Authors : Clarisse Bardiot, Emmanuel Château-Dutier, Edward Gray, Alexis Grillon, Jaymes Kalala, Victoria Le Fourner, Mathilde Wybo
In November 2020, the DHnord conference took place online. We tried to make the most of the situation, and as several people asked us questions about the organization of the event, this post will relate how we did it, as accurately as possible, so that the experience can be used by other conference organisers.
Every year, the DHnord conference, dedicated to digital humanities, is organized at the Maison européenne des sciences de l’homme et de la société (MESHS) in Lille around October-November. Over the years, DHnord has become a key event in digital humanities, both nationally and regionally. Since 2019, it is directed by Clarisse Bardiot, scientific head of the digital humanities axis at the MESHS, in partnership with the Centre de recherche interuniversitaire sur les humanités numériques (CRIHN) in Montreal, one of whose members is the scientific co-director (in 2020 Emmanuel Château-Dutier). The administrative team of the MESHS is in charge of its organization. An added wrinkle: the permanent position of Digital Humanities Coordinator is currently vacant and Edward Gray was recruited only in July 2020, a few months before the date of the conference. Moreover, in 2018, it was decided to establish a biannual rhythm: in odd years, a conference in French on a broad topic; in even years, a conference in English on a specific topic. Finally, this colloquium has the particularity of operating on the basis of a call for papers and invitations from researchers, all validated by a scientific committee that is renewed each year.
The theme of the 2020 colloquium was communicated at the close of the 2019 edition (large corpora of images in art history) as well as the names of the scientific co-directors. At the end of the winter, the co-supervisors form the Scientific Committee, which met for the first time in February to define the call for papers and the title. The conference title was decided: « La mesure des images : approches computationnelles en histoire et théorie des arts » (The measurement of images: computational approaches in art history and theory). The call for papers was published within the usual « normal » deadlines with a response date of 05/06/2020.
In the meantime, the great confinement was declared… In spite of the uncertainties that weigh on the autumn, and a certain optimism about the possibility of conducting the conference in person, we decide as early as May to switch the colloquium online. The scenario is as follows: whatever happens, foreigners will probably not be able to travel; if there is a possibility of face-to-face meetings, this will be reserved in priority for the workshops. It was a wise decision: on the dates of the conference, France was once again confined …
It remained only to imagine the form of an online conference. Online seminars are flourishing everywhere, we are experimenting with online courses with students, sharing good and bad experiences. Based on these discussions, several guiding principles emerged:
- Take advantage of the situation to turn it to our advantage and invite the greatest international specialists (who would probably never have come to Lille). We were surprised by their enthusiasm and all responded favorably, far beyond what we could have imagined.
- To privilege live round-table discussions with exchanges between the speakers in order to preserve the lively if not convivial dimension of the colloquium. The round tables are recorded but agreement for rebroadcasting is requested a posteriori to allow the right to make mistakes and to privilege the live broadcast.
- Record the papers beforehand and broadcast them two weeks before the conference to encourage exchanges during the round tables, and allow the public to consult them according to their schedule while avoiding sitting behind the screen for several hours at a time. This also helps to reduce technical problems.
- Augmenting and editorializing the videos thanks to the MemoRekall software: the conference is very specialized and this makes it possible to explain the references mentioned by the speakers, to attach the bibliography, to refer to the code and the corpora if necessary, etc. This is particularly useful for students.
- Extend the duration of the colloquium to 4 days (and not two and a half as initially planned) with an appointment each day for a round table discussion from 1:30 am to 5:00 pm Paris time, which allows North Americans to participate.
- Choosing to move the roundtables to YouTube, Twitch and CanalU to avoid the audience from connecting to the zoom session and relaying the questions asked on a dedicated chat or on twitter.
- Propose a poster session in the form of 4 images that can then be broadcast on Twitter. Templates have been provided, to ensure graphic harmony and to avoid file problems.
- Define a graphic identity declined on all media and completely abandon paper supports
The formula was a real success: 200 attendees registered to the conference, with a large international representation; more than 100 people connected every day for the live round tables; more than 1,300 unique views on the videos at the end of the conference.
