given similar treatment to other metadata found within the record; now subject terms are given a boost more similar to how Title or Author are treated.
The boost should only apply when query terms match specific subject terms and not when subject terms are found in other fields in the record. When searching on “computational linguistics,” the title “foundations of Computational linguistics” shouldn’t be affected, since “computational linguistics” is not a subject term for this item. However, the title “Readings in Machine Translations” should get a boost since computational linguistics is a subject term in that record. Also, the title “Japanese Computational Linguistics” should get a boost because it also has computational linguistics as a subject term.
Note: Because relevance ranking is a complex calculation of multiple factors, actual influence of subject terms will be determined by the search parameters at the time of execution. Also, you may not be able to recreate these exact examples in your instance of Summon due to differences in the search configuration at the time of searching, including differences in content rights, configuration, and facets applied, for example.
Beyond the fixes and improvements listed above, we have also made the following changes to improve performance and usability with Summon 2.0:
- Addressed issues that were causing the Summon search box generated by the Summon widget to disappear
- Addressed issues with text appearing out of margin in the new Summon results page preview pane
- Removed “Indian” as a language option in Advanced Search Language dropdown
- Corrected issues with the MLA Citation format