The amount of content in online products has grown rapidly, and it has become necessary to give users some guidance about how to find what they are looking for. Many of the standard search methods of past years no longer work well. This panel discussed some of the issues surrounding this problem.
Tom Beyer, Director of Platform Services, Safari, Inc., said that the goal is to move people around the platform and allow them to search as easily as possible, as well as allowing browsing, which is aimed at less experienced users. When they get to a page, they need to be oriented and shown what the content collection is about. To get them where they want to go, you can provide links from the page, have branding information about a journal or book, or (for a book) show them where the chapter is in the book. But can we do more? We could use semantic enrichment technology to related article links, but there is not a lot of information that you can give there because it may not be clear why things may be related. All this work requires lots of extra metadata which is expensive to produce.
Will Schweitzer, Director or Product Development at AAAS, said that solving the problem depends on how much effort a publisher wants to invest. Science goes back to the 1800s; there are a lot of connections, and an incredible amount of metadata is needed. Trying to create a seamless user experience is next to impossible. Are we delivering something of worth to the users by doing this? They spend about 3 minutes on a search, look at 2 or 3 pages, find the PDF, then go away. Social media users are on the site for an even shorter time: about 1.5 minutes. What do users expect? Improving the metadata may be more valuable in the long term. The search process is not on the publisher’s platform, and vendors don’t know enough about their users.
Michael Levine-Clark said that from a library perspective, librarians are not worried about spending time on a platform, only about getting users to the right content. Relationships will change with every project; it’s not as simple as a keyword search. They might be searching from different perspectives, and some topics will be tangential to their main search. Undergraduates search today, then tomorrow they search for something different. They may not care if they don’t find the related things ever again.
Will agreed with Michael and said that most publishers have applied taxonomies in ways invisible to the users. If the article a user wants is not in the top 3 results, they consider it a failed search.
Michael continued and said that we are falling short of what Google does. We have different content types in searches, and different needs. We have amazing primary source collections, but you would never want them to come up in the top 3 results because they are not what most people are looking for. How do you make them discoverable and make connections to other related material? We may be close to doing it with articles, but not with anything else. We are investing in discovery tools to keep our users in the confines of the library platform, but maybe we should be teaching them how to search Google more effectively to get to our content. Our collections also have other things such as books, DVDs, archives, etc. It is important for users to know that the library is an important resource and they should come to it. Branding should identify the material as coming from the library but that is not happening very well today. Typically the branding of a PDF is at the bottom of a page, so it is often not seen.
After these introductory presentations, the following conversation took place (questions asked in italics):
Is there a learning pathway that would be appropriate?
Will: Different parts of the content of Science are on 2 different platforms: news and research. The news platform uses mass media sources, graphics, and related content. They are getting a lot of traffic in the form of alerts from this. Scholarly publishers are looking to newspapers and what they are doing, such customizing the home page based on what they know about you and what you did in the past on their site. Not many scholarly publishers are thinking about their users that way.
Michael: Some of that is troubling to librarians; if you are tracking users that closely, you are starting to get into privacy areas.
Are users taking the time to personalize their preferences and help the publisher curate their content?
Tom: It is increasingly popular. We are helping people to have ways around the content and have the ability for people to save things and track them.
Michael: It is intriguing to think about what you could do: perhaps connect to the student information system and use their profile data. That is very scary!
Will: We are looking at social traffic and using altmetrics to learn how it has impacted our products.
Tom: There has been a resistance to initiatives that surface related content belonging to a different publisher. We can crosslink a series of related products, which has been difficult to convince publishers that it is good to do.
Michael: There is a romantic sense of the days of serendipitous discovery of browsing the shelves. It doesn’t work so well because one library may shelve a book differently from another. Not every book has a single subject, so the old environment of shelves and call numbers is limited. Today, it is fascinating to think about what you could offer and get to a virtual shelf and browse. That would be very valuable: from an unsophisticated search, find something that touches on what you are looking for. It would solve some of the tension as we have moved more print collections to offsite storage. There are many different connections that you could have among content.
Will: We have the technologies today with sophisticated A&I services. They are very complicated tools that are not user friendly, and they are expensive. How do we make it better and cheaper?
Tom: Can we get access to enough content such as all the resources in the library? If we can, we can provide some sort of a solution. Can you then say something about it; can you give directions to related content, etc.?
Michael: When you think about all the possible connections that somebody might want to make, the subject connections and authority ones are easy, but it is very difficult to figure out how you would do that in a meaningful way with methodologies and related topics.
Tom: It comes down to a set of links and the context you can give users.
Will: We have devalued the role of indexers. It will take all data sources to deliver a list of curated links.
Tom: Many of our products are essentially annotated bibliographies.
What is coming next?
Tom: I am skeptical about a technology-only solution to the problem. Humans still do a better job of curating content.
Michael: We are purchasing metadata, and its quality is going down. We need better quality and need to figure out how to have linkages.
Will: Pubishers must decide what their role is. It has never been about curating content; it has been on distributing it. We have not been good at finding people to help us along the way.