The main lesson is that the online conference shifts the workload: the upstream work is considerable and the technical and communication teams are very much in demand. On the contrary, since there are no more train ticket orders, mission sheets to fill out, coffee breaks to take, restaurants and hotels to book, the administrative and financial teams are much less involved. It does not seem to us that an online conference will replace in person meetings as the the formula for academic conferences either: if all the people invited have responded favourably, it is because we have met them « in real life » in face-to-face conferences or during other scientific exchanges; finally, informal exchanges, where future collaborations are often decided, are terribly lacking.
We detail below the main points with our working documents, which you can freely reuse and adapt. Regular meetings with the whole team (monthly then bi-monthly), a great flexibility of adaptation, listening and a lot of good mood are the main ingredients. We weren’t able to escape last minute bugs, the crash of a video platform (but we had two others working) and two or three other surprises but this (almost) went unnoticed!
Undertaking an international conference entirely online requires, above all else, planning and efficient organisation. Retroplanning was key, as was making room in this schedule for unplanned delays- because uncertainty is the only thing certain in 2020. We also needed to keep the team on the same page, and to keep our communication with our participants succinct, clear, and efficient.
After identifying our participants and ensuring that they were aware of our general plans, we then set out to ensure that their job – that is, creating their presentation and sending it to us – was as simple and clear as possible. Thus, we chose to ask the participants to send us only the video files, deciding to handle the editing on our end- ensuring a degree of harmony and standardization in the videos. We elaborated a guidelines document that not only explained in detail how to use OpenCast Studio, but also gave tips for how to effectively film oneself. While these may seem like simple recommendations (avoid having windows behind you, wear simple colors) experience over the last 10 months has shown that these are important considerations. We were also flexible when other participants asked to use other platforms. What was important for us was to take the guesswork out of filming, so that the speakers could focus on making a great presentation.
It was also vital to keep the entire team on the same page. To do this, we took advantage of a private chat channel in our mattermost chat, pinning messages as necessary. We also used a shared file system- those within the ULille system used Nextdoor, however due to the difficulty of integrating those without ULille accounts, we also used Google Drive and Docs. In these docs, we kept a journal- a running list of recaps of our previous talks (we took notes in a collaborative manner as meetings progressed, ensuring that we understood)- as well as other documents, such as the aforementioned guidelines, and several spreadsheets dedicated to following the progress of the various aspects of the conference. The ability to share these spreadsheets with other, non-permanent team members allowed us to rapidly and easily transmit information and work at our own pace. This allowed team members to instantly check what was done, what was still being worked on, and what needed to be done. With team members working from home, and in the case of Emmanuel, a continent away, these elements were invaluable.
In preparing for the day itself, we sent an email to the participants in the round tables the week previous, with the complete zoom connection information. When creating the zoom, we also made sure to enable data-centers in regions where we would have connections from participants, such as Canada to cover North America. While the team connected and got into place 30 minutes before start, to ensure that everything was running well and to discuss any last minute matters (or simply to share a coffee and a treat), we had participants connect 15 min early to ensure that their cameras and microphones functioned well. We then did a last minute reminder of how the session would proceed before going live. By repeating this information and making sure our participants were on the same page, we made ourselves prepared for any surprises.
In addition to the moderators, there were several team members connected. The Regie handled streaming, and another, separate team member was on standby for any technical issues/to allow any participants to enter. We fortunately did not have to make use of this, but it is best to be prepared. Other team members worked on the communication aspects, livetweeting the conference or composing drawings that represented the discussions. A final team member watched for incoming questions from twitter (often with the help of other colleagues who were dedicated to surveilling these platforms and relaying them to this person), the dhchat, and the questionnaire form, compiled them, and relayed them to the moderators. We used a combination of zoom chat to relay to the moderators – so they could focus on one program – and in the case where the moderators used an online framapada for organizing their own questions, audience questions were added to the pad. One common experience was that audience engagement was somewhat more difficult to achieve, so it is necessary to prepare some questions ahead of time.
A final note about organizing the ateliers- we originally planned for 20 people in the audience to allow discussion during the workshop, but ultimately , and asked those that signed up to confirm their presence the week before the conference. The response rate was really rather low, and we soon learned that we were best off inviting far more people than planned to the sessions, since online events are much easier to decide to skip out on at the last minute.
