A new feature to fuel discussions about humanities and social science books
OpenEdition Books has an exciting new feature. Now, with “Review of Books”, publications can be linked to book reviews on OpenEdition Journals, helping to fuel discussions about humanities and social science books. We interviewed Elodie Faath, the OpenEdition Lab department manager and R&D project manager, Mathieu Orban de Xivry, the OpenEdition Lab R&D developer, and Alice Pierre, the Books department editorial officer to find out how Review of Books was developed and what the future might hold for their research collaborations.
- It all began with the AgoraWeb research project. What was the purpose of the project?
Elodie F.: AgoraWeb was an R&D project funded by the Provence-Alpes Côte d’Azur Regional Council between 2012 and 2015. It was the second collaboration between OpenEdition and Laboratoire d’Informatique et Systèmes – UMR 7020 (LIS, formerly LSIS). The team at AgoraWeb came up with the idea of a tool that would recommend books automatically based on dynamic analyses of the web, its content, and their implicit and explicit interconnections.
AgoraWeb was originally used to create a website intended to spark discussions about humanities and social science books. It worked by finding book reviews and comments about scholarly publications on social media and by providing a dedicated space for conversations and debates about the books.
To succeed, the team needed to develop robust, large-scale digital methods to:
- Effectively crawl the web for relevant information and comments on books
- Automatically detect and analyze text that referred to the books
- Design a system that offered recommended reading based on certain data
The research team consisted of two researchers, Patrice Bellot and Frédéric Béchet, and a doctoral student, Hussam Hamdam, who defended his dissertation in 2015. The development team was composed of an information specialist and a developer in charge of creating a prototype of OpenEdition Review of Books.
- Why did you decide to turn it into a feature of the OpenEdition platform instead of building a separate website?
Elodie F.: The goal of AgoraWeb was to create a meta-review that would find and analyze all book reviews (identified by the document’s metadata or detected automatically), collect and sort bibliographic information about the books, and enhance the website with a recommendations tool.
The prototype for OpenEdition Review of Books was an experimental space that solved several of the project’s research problems at once, including how to process corpora and how to evaluate and analyze results. The prototype also had to imagine possible navigations based on data description models, as well as improve the environment used to distribute and promote reviews on the various OpenEdition platforms and optimize user interfaces.
Although the results were promising, a reference website would have required more content creation and editing. Instead, we decided to turn it into an add-on to promote the books and reviews available on OpenEdition Books and OpenEdition Journals.
- The project was a collaboration between OpenEdition Lab and OpenEdition Books. What was each team responsible for?
Mathieu O.: One of the first things we talked about was how to describe and detect the most relevant information so that we could identify a book from its review. The problem is that sometimes metadata was missing and sometimes it wasn’t, while other data was scattered in different places. For example, sometimes we would have to extract the name of an author (using Bilbo automatic citation software) from a complete bibliography, even though it usually appears in the metadata, in a specific field. By collaborating with the Books team, as well as the Journal and Data teams, we were able to refine our understanding of metadata descriptions. OpenEdition Lab then selected the most useful metadata. The Lab and Books teams also helped determine how to position and display information on the platform pages and avoid potential detection errors.
Alice P.: The Books team worked with us throughout the project to ensure that we were giving Books platform users what they needed and provided the best user experience possible. We worked especially hard on the front office template displays and pop-ups, the back office plugin functions and displays for Lodel users, and on the informational pages that explain how to use the new feature.
- How can users find a review from a book’s page?
Alice P.: When one or more reviews are associated with a book, a “mentioned by” button appears on the book’s home page on the left beneath the photo of the cover. If you click the button, a pop-up window opens, listing reviews of the book.
From there, you can see the review’s author, title, URL, and DOI. To read the review, simply click the URL and it will take you to the appropriate OpenEdition Journals page.
- How many books are linked to Review of Books?
Mathieu O.: Currently, 13% percent of our books are linked to one or more reviews, totaling about 1,400 individual books. However, some books have multiple reviews.
We really aren’t sure how many books have reviews. It very well may be twice that number. The problem is that the more we expand the algorithm to find more books, the less accurate the algorithm becomes. It starts returning false positives. If we tighten up the search parameters, we get extremely accurate results but far fewer books. We decided that the accuracy of the links between the books and reviews should be greater than 90%, and we are currently at 91%. We’ve included a manual way to report any false positives that slip through.
- Are there any plans to link books to content on other OpenEdition platforms?
Elodie F.: During the Inter-Textes project, we discussed the possibility of linking books to reviews on Hypotheses. Chahinez Benkoussas, a doctoral student who studied the project, began by looking at book review detection markers on OpenEdition Journals and Hypotheses. She described her initial results in a blog post on OpenEdition Lab. We took her results and built on them using work by another doctoral student, Hussam Hamdam, on sentiment analysis, which led us to create a piece of software called echo. The idea was to complement the book review markers that Chahinez identified using sentiment analysis markers on Hypotheses. Mathieu has uploaded the code for ROBOH (Review of Books on Hypotheses) to GitHub.
While the software categorizes Hypotheses book reviews with 80% accuracy, the most challenging part is correctly identifying the books being reviewed. Searching implicit citations in a text will likely be useful. We hope that the recent development of search software called Bilbo2 will also help.
- You said that the goal was to fuel discussions about humanities and social science books. What else can Review of Books do? Do you have ideas for future research projects?
Elodie F.: The primary goal of the project was to promote book reviews on OpenEdition Journals and to link them to books not only on OpenEdition Books but also on other publication websites. It also does the opposite: it connects our books to mentions outside of our journals. At the time, the project was a complex and ambitious one, but now we want to reach out to others and introduce them to the benefits of our tool.
There are three main ways we might open up and expand the reach of Review of Books in the future:
- By sharing our information in a variety of standard digital formats to make it accessible to as many people as possible. For example, we could add it as metadata in descriptions of books or reviews (and especially in DOIs). We could also develop a function that would allow partner libraries and institutions that support open-access development by paying for Freemium content to easily access the metadata and offer it to their users.
- By promoting our reviews by searching for reviewed books mentioned on publication websites other than OpenEdition Books.
By going one step further and opening it up to the public. We would work with specialists in digital publishing, such as La Vie des idées and Nonfiction, and with digital book sellers to construct a shared database and API (application programming interface) that would let anybody add links to potential reviews and then associate the reviews with their books, regardless of what platform they use.