FEAST is a regular forum to discuss formal approaches to sign language grammar (in particular in the generative tradition), experimental approaches to sign languages, and their interaction. The eighth meeting of the “Formal and Experimental Advances in Sign language Theory” (FEAST) colloquium had to take place in Hong-Kong in June of this year but due tothe increase of COVID-19 cases in Asia, by mid-February, it was decided to displace the conference in Paris instead. Mid-March, with the worldwide lock-down, the whole conference had to be re-imagined.

It was finally held online on June 23-25, 2020. The event was co-organized by the Université de Paris /CNRS (LLF and Labex-EFL), the Université Paris 8/CNRS (SFL), and the Ecole Normale Supérieure/CNRS (IJN) and ERC Orisem (PI- Philippe Schlenker)).

The conference had English and ASL/International Sign as official languages, all the budget was used to provide both offline translation and online interpretation. The conference included 10 talks and 12 posters with a recommended watching program starting on Tuesday, June 23^rd and ending on Thursday, June 25^th. The full program can be seen here.

TL:DR

Structure:

- Asynchronous talks and posters. Talks and lighting poster videos were stored on Youtube but embedded on the website on dedicated pages such that all the conference videos were visible from the official conference website.

Advantages: 1. asynchronous watching options made it easier for people to attend from many different time zones, 2. It guaranteed better video quality (no internet connection issues), 3. It allowed to translate videos ahead of time and to provide the transcript of the videos, 4. People enjoyed watching the talks following their own rhythm, to be able to pause them or increase the speed, 5. Youtube have no limitations regarding the number of simultaneous viewers.

- Live talk Q/A sessions. Talks Q/A were held on Zoom. The link to access the Q/A of the day was sent each morning to avoid Zoom bombing.

Advantages: 1. Having live sessions was livelier, 2. It gave the feeling of a ‘true’ conference since the moderator would announce the next talk/poster session or event to which attendants could participate.

Disadvantages: One has to verify that they have the correct tools needed to mail so many people, creating a mailing list would be a good option.

- Recommended watching schedule. The conference had an official program which indicated the talk to watch (with a clickable link to access the correct page) followed by the time of its Q/A live session, such that people willing to follow the program could watch the talk and then connect right away to the associated Q/A. Events in a day were comprised between 2 to 6pm (Paris time), to allow people from all around the world to assist to a maximum of live Q/A.

- Questions had to be asked using “raise hand” Zoom option. We decided to deactivate the chat feature, because (a) Zoom keeps track of raised hands, but chats appear immediately (b) it was not fully accessible to Deaf participants, since we wanted all discussion to be translated into sign language (c) there is no reason not to just raise your hand and ask your question in person. Even when chat is deactivated, though, the *host* can still send messages to everybody. We used this function frequently, to send reminders about deactivating camera and sound, and to send the link to the next event at the end of the Q/A session.

- Zoom session organization. For each talk, we had a moderator, who would start each session with technical announcements (how to raise your hand, etc.), and who would call on people in the order they raise their hands (which is indicated in the Zoom participants list). In addition to the moderator, we had a separate person who was the "host" of the Zoom meeting. This person would record the session, lower participants hands if they forgot to, mute and close people's videos if they forgot to, and had an action plan in the event of Zoombombing. The duties of both people are included in the check-list at the end of the document.

- Zoom session display. Since all Q/A sessions were interpreted, we wanted to make sure that the display was not too cluttered, so the moderator, at the beginning of each session, asked attendants to close their video and audio such that the only visible people were (a) the moderator, (b) the speaker(s) (max 2, even for larger co-authorships), (c) the interpreter, and (d) the person asking the question. In "gallery view," this provides a very comfortable viewing display.

- Replay. Zoom sessions were recorded and remained online (on the talk’s page) for a week. Talks and poster video remain online without a time limit, unless authors asked otherwise.

- Poster discussions. At the time of the official (i.e. according to the watching program) beginning of the poster session, people were invited to watch the montage video of the lighting talks (put online asynchronously) and download/see the posters (on their respective webpage). They had then 45 minutes to go to a dedicated Skype chatroom where the poster session occurred via written chats or recorded videos in sign language. This way, all the discussion was in principle available to all attendees.