With an entirely online event and the broadcasting of pre-recorded conferences, it seemed particularly important to us to be able to find solutions to guarantee, in spite of everything, high-quality scientific exchanges. Indeed, we were acutely aware of what we were losing with a remote event in terms of informal and spontaneous meetings and discussions that often make up the salt of scientific meetings. It therefore seemed to us relatively essential to give a large place in the course of the event to discussions through the organization of several round-table discussions which would be broadcast live on a rather limited time slot but throughout the week.
In order to ensure that the discussions would be properly conducted, it was essential that a leader be designated for each round table to prepare the discussion by carefully discussing his or her intervention with each speaker. Thanks are due to Nuria Rodriguez Ortega, Taylor Arnold, and Everardo Reyes for agreeing to play this role and for their investment in the preparation of the roundtables. Based on these preliminary discussions and having reviewed the recorded communications of the various participants, a sequence of each round table could be established, thinking about the dialogues that could be established between the various stakeholders. These sequences, organized by major thematic issues with sub-issues, were intended to organize the discussion during the roundtables. They were therefore communicated to the participants before the event so that they had time to think about them before the event. With the recorded lectures online, all speakers had also had time to think about their colleagues’ presentations.
Speakers were welcomed into the roundtables half an hour before the event was launched to ensure that there were no technical problems and so that the speakers could begin to get to know each other in informal exchanges. Before going on line, they were reminded of how the event would be run, stressing the importance of having as fluid and informed a discussion as possible, so that they could of course question or bounce back on each other’s comments. Several channels were set up to collect questions from the public on Mattermost or through a hashtag (#dhnord2020) on Twitter. The roundtable facilitators were relieved of the task of collecting questions by designated individuals so that they could concentrate on the discussion. These questions were communicated to them in the videoconference instant messaging system at the appropriate time.
The discussions during these roundtables were particularly rich and of high quality. In addition to the choice of participants, we believe that much of this was due to their preparation. On the other hand, there were not many questions. A good part of the registered participants registered on Mattermost but the indirect contact with the speakers may not have facilitated the discussion. Facilitating the discussion on the Mattemost channel throughout the event may have facilitated the gathering of questions and reactions from the audience. For live events, it may be necessary to allow the audience to connect directly to the videoconference, however, we had favoured securing the bandwidth of the videoconference and multi-channel broadcasting.
One of the more unique elements of our conference were the pre-recorded videos that were then enhanced with MemoRekall, an annotation software developed by Clarisse Bardiot. While relatively simple to learn and implement, MemoRekall allowed us to showcase a new and innovative approach to conference papers. With this software, we were able to mark out the principal sections- or “chapters”- of a presentation, as well as provide “footnotes” that contained links to papers, websites, and datasets. Thus, our audience has easy access to further explore, something that is quite limited in a traditional conference format! Even more so because these MemoRekall capsules are hosted on the website of the MESHS, and will remain accessible for the longterm. Thus, #dhnord2020 will continue to be a resource far beyond November 2020.
In creating these capsules, we asked our participants to send along their presentation slides as well as any references/breakdowns they wished to have added to the capsule. They often submitted their slides which had elements of “chapters,” and more rarely, time stamps. Each video started with placing a note, that lasted for the entire video, about the conference, as well as the author of the intervention and a link to submit a question to be asked during the round tables, as well as an upload of the slides if these were given. From there, we watched the videos and added references as they came up – giving 30 seconds minimum for links, and making “chapters” as best we could. A second detailed watch was necessary for quality control and assuring that the “chapters” were relevant, especially when these weren’t already clearly defined by the author. Depending on how many references there were to be added, as well as what was provided by the participant, each video took between 2.5 and 3 hours to complete on average. Once completed, they were then sent to the authors for approval before being placed on the conference website.
Example of a capsule edited with MemoRekall. view fullscreen
The communication of the #dhnord colloquium was implemented by the MESHS communication department, supported by the Digital Tools and Digital Humanities departments and by the members of the organizing committee. These tools were deployed in the context of the health crisis linked to the Covid-19 pandemic during which two lockdown periods were imposed (spring and autumn 2020), involving massive recourse to telework.
