November 11. 2020, Wednesday. Just another day. At 9.30 on the dot, all 24 students are in class. Unshowered, sleepy little dots with initials. Torsos freshly filled with breakfast share the screen. I forget to turn off my mute and speak and gesticulate for almost a minute in front of my computer. There is no response, I feel alone in my bedroom, living room, and study. I switch on the microphone and sound, and am reintegrated into the group with a warm grin, repeating what I have just said. My fellow students understand me. How will the computer understand me? Students upload the hour-and-a-half recording and have it transcribed. How often does a sentence end with “so”, how often is “not” said, and what happens when I start the next line with the end of the line? They had texts written and – illustrated with screenshots from the session — compiled in a small book of poems.

The poems are based on the automatic transcription of the online meeting on November 11. 2020. 21:39:49 hours of video material were analyzed, resulting in 113 A4 pages of text or 113,546 words. In the case of the online meeting, the technical implementation took place via the YouTube service. Speech confusion, connection problems, and speaking too quietly led to errors and gaps in the transcription. The students left the discrepancies as they were. Paragraphs were generated by the program itself, according to the typical length of sentences or pauses in speech. The poems were mostly generated by filtering for specific terms. The images are based on screenshots from various videoconferences. The full transcript of the analysed videoconference is attached. The program does not take into account intervening noise, various transcription errors, and special incidents, which are noted here over the entire length.