Advantages: (a) when you went to a poster chat, you could see the questions that people had asked before you arrived (b) it can serve as a record for the authors of all the comments they received, (c) if you don't have time to see all posters, you can take a look at the discussion later (d) Some people in Asia also left questions on the chats in advance, before going to bed. You can see one example of a poster chat here: https://join.skype.com/amGwQbSB90tp

- Social events. We tried out http://gather.town/. It's not ideal -- hard to understand at first, the videos are small, and it doesn't work with all browsers. But it was pretty fun and silly if you can get it to work and take the time to explain it to other people. Alternatively, we shared a google spreadsheet where people could post links to social gatherings if they wanted to host their own Zoom session, etc. before or after the day's talks.

Challenges:

- Chinese firewalls. (which also cover Hong Kong currently) were an unexpected problem, blocking both Youtube and Google Sites. The relevant participants used a VPN to the extent possible.

- HTML emails. Fancy emails realized with mailchimp are apparently blocked right away (not even redirected in the spams) by some academic and professional inboxes. Hence, future editions should create a mailing list (to allow sending mails to many people at the same time) and send simpler emails.

- Navigating the program. Accessing all info could have been slightly easier. Example: the posters were a bit hard to navigate, since you needed to click on the poster’s page then click on the chat link. For a given poster session, it would be better to have all posters and chat links on a single page, for a given poster session.

- Zoom attendee restrictions. make sure that your Zoom license matches the maximum number of attendees that you expect.

- Social events would probably be more successful if they were happening in-between talks, so in the future, time should be allocated to that. The problem for conferences is that there are too many people for Zoom sessions to work. Gather.town offers one possible solution to that, though not yet perfect. Another would be to just have *several* Skype rooms (this is not possible with Zoom) opened at the same time, so people can switch between them to chat in smaller groups.

The long (detailed) journey

Challenges of the organization:

A conference organized in a three month-delay

FEAST 2020 call for papers was published by the University of Hong-Kong on January, 11^th, 2020. By the end of the month due to both the politic situation in Hong-Kong and the increase of COVID-19 cases in Asia, the decision was made to held the conference in Paris instead in a traditional format. After warning all the authors who had already submitted a paper and updated the conference information regarding location, an emergency organizing committee (Charlotte Hauser, Jeremy Kuhn, Caterina Donati, Carlo Cecchetto and Carlo Geraci) agreed upon a budget, booked the conference venue, and reserved the social dinner restaurant all in less than two weeks.

In March, country frontiers were starting to close and Italy began its lockdown, it hence became clear that having a physical conference would not be possible. We considered canceling the conference altogether but we had received 67 high-quality scientific contributions for the conference call-for-papers (the deadline for submission was March 4^th). Hence, we made the choice of having the conference fully online and proceeded with the reviewing process to select 10 talks, 5 alternate poster/talks (in case someone was not willing to present their results on an online conference) and 5 posters. With the ever-growing number of conferences canceled or re-scheduled for the Fall, we considered that it would be safer to hold the conference out of this busy period and before the official summer vacations (beginning of July), leaving us with a three month-delay for the whole organization. Presenters received their notification of acceptance on April, 24^th and had only one week to confirm their participation. Poster and talks’ videos were set to be sent within a month (May, 25^th), leaving two weeks delay (June 15^th) for interpreters to send back the translations’ video. Charlotte Hauser and Justine Mertz had then one week to edit together all the videos and put them online, on hidden dedicated pages.

Online

The conference format was of course the first challenge, how should we proceed to have a fully online conference? On what support should we host it? How could we conduct Q/A sessions? We benefited a lot from the experience of two previous online conferences: CUNY and GLOW. The organizing committee of both conferences had radically different strategies: CUNY was held live, following a pre-established program, only poster sessions were asynchronous, while GLOW was completely asynchronous and spread over a two weeks period.

CUNY organized all the synchronous talks and Q/A sessions on Zoom webinars, and hosted poster sessions on OSF. Through the use of Zoom webinars, questions could only be asked using the written chat and they were visible only by the host and panelist. Participants had no control over their sound and video, as such, session’s host had to select questions and unmute the correct person for each question selected. For posters, the PDF and video-presentation were put in an OSF folder and discussions could be held through the chat integrated on the website. (All the details of the CUNY organization can be found in this really helpful article: https://blogs.umass.edu/cuny2020/2020/04/12/virtual-cuny/ )

GLOW asked all presenters to put their slides, talks-videos and posters on their dedicated OSF folder, discussions also happened through the written chat.

From these two events we learned that: 1) having a fully asynchronous event is not very engaging, 2) we need to have one website with the whole conference searchable (instead of multiplying the platforms), 3) the questions/answer sessions had to be held live to allow both engagement and accessibility (for people who prefer sign language rather than English), 4) poster sessions discussions should also allow for bilingual discussions, not only written text.