The graphic identity of the conference and its variations
The first of the communication actions was the definition, then the declension, of a graphic identity concerned with embodying the theme of the 2021 edition of #dhnord devoted to the study of images in human and social sciences through digital tools, taking care of the aesthetic, simple and warm character of the visual. After much debate on a series of diverse and varied first drafts, we decided on the concept of an ASCII image. After deciding to use the ASCII format to make the conference image, we chose Ada Lovelace, a pioneer in computer science, as the primary image for the conference. Her portrait was transformed into an ascii image using free software. The poster was then enriched by some computer circuits drawn by hand by Adèle Huguet, mediator and scientific illustrator, with whom we had the pleasure of working during the 2019 edition of #dhnord. The landscape format was chosen because it is better adapted to digital communication.
Poster of the conference #dhnord2020, MESHS, 2020 (English version)
Poster of the conference #dhnord2020, MESHS, 2020 (English version)
The visual was then declined on different supports (website design, digital signature, entry and exit cards and video capsules design, social networks design, posters). A digital program was developed and illustrated by the portraits of the conference speakers also « transformed » into ascii images, after having made the requests for authorization to transfer image rights. For all media, the 100% digital format led us to choose and adapt formats according to different needs (screen size 1920×1080, landscape format, long version of 82 pages for the program). The themes of the conference were also illustrated by Adèle Huguet, in the continuity of the graphic charter.
Plate of illustrations made by Adèle Huguet for #dhnord2020, MESHS, November 2020.
Plate of illustrations made by Adèle Huguet for #dhnord2020, MESHS, November 2020.
Information channels (internal and external)
The call for papers of the colloquium was first distributed in spring 2020 via the SciencesConf.org website and the MESHS website. Next, the scientific content (program, original presentations – example, presentation of the speakers not having made an original presentation) and practical information (« How to access online to the #dhnord2020 conference?« ) concerning the conference were gathered in a single point (MESHS website). The pages of the site, as well as the digital program, have been written in English and French. Shortcut urls and anchors in the pages were created to facilitate the communication of information on specific contents. Hyperlinks were used extensively throughout and required extensive checking. The information concerning the conference was also communicated on several portals (Calenda, Culturables, OCIM, agendas of the partners and universities supervising the MESHS). It should be added that the international dimension of the 2020 edition of #dhnord aimed to mobilize a larger scientific audience than in previous editions. This required systematic translations of all media. In addition, a new feature was created on the website to allow a selection of the time zone.
The Twitter feed #dhnord2020 was used in advance of the colloquium to disseminate information about the event itself, the posting of the pre-program and the full program, the opening of registrations and the posting of annotated videos, ten days before the start of the colloquium. Illustrated threads were written by theme, then by round tables. The accounts of the speakers and their institutions were identified upstream in order to easily mention them in the messages relating to their interventions. During the event, the streaming of original interventions, workshops and round tables were announced via this thread. Reminders were scheduled in parallel on the Mattermost chat. The « Posters » session was also organized on Twitter on the 3rd day of the conference. The Twitter feed was also used to relay certain questions.
As a reminder, participants had three ways to interact with the speakers and moderators of the round tables: the #dhnord2020 Twitter feed, the Mattermost chat and a LimeSurvey questionnaire exclusively dedicated to questions (« Ask your questions/Ask your questions »). The feed was fed and enriched by the « live-tweeting » of the members of the organizing committee and conference participants. In total, between November 15 and 20, 27 messages were posted from the MESHS institutional account, most of them in English. As a result of this activity, the MESHS account welcomed 55 new subscribers during the month of November, which is 3 to 4 times more than a usual month. All indicators show an intensified activity for the month of November 2020.
Screenshot of Twitter analytics as of 30/11/2020 for the @MESHS_Lille account
Screenshot of Twitter analytics as of 30/11/2020 for the @MESHS_Lille account
The Mattermost chat was useful both externally and internally. Since participants were unable to connect to the roundtable zoom session (reserved for speakers and moderators), the objective was to provide a space in which participants could express themselves, ask questions, and easily exchange information. Several discussion channels were created, by thematic sessions and round tables. A digital humanities monitoring tool was also integrated. This tool made it possible to disseminate practical information, but the Mattermost chat was rarely used as a discussion space by the participants: this is a potential area for improvement for 2021.
