During the spring and this summer, I worked on and off, in my free time, on a side project about French researchers and their publications. This is an ongoing work, it will probably move slowly and last for some time, but I think it’s interesting to start documenting the first results. Anyone with a passing knowledge of the French Higher Ed landscape will know this is a daunting task: polite decision makers talk of a fragmented system; less polite people confess it’s a terrible mess, a jumble of overlapping structures and competing loyalties, with Universities, University groups, Ecoles, the CNRS, labs big and small with multiple parent organisations. And changing all the time: they split and regroup like crazy cells in a petri dish. ...

Depuis janvier je gère la publication d’une lettre d’information professionnelle, hebdomadaire, “Doc. - Voir aussi”, sur des sujets de documentation. On a envoyé 22 numéros de la lettre et le logiciel utilisé pour gérer la lettre donnant quelques éléments statistiques, je me suis dit qu’il serait intéressant de faire un billet à ce sujet. Doc. Voir aussi - quelques statistiques 510 personnes se sont abonnées avant que la 1e lettre ne soit envoyée, suite à l’annonce de sa création sur les réseau sociaux et sur les listes professionnelles. L’envoi de la 1e lettre en janvier a engendré une nouvelle vague d’inscriptions, un peu moins de 100 personnes supplémentaires. Depuis c’est stable, une ou deux personnes supplémentaires par semaine tout au plus. On peut en déduire que le lectorat est venu des canaux de communication professionnelle, mais pas du bouche à oreille. La liste a 650 abonnés aujourd’hui. ...

TL;DR un pas dans la bonne direction attention à la définition des attributs minimums nécessaires particulièrement adapté à la situation en France on pourrait déjà plus utiliser Shibboleth Nathalie Clot attire mon attention sur un débat en cours autour d’un effort pour créer une norme NISO pour la gestion de l’accès aux ressources électroniques. La recommendation est RA21: Resource Access for the 21st Century, et le billet sur lequel pointe Nathalie est What Will You Do When They Come for Your Proxy Server?. De quoi s’agit-il? ...

With the caveat that my last.fm account was disconnected in April and I didn’t realize it before the beginning of May, I have a few stats about the music I listened to in 2018: 3,645 tracks, some multiple times: 6,245 listens. that amounted to 15 days and 4 hours of music 690 different artists, of which 66% were artists I hadn’t listened to before 971 different albums (84% new) My top albums were: ...

Two years ago I started learning the Go Programming language. I’m not a professional dev and so couldn’t devote as much time to it as I would have liked, but I did continue to learn and got better at it. One of the issues I outlined in the post linked above from 2 years ago as I was beginning, was that I didn’t find a Go web framework I liked, but https://gobuffalo.io has emerged in the meantime. It is inspired by Ruby on Rails, the documentation says. But I don’t know Ruby on Rails, so what’s meaningful to me is that Buffalo: ...

This isn’t much of an article, the information here is quite readily available elsewhere, but I’m just documenting how I set up a Raspeberry Pi 3 with a removable hard drive to work as a DIY NAS at home. Just so I can come back to it if I have to do it again. Note: was fun and all, and it does work, but not very smoothly, and it’s a bit of a pain. I should pay Dropbox | Google | Amazon and be done with it. Anyway… [hits publish] ...

The Bibliothèque nationale de France (BnF) has opened APIs for a number of services, including their Gallica digital library, which contains 4.3 millions of documents: digitized books, maps, photos, musical scores, etc. from the collections of the BnF and partner libraries. I thought it’d be a nice little project to do a small set of client functions for those APIs: I’d practice Go, and familiarize myself with the APIs. So I started doing that. First result is here: ...

Following an experiment with the Google Cloud Speech API, and another with the Azure Vision API, I continue my experiments in Machine Learning for archival materials with Machine Box. I read Machine Box founder’s Mat Ryer’s post about How (he) built an image proxy server to anonymise images in twenty minutes and based my expriment on his. He uses an image, detects a face in it, removes it and passes the modified image on. And reading the machine box documentation, I saw it also had functionalities to train Machine Box to recognize a particular face, and then use it to find similar faces in other images. ...

Following an experiment with the Google Cloud Speech API, where I tried to extract information from sound files - transcripts of speech, keywords, etc. - I thought I’d try another Machine Learning API, with the same overarching goal but a different project. The overarching goal is still to use the Machine Learning APIs provided by such internet giants as Google or Microsoft and try to somewhat automate the treatment of archival and library material. ...

This past year, we’ve heard a lot about the progress of Machine Learning algorithms and that got me thinking. Libraries and archives have a lot of sound files around, which can’t be easily used: for the most part, if you want to know what’s in them you either have to listen to them, or hope that an archivist did and provided the necessary metadata, which is quite time consuming, and there’s a lot of backlog around… ...