We decided that all talks and posters should have their dedicated page on the conference website, and that all should be accessible from the conference’s program page. All these pages were prepared ahead of time (and remained hidden from navigation) with the name of the authors, the title of the talk/poster and the place to embed the associated video.

Talks and posters’ videos had to be asynchronous because we could not risk that connectivity issues alter the on-line interpretation and the watching experience. Once edited with their translation (see next section), all videos were put on the conference Youtube channel, unlisted, allowing to embed the videos on the conference website. Storing the videos on Youtube also made the integration of subtitles easier, it additionally allowed us to track the number of views of each video and ensured that there would not be any issue regarding the number of people watching videos at the same time.

All Q/A sessions were held live on Zoom and recorded. The Zoom links of each session (1 per day) were sent by emails (see example here: https://mailchi.mp/79c5c294b0e2/guidelines-for-talks-and-posters-12520252?e=324ced2d63 ) to avoid Zoom-bombing. For each Zoom session, there was a technician in charge of closing videos and sound of people not following the rules. The moderator had the role of reminding people of all the rules and giving of the floor to each person who had raised their hand by using the dedicated ZOOM device. S/he additionally had a visual reminder always next to her/his face to ensure that the sessions’ rules were known from everyone watching (see picture).

All asynchronous videos were available online on Monday, June 22^nd but participants had a recommended watching program, which integrated each talk just before it’s live Q/A session. At the end of each Q/A session, moderators also announced what was the next event in line in the program (next talks’ name or poster session or social event). The talks and posters’ videos remained online without a time limit (unless presenters were against it, in which case the video only stayed for a week), the Q/A sessions replays were only available for a week.

On the conference website, we provided a page dedicated to explaining the whole conference process, along with video tutorials on how to ask questions (https://sites.google.com/site/feastconference/feast-2020-online/how-to-attend?authuser=1 ).

Accessibility

Talks and posters videos

With a conference on Sign Languages (but this should be the case for any conference), we had to make sure that the whole conference was accessible to both hearing and Deaf attendants. This accessibility started with the conference website itself, on which pages were displayed both in English and in International Sign (IS), thanks to Mirko Santoro. Of course, we had planned to hire interpreters but the big question was whether the conference should be interpreted live vs. offline. The talks and posters interpretation was balanced in gender with two males and two females interpreters.

To facilitate their work and ours (for editing video), we requested that presenters record their presentation (30 minutes max. for a talk and 3 minutes max. for a poster), leaving ¼ of their slides free to insert the interpreters, and respecting a slow pace of speech (i.e. wait few seconds before switching to next slide, never speak as you display an example or show a video…), just like is usual in interpreted conferences. We put all video presentations on a shared dropbox with the interpreters, who returned back a video translation to be edited with the video presentation. When we edited interpreter’s videos, we made sure that the size of the interpreter was as big as possible, to ensure the best possible visual display between slide’s text and interpretation (see picture below). We also made sure that the interpreter’s eyegaze and pointing direction were coherent with their position on the slide (see again the same picture).

Whenever presenters did not follow the instructions, we had to diminish the size of the slides, to ensure that the interpreters had all the place needed. Similarly, when presenters did not respect the slow pace compatible with interpretation, we had to edit the videos to add stops, so that the slide would remain displayed as long as the interpreter had not finished interpreting it. Gladly, this only happened once. By the end of that process, all talks and poster’s videos where thus bilingual between English and IS.

Additionally, we used Otter.ai to create automated captions for each video. All the transcripts were sent to the presenters so that they could review them and send them back before the conference. We put the transcripts as subtitles on Youtube and also provided on, each relevant page, the link to download them for people wishing to read them. Slides were also downloadable from each talks/poster’s page.

Q/A sessions

From GLOW’s experience, we learned that having asynchronous Q/A sessions is not that attractive or engaging for participants, so we decided to have it live, with live interpretation. On Zoom, we found that the best way to guarantee an optimal visual display was to select the ‘gallery view’ and the option ‘hide participants without video’. With this display, only the presenter(s), the interpreter, the person asking a question and the moderator were visible on-screen.

To avoid reducing too much the window with the interpreters, at most two presenters for each talk could have their camera on at the same time. Questions could not be asked through the chat. This ensured the respect of the order in which people asked questions but also, following Dr. Santoro’s recommendations, to help Deaf people knowing where to look when a question is asked. People asking questions were asked to unmute their sound and video by themselves, introduce their sign name, and wait for the interpreters to be ready before asking a question. Asking questions was possible through the “raise hand” Zoom option, whose location and use was explained at the beginning of each session and through a visual reminder printed by all moderators (see picture in previous Section).