Internally, the flow of shared information was quite intense before and during the conference. A private « organization » channel, associated with the team’s regular zoom meetings, was created and proved to be very useful, allowing us to interact easily and pinpoint important messages and come back to them as needed. Several shared documents were also created. Although team members were able to meet on site during the four days of the conference, facilitating exchanges and information sharing, the Mattermost chat and other collaborative tools were invaluable. Their use should be preserved for future events involving several internal services and external partners, as was the case this year with the CRIHN.
Forms and mailing lists
A LimeSurvey registration form was used to gather conference registrations.. This tool is important to know the future audience of the conference as well as the main demographic information of the audience. On the basis of this form, two mailing lists « participants » and « speakers » were created with the help of the MESHS IT department. During the fifteen days preceding the colloquium, the « participants » list was updated daily. It was very useful to communicate practical information about the online conference (how to connect to offline and live content). This direct link with the participants is indispensable and it would be interesting to think about the future to better schedule the sending of indispensable information (confirmation of registration, putting new content online, starting live sessions, etc.). These mailings must be regular, without being invasive. In the case of a technical system in which several connection points are proposed (« virtual » spaces where the conference takes place), it is necessary to support the participants as well as possible in their participation in the event.
Two other forms have also been created and proposed to participants and speakers: a dedicated form to collect questions and a satisfaction form to collect opinions and areas for improvement. This questionnaire was developed at the end of the conference, at a time when the team had a better view of the final result.
A preliminary study of the means that would allow the organization of an online conference was carried out beforehand. It emerged that while « all integrated » solutions are attractive, the quality of the service is not guaranteed, and makes the institution totally dependent on the existence of the service for its activities. We therefore chose to build the colloquium with our own means. It should be noted that a certain number of solutions, complementary to what will be described later, have been identified, without all of them having been put in place (due to lack of time or means), such as, in particular, Mumble and the mailing lists.
Given this observation, the choices for broadcasting the conference were therefore made by relying on our solutions, and by choosing two ways of broadcasting videos :
- pre-recorded programs that allow for reliable pre-broadcasting. They constitute the backbone of the conference.
- streaming programming, partially live, which presents more risks of breakdowns and hazards.
Videos on demand
Concerning the pre-recorded videos, in this period of alternating confinement and forbidden travel, we have chosen to make the speakers autonomous via an Opencast type service, by providing them with a practical guide to improve, as much as possible, the quality of the intervention.
These videos were then incorporated into the graphic ecosystem designed by the communication unit, integrating both the presentation, and the capture of the speaker during his presentation on the background provided by the graphic charter of the conference.
They were then integrated into MemoRekall to link to it an augmented content with which the spectator could interact.
Streaming: Preamble
Concerning streaming, we were inspired by the architecture imagined by Pascal Dayre (IRIT Toulouse) in the framework of JDEV2020 (You can find some thoughts on this event here) and built around the federative project « Constellation » of which MESHS is a part. By chance, the MESHS took the responsibility of the diffusion of the communications during the event.
Global video diffusion schema
Global video diffusion schema
At the heart of the broadcast lies an OBS studio (with a second machine ready to take over if necessary), which transmits the stream. It allows the integration of video, sound, music and the capture of external sources (such as videoconferences).
The stream is then pushed to an NGINX server equipped with an RTMP module, which in turn pushes this data to different output platforms or other relay servers.
So far, this part is identical to the device deployed for the JDEV2020.
NGINX RTMP Relay ServerNGINX RTMP Relay Server
On the other hand, unlike the JDEV2020, the #dhnord2020 conference, although aimed at a technophile audience, nevertheless reaches spectators with more specific, less generalized computer knowledge. Therefore, we chose to use mainly consumer video distribution platforms (Canal U, YouTube and Twitch), and not a constellation of RTMP streams whose use requires the installation of other software such as VLC.
This compromise makes it possible to offer both non-GAFA institutional solutions and services whose reliability has been tested to ensure that they are operational. We are convinced that offering the user choice, putting open solutions and « comfort » platforms in competition with each other is an essential step in reaching all audiences while guaranteeing a form of digital sovereignty.
