The air is starting to cool in the Puget Sound and the leaves have turned. This of course means that students and faculty have migrated back to their respective institutions and are embarking on another year of learning, research and breakthroughs. It always brings me great joy to think about the potential impact the collective research community can bring to the world each year. To that, a tip of the cap and a hearty “Welcome Back” from the Microsoft Academic Team.
Of course, the Microsoft Academic Team does not follow the same schedule as academics and like many of you, we have had a busy summer. Together we have added a few features to the site and addressed some issues that we have found through user feedback and testing. I’d just like to take a moment to recognize those of you that have submitted feedback, thank you! We review all the feedback and appreciate your perspective and suggestions.
To name a couple user-initiated features:
- Ability to export citation lists to RIS format (import to EndNote / Zotero)
- Export the list of papers associated to your profile to various formats
- Conference calendar improvements on the conference analytics page (opens in new tab)
- Period characters (.) in usernames
Reading Lists
In addition to user-initiated features and fixes, we have been working on a couple major features for the site. Our users really like the “Cite” feature on our site. It allows you to select a list of papers throughout the site to build a bibliography and export that list to various formats. However, as we looked at that feature, we wanted to create something that would allow users to carry that list across user sessions and add the ability for users to also place those papers into custom lists. In this way, the Reading List feature can be a collection of bookmarks that allow users to come back to papers to consume later. And, like the cite feature, we allow users to export those lists in various formats.
Organic growth of this feature has been great to see since we released it a couple weeks ago. This seems to be a great time to announce this feature as well with the new academic year upon us. You can use this feature to collect papers that you would like to read that might involve research you are starting in on; or if you are a faculty member, generate and export a reading list for your students.
Since the launch of the feature, we have added some additional tabs to the reading list interface that show related papers to those in your list, top references from papers in your list and top citations for papers in your list. We believe this will help you find papers that you may not have known about (related papers) as well as understand both the influences (top references) and the impact (top citations) papers in your list have had. Sign into your Microsoft Academic account, click on the icon anywhere you see a paper on the site and give it a go.
Experimentation
To integrate users into our feature pipeline we will run experiments on the website from time to time. We have been experimenting recently with how we understand search queries made on the site and how users interact with the results given. We have looked at the usage of the search feature and have found that most of our users are conditioned to more traditional search engines that run lexical or keyword search to return relevant results to the user.
When we re-launched Microsoft Academic ~4 years ago, we decided to employ pure semantic search in our search box to deliver highly accurate results based on entities we interpret from your query. If you know what you are looking for, typing a title or author into the search box would give you a single result in the best case. This approach makes some assumptions about our users. One assumption is that users are interested in a specific entity or set of entities and would like to see results that reflect an association with those entities (Example query: papers by Eric Lander at Broad Institute). It also assumes users understand this semantic search behavior and can craft a query to get the desired results. In some cases, the combination of these assumptions can result in user de-sat for those conditioned to lexical or keyword search.
A few months ago our engineers embarked on an investigation to see if we could use our existing indexing product (KES (opens in new tab)) to return a mix of semantic and keyword search results. This work has been really exciting to watch and at the end of the summer we felt that the technology was mature enough to deploy to our production servers as an experiment powering the search feature on our site. What this means for our users is that we will try to understand your query better and present semantic results when we can, but also mix in keyword results where appropriate. This blended result seems to be working as we have seen a favorable uptick in search result interaction since the launch of this experiment.
Over the next few months we will be refining this new way to interpret user queries and surface highly relevant entities from our graph. There is a lot to talk about regarding this new technology and we will be explaining our approach in an upcoming paper and blog post; so, stay tuned for that. Until then, head over to Microsoft Academic (opens in new tab) and give our new search a try and if you have feedback one way or another, let us know via the feedback tool at the lower right of the website.