For poster Q/A sessions, we wanted to find a way to ensure: 1) that the conversations were easy to access from a technical point of view, 2) that the informal tone that is typical of in-person conferences was preserved, 3) that people reaching the discussion could be able to see previously asked questions, 4) that English was not the only option to interact, 5) that no-one would be left behind based on the language they understands. To achieve all these objectives, we chose to use Skype group chats. Skype group chats were created ahead of time, one per poster, they allowed to send text messages but also direct video messages. This last option was crucial in avoiding the hassle of recording a video, saving it somewhere on one’s computer and loading it then as an attachment to a webchat. Each poster’s group chat was accessible through a clickable link on the poster’s page, making it very easy to access. We asked people to not launch video live conferences, since we could not provide simultaneous interpretation of four posters, and since we did not want to risk to have people speaking to one another, hence excluding part of the attendants from the discussion. While we received remarks that the poster discussions should have been held also on Zoom, we believe that the option we found was the most accessible and equalitarian, in the absence of interpreters.

Lively conference around the world

One aspect that we needed to take into account when agreeing on the program was to ensure that everyone, around the world, would be able to participate to live Q/A sessions, or at least, that it would be as optimal as possible.

After comparing the main time zones (Japan, Hong Kong, West Coast USA, East Coast USA, Paris, Brazil…), we agreed that the best timespan would be between 2pm (early morning in the US, evening in Hong-Kong) to 6/7pm (Noon in the US, night in Hong-Kong) Paris time. We also made sure that talks from US presenters would be put later in the day, while Asian presenters would be put earlier in the day. Poster sessions were in the middle of the day so that any presenters could participate in the discussion.

The asynchronous talks, and the fact that we put the videos online one day before their recommended watch time, allowed participants to watch talks whenever they could, ahead of time, if the ‘official’ diffusion was too early or too late in their timezone.

One thing to note however is that people from mainland China or Hong Kong could not access the videos without a VPN because both the conference website and the youtube channel are related to Google, hence they are blocked by the Chinese firewall.

Social events

This is probably the point on which we have the most improvement margin. There were two official social events: one planned on Tuesday 23^rd, at 6pm (Paris time), for which we had a shared google sheet on which we invited people to propose their own meeting and provide the link to access it (following CUNY’s model), and another which offered to watch together the premiere of the SIGN-HUB documentary “We were there… we are here”.

The first event did not receive a lot of adhesion, the second, however, was a success. It even overloaded the maximum capacity of our Zoom session (100 people) that was held for the live introduction to the documentary. Gladly, the actual watching of the movie was on Youtube, hence presenting no maximum viewer limit.

For future events, we think that social events would have more chances to succeed if they’re planned in-between talks, so that people can discuss after a session, while they wait for the next session. For future events, we would like to use gather.town, as this platform is a funny way to reproduce conference informal conversations. It is a virtual room, in which you can move away or towards other people to start discussion, people too far from you cannot interact until they come closer.

Attendance figures:

713 people registered to the conference. We think that the vast majority did so to watch the SIGN-HUB documentary, nonetheless, since the documentary was hosted on the same platform as the talks, all the people registered have learned about the FEAST conference. Additionally, all registered participants received the daily emails which contained the whole program of the day.

180 people on average have watched talks’ videos, with a minimum of 111 and a maximum of 321 views depending on the talk.

88 people max. participated simultaneously to talks’ Q/A sessions on Zoom. On average, talks Q/A sessions had 70 people connected, the minimum has been reached on the last day with 57 people connected. Replays of the Q/A sessions were viewed 15 times on average. In previous FEAST editions, the number of people registered was between 50-60 on average.

263 people watched the live premiere of the SIGN-HUB documentary. To this date, the documentary has been watched 947 times.

Feedback from participants (37 answers):

How easy was it to find the details about the conference attendance, registration and so on?

You found the asynchronous presentations... (indicate how much you agree/disagree by sliding the button on the 6 points scale)

You found the talks Q/A .... (indicate how much you agree/disagree by sliding the button on the 6 points scale)

Out of 10, how many talks and talk's Q/A did you watch?

You found the posters session group chat... (indicate how much you agree/disagree by sliding the button on the scale)

Out of 12, how many posters and poster sessions did you consult?

Thank you for your answers, you are...