Streaming: Prepare for the worst, hope for the best
For each of these steps, a fallback solution was available. Thus the OBS studio was not only duplicated on two machines, two networks, two electrical sources, but it could be bypassed by pushing the stream directly to the relay server from the videoconference (Zoom in this case).
The multicast server could be bypassed by sending the stream to only one of the platforms.
The relay platforms themselves were multiple.
However, it should be pointed out that the various standards used by the platforms, even though the original stream was a single one, could lead to technical complications (which is what caused us to be unavailable on one of these sites for part of the event).
Finally, the visual media dedicated to the broadcast included messages, which could be manually activated, to signal potential problems or delays. However, they could only be broadcast via the control room or by sharing a zoom screen in case of emergency.
Streaming: round tables
In order to guarantee a maximal quality during the sessions, and given the international nature of the speakers, we had to rely on a proprietary solution (Zoom). However, we wished to isolate it, for security and portability reasons, from the spectators. Also, these sessions were not accessible to the public, and one person from the team was in charge of accepting entries into the session. They were retransmitted using OBS window capture and integrated into the streaming described above.
Streaming: what content?
We thought it was essential to integrate pre-recorded videos into the streaming.
Indeed, we know that the perception of the spectators towards an online conference differs greatly from the same face-to-face event. Offering the opportunity to view the videos (not augmented) just before the round table dealing with the same theme helps to facilitate accessibility to the conference’s contents. However, this creates an additional complexity in the communication of schedules.
We also wanted to accompany this broadcast with a prior streaming period in a « waiting test pattern » (one hour before), in order to remind the audience of the broadcast times, and then a short « introduction test pattern » (one to two minutes) in order to attract the viewer’s attention just before the broadcast, thanks to a different sound design, which is not without reference to the « three shots » signaling the beginning of a play. This practice, which does not correspond to the modern uses of « live streams » or « youtubeurs » that begin their broadcast without any precondition, is however relevant in the case of accounts that have neither subscribers nor a loyal audience (which was our case). These test patterns were based on on-demand video cards, and dressed with sound effects and music available in Creative Commons 0.
On most of the broadcast « scenes », different graphic elements that reflected the graphic identity of the conference were manually activated to adapt to the circumstances (reference to question forms, hashtag, attention message, message of resumption of broadcast, delay or technical problem).
Graph detailing the number of spectators during one of the streams
However, it should be pointed out that excessively long streams, although allowed by most platforms, are not without potentially causing problems of post-processing.
Streaming: how to accompany the viewer?
The main obstacle remains the interaction between the spectator and the moderators / round table and to be able to offer clear and explicit information to the spectators on the hazards of streaming (delays, breakdowns, etc.).
During this conference, we proposed different solutions:
- A (unique) form to ask questions : although relatively simple to use, it was surprisingly little used (too much impression of complexity?, subjects not very conducive to questions).
- A Twitter hashtag to ask questions and interact on the sessions
- An instant messaging service (such as Mattermost) equipped with centralized authentication support: There was very little direct interaction on this channel (which hosted about 1/3 of the registered conference attendees). However, it seems that the tool was appreciated for the reminder to start broadcasting sessions. However, despite our efforts, this tool, which was supposed to provide a relational aspect closer to the traditional colloquium, remains a failure.
- A system for converting the different time indications in different time zones on the MESHS site.
The formula was a real success: 200 attendees registered to the conference, with a large international representation; more than 100 people connected every day for the live round tables; more than 1,300 unique views on the videos at the end of the conference.
The main lesson is that the online conference shifts the workload: the upstream work is considerable and the technical and communication teams are very much in demand. On the contrary, since there are no more train ticket orders, mission sheets to fill out, coffee breaks to take, restaurants and hotels to book, the administrative and financial teams are much less involved. It does not seem to us that an online conference will replace in person meetings as the the formula for academic conferences either: if all the people invited have responded favourably, it is because we have met them « in real life » in face-to-face conferences or during other scientific exchanges; finally, informal exchanges, where future collaborations are often decided, are terribly lacking.
URI/Permalink: