Bloglillon

Bundling some things I posted elsewhere, exploring


Leave a comment

Coursera’s Global Translator community: Transifex subtitle translation interface

NB this post is made to illustrate a reply to Transifex Support about the video player in the subtitle translation interface for the Global Translator Community. Transifex Support sent me a link to their Translating Subtitles in Transifex resource that explains about

It would be great if that were how it works in Coursera’s GTC projects, but it doesn’t.

From November 1 to November 18 2014, in GTC projects, the video player in the translation interface appeared like this (1):

screenshot of a Transifex Italian translation page for GTC subtitles

i.e.  not playing anything, just a black rectangle with the original subtitle overlaid

And since Nov. 19, 2014, the player has entirely disappeared.

[update Nov. 27, 2014: screenshot of the same https://www.transifex.com/projects/p/coursera-android/translate/#it/43/21811835 translation page, without any player:

screenshot of the page indicated above, taken on November 27, 2014

/update (1)]

Now the Translating Subtitles in Transifex resource also says:

If you want to see the default editor view without the video, click the gear icon and uncheck the box next to “enable video editor.”

So I checked if I had done that inadvertently, but I didn’t: the box is checked.

Could it be that the issue lies with the first part, i.e. that the maintainers of Coursera’s GTC projects made an error in adding the video links to create players?

(1) click on the picture to enlarge it


Leave a comment

Coursera’s Global Translator Community: puzzling Translator Agreement(s)

Second update Nov. 6 2014

Access to http://translate-coursera.org/lander/terms.html, mentioned in the description of the D version of the Translator Agreement, has now been blocked, as well as all the other pages of the http://translate-coursera.org/lander subdirectory.

Update Nov. 6 2014

Yesterday, I found out that there is a fourth version of the Translator Agreement. I’m updating this post in italics accordingly

Three (actually four) Translator Agreements

There have been three versions of the Translator Agreement in the 6 months of the GTC:

(For easier comparison, I had put the  first three versions side by side in a table, which can be downloaded from https://pdf.yt/d/9VA7ZY8lU1PYUYnb. Should there be differences between C and D, I’ll make a new table).

While B only corrected some obvious mistakes (typo, wrong contact address), C’s has introduced several important changes. Here I’ll concentrate on those to the part defining the property of volunteers’ translations.

Ownership of volunteer’s translations and “work for hire”

C adds a new sentence at the beginning of this part:

Coursera’s licensors place strict obligations on Coursera to protect their intellectual property rights in the licensed content included as part of Coursera courses.

This addition leads to a rewording of the following sentence: where A and B formerly read:

As between Coursera and you, Coursera owns all right, title, and interest to:
1) the copyright or other intellectual property or proprietary right to the translations and translated works (collectively “Translations”), and
2) the Coursera name and logo.

C now reads:

As a result, Coursera must require its translators to agree under this Agreement, that Coursera owns all right, title, and interest to: [rest of the sentence unchanged]

So what was presented as Coursera’s decision in A and B is now presented in C as something imposed by its licensors, i.e. by the universities and institutions who actually provide the courses.

C also changes significantly the part defining volunteers’ translations as “work for hire”.  A and B read:

YOU EXPRESSLY AGREE THAT ANY TRANSLATION SERVICES YOU PROVIDE WILL BE DEEMED A “WORK FOR HIRE,” UNDER SECTION 101 OF THE U.S. COPYRIGHT ACT, IN EXCHANGE FOR GOOD AND VALUABLE CONSIDERATION, THE SUFFICIENCY OF WHICH IS ACKNOWLEDGED.

But the corresponding part in C says

To the extent that you are creating translations and translated works (collectively the “Translations”) at the request and for the benefit of Coursera, you agree that the Translations will be works made for hire to the extent permitted by applicable law, and Coursera will retain all copyright, patent, trade secret, trademark and any other intellectual property or proprietary rights (“Intellectual Property Rights”) in the Translations

So C replaces the reference to Section 101 of the U.S. Copyright Act with the vaguer “to the extent permitted by applicable law”.In fact, Section 101 of the U.S. Copyright Act actually says:

A “work made for hire” is—
(1) a work prepared by an employee within the scope of his or her employment; or
(2) a work specially ordered or commissioned for use as a contribution to a collective work,(…) if the parties expressly agree in a written instrument signed by them that the work shall be considered a work made for hire.

Definition (1) does not apply to GTC volunteers, as section “8. Relation” says “Nothing in this Agreement creates or will be deemed to create an employer-‐employee, partnership, joint venture, or agency relationship between Coursera and you.” And as to definition (2), Coursera never provided volunteers with a written instrument each should agree to and sign. Therefore the reference to that section of the U.S: Copyright Act did not make sense.

However, after that, the Agreement says:

If any of the Translations do not qualify as works made for hire, you hereby assign to Coursera all right, title and interest and all Intellectual Property Rights in the Translations, and if requested by Coursera will deliver a written assignment any other documents necessary to establish Coursera’s Intellectual Property Rights.

(from the C version, which summarizes the same concept, expressed in a heavier and more long-winded way, in A and B)

In other words, whether their translations qualify or not as work for hire, volunteers must cede all rights on their translations, including their moral rights, to Coursera.

No more “good and valuable consideration”

And then,  C scraps “IN EXCHANGE FOR GOOD AND VALUABLE CONSIDERATION, THE SUFFICIENCY OF WHICH IS ACKNOWLEDGED.”

This  is a soberly realistic deletion, because as explained in the course shell / collaboration platform that Coursera has chosen for the GTC: no recognition at all is granted to volunteers who translate fewer than 2500 words “of consistent  high-quality”.

For those who meet this requirement, there is a mention in the public Meet Our Translators page. That mention is not likely to impress a potential employer. Anyway, that page only lists 1011 [unchanged Nov 2] * volunteers, against “14064* persons contributing” indicated  by Transifex, the online collaborative translation tool chosen by Coursera for the GTC,  except for the 12183* Russian translators, who use a tool made by ABBY LS – see http://coursera.abbyy-ls.com/En – and are therefore not included in the Transifex figure.

Hence, that  “Meet Our Translators” only lists  less than 4% of volunteer translators.  Even by a very conservative estimate of only 100 words translated in average by each of the 24033 unmentioned volunteers, Coursera is getting over 2’400’000 words translated for free, without offering any kind of recognition to their translators.
Then for volunteers who translate 15’000 words, there is a Statetement of Accomplishment, and above 75’000 words, there is a Statement of Accomplishment with Distinction. However, these are the same Statements of Accomplishment that Coursera is presently scrapping for course participants, and which Daphne Koller merrily debunked as mere personal mementoes and explained how to easily fake in a hangout for the GTC on May 15, 2014 (from 18:37 in the YouTube archived version).

Moreover, Coursera uses a script that is meant to assign these forms of recognition by elaborating the statistical data provided by Transifex. Hence the Russian volunteers, who don’t use Transifex, are not covered, and even for the Transifex-using volunteers, this script often doesn’t work properly.

Therefore, since October 23, 2014, Coursera is asking volunteers who are entitled to a form of recognition but didn’t get it, or who translate into Russian, to fill a recognition request … in a Google Drive form: in spite of the fact that it recognizes in its public Web page describing the GTC that these Google Drive forms can’t be accessed in China and in other countries.

So all in all, it does indeed make sense to stop pretending that Coursera offers translators “good and valuable consideration” for their work.

Self-contradiction

All three versions of the Translator Agreement say, under “10. Miscellaneous”:

Except where expressly provided otherwise, the Agreement may only be amended in writing signed by both you and Coursera.

And Coursera has not submitted to GTC volunteers any written text about amendments for them to sign before changing the Translator Agreement from version A to B, and then to C.


* As of Nov. 4, 2014. However the 4% proportion of volunteers getting any form of Coursera recognition has remained fairly stable so far.


Leave a comment

Coursera’s Global Translator Community: background

This is meant to be the first of a series of posts about the Global Translator Community (GTC), which Coursera launched on April 27, 2014: see their Introducing Coursera’s New Global Translator Community blog post.

The GTC is not Coursera’s first attempt at having the subtitles of course videos “crowdtranslated” by unpaid volunteers. So here is some information about its previous attempts, as I experienced them.

Coursera’s Amara team (August 2012 – February 2013)

I first joined a Coursera course because I was puzzled by the issues volunteer subtitle translators were reporting  in the Amara Help forum, mainly about the provided unusable original subtitles automatically produced by voice recognition.

Actually, I first tried to join Coursera’s Amara team, but got no reply to my application. From the outside, it looked like an attempt to copy the settings of  TED’s Amara team, with a workflow of translating-reviewing-approval tasks, but without any of the resources and feedback opportunities TED offers volunteers, and with a big hurdle: the unusable original subtitles that got automatically added (see Ambrose Li’s Things to watch out for if you want to work on Coursera’s subtitles:)

Then at the end of 2012, Coursera staff removed this workflow: anyone with an Amara account could now edit and translate the subtitles of their team’s videos. When they also stopped adding the unusable automatic original subtitles, volunteers were finally able to work normally (see the Amara autocaptions for Coursera videos topic in the Amara – Deaf & Hard of Hearing Discussion List for more details).

However, Coursera staff deleted the Amara team at the end of February 2013, telling volunteers they might copypaste the .srt  subtitle files they produced in subpages of the private Coursera students’ https://share.coursera.org/wiki/ (accessible only when logged in with a Coursera ID), adding that Coursera techies might or might not use them in the courses. Some volunteers complied, some found that a bit too daft and continued using Amara, and linked to the Amara subtitle pages instead.

Coursera staff also announced that they were working on a single tool for translating the site’s interface and the subtitles.

Coursera’s Global Translation Partnership (May 2013 – end unclear)

Then there were no news until May 14, 2013, when they wrote the Coursera Partnering with Top Global Organizations Supporting Translation Around the World blog post: instead of developing their own tool, Coursera was going to use Transifex for both things, and involve only said Top Global Organizations, and only for translations “many of the most popular language markets reflected by Coursera students: Russian, Portuguese, Turkish, Japanese, Ukrainian, Kazakh, and Arabic. Each Coursera Global Translation Partner will begin by translating 3-5 select courses, with the majority of translated courses being available by September 2013. ”

By September 2013, that goal was very far from being reached, possibly because Transifex is a great tool for translating interfaces, but though it can cope with subtitle files, which are just text files with a funny extension, it’s not a subtitle translating tool: no video player where you can check your work on the video. Therefore, as I had pointed out in a May 31, 2013 comment to the mentioned blog post, providing translators with with human-made, accurately revised original subtitles would be essential if they were to be translated with Transifex.

So maybe Coursera didn’t do that, and/or maybe the Top Global Organizations had entered that partnership without realizing how much work in such inappropriate conditions would be entailed. Just a hypothesis, as Coursera hasn’t made public any post-mortem about that partnership – nor about the Amara team initiative, for that matter – and never officially announced its end.

The GTC compared to these previous initiatives

On the one hand, the GTC is a return to the Amara team’s crowdtranslation by any volunteers. However it also has “partners” for some languages, though except for the Lemann Foundation (Brazil), which was already partner in the Global Translation Partnership, the other partners have changed.

As to the original subtitles meant to be translated by volunteers, Coursera has changed paid provider since the Amara team,  now using one who  crowdsources subtitling to humans, as Coursera Staff explained in the May 15, 2014 Global Translator Community Hangout with Daphne Koller (from 50:00 ca – see also the captioned version with transcript), but without telling who is this new partner, let alone how much the crowdsourced subtitlers get paid, if anything.

As with the Global Translation Partnership, the GTC’s  tool for translating subtitles is Transifex.

Another very important convergence between the Global Translation Partnership is the GTC’s Translator Agremeent that volunteers must accept when they apply to join.

That will be for another post: because this agreement is worth a post of its own , and also because presently its page, which is linked to on “terms and conditions” in Coursera’s public Web page describing the GTC, now only yields an XML Access Denied message :

<Error><Code>AccessDenied</Code><Message>Access Denied</Message><RequestId>DC7BD639FE7B7801</RequestId><HostId>Q8soFADQDwa2H+pNlwRFyvBHAAlaIFaRzZsbVbzrj/amDhg62HJko0ajmdpifg8s</HostId></Error>  .

Update November 4, 2014: though I notified both Coursera’s support and the GTC’s admins of this issue on Oct. 28, the link on “terms and conditions” in Coursera’s public Web page describing the GTC continues to produce this AccessDenied message.

What happened is that Coursera assigned a new URL to a new version of the Translator Agreement. They did put the correct short URL – http://goo.gl/W5WY0J redirecting to https://d396qusza40orc.cloudfront.net/translations/updated_Coursera_translator_TOS.pdf – on the 4th page of the Google Drive form for subscribing to the GTC.

However they kept the old obsolete one in the link on “terms and conditions“.


Leave a comment

Corsi Coursera “bloccati da sanzioni economiche US”?

Così appariva la pagina di login ai corsi Coursera in Siria il 28 gennaio 2014. Trascrizione del messaggio:

“Our system indicates that you are trying to access the Coursera site from an IP address associated with a country currently subjected to US economic and trade sanctions. In order for Coursera to comply with US export controls, we cannot allow you to access to the site.”

Cioè: “Il nostro sistema indica che stai cercando di accedere al sito di Coursera da un indirizzo IP associato a un paese attualmente sottoposto a sanzioni economiche e commerciali US. Affinché  Coursera adempia ai controlli di esportazioni US, non ti possiamo autorizzare ad accedere al sito”.

Ulteriori spiegazioni in Update on Course Accessibility for Students in Cuba, Iran, Sudan, and Syria sul blog di Coursera, prima pubblicato il 28 gennaio, ma purtroppo poco chiaro perché frettolosamente aggiornato il 29 gennaio. Cioè:

  • Il 28 gennaio quelli di Coursera hanno impedito agli studenti iscritti residenti a Cuba, in Iran, Sudan e Siria di fare il login per accedere ai corsi per, bloccandoli in base all’indirizzo IP del computer (o di altri aggeggi che vanno online), per ottemperare alle sanzioni economico-commerciali decretate dal governo US. Si scusavano, dicendo di essere in trattative con l’amministrazione US per ottenere il permesso di togliere questo blocco.
  • Il 29 gennaio si sono accorti che per la Siria, c’era un’eccezione alle sanzioni per quanto riguarda le offerte educative, e quindi hanno tolto il blocco per i residenti in Siria.

Cioè: adesso Anas Maarawi e gli altri studenti in Siria possono accedere ai loro corsi Coursera, però quelli residenti a Cuba, in Iran e in Sudan rimangono bloccati. E quelli di Coursera sembrano essere passati ad altro: nessun aggiornamento su quelle annunciate trattative con l’amministrazione US.

Assurdo?

Certo, è una situazione assurda. Corsi universitari offerti gratuitamente non sono patate vendute per soldi. Questa assurdità è stata rilevata nei commenti a quel post Update on Course Accessibility for Students in Cuba, Iran, Sudan, and Syria, nella comunità di Coursera su Google+, e da parte degli enti universitari che forniscono corsi tramite Coursera: vedi ad es. le reazioni dei responsabili dell’EPFL e dell’Università di Ginevra nel riquadro a destra di Des cours en ligne d’universités suisses bloqués par les Etats-Unis (Marc Renfer, RTS Info, 30 gennaio 2014)

Però Coursera è un business

A livello di immagine degli Stati Uniti, l’estensione di quelle sanzioni economiche a corsi online è senza dubbio un disastro. Tuttavia, come visto all’inizio, il blocco per IP imposto da Coursera per ottemperarvi riguarda il login, cioè la cosa che fa che i corsi Coursera non sono affatto MOOC, con la prima O che starebbe per Open, aperto.

In effetti, quando  , da iraniano residente in Iran, ha espresso il suo sgomento davanti a quelle sanzioni nel gruppo Coursera su Google+, gli ho chiesto di vedere se poteva ancora accedere ai contenuti del progetto OpenCourseWare del MIT, non chiusi da login: può.

Sta di fatto che Coursera è un’impresa dichiaratamente commerciale,  e l’obbligo di login per accedere ai contenuti è caratteristica di questa impostazione commerciale. Come rilevava Daphne Koller  in What we’re learning from online education (TED talks, giugno 2012 – dalla trascrizione dei sottotitoli italiani):

Ci sono enormi opportunità da sfruttare da questa struttura. La prima è che ha il potenziale di avere uno sguardo senza precedenti sulla comprensione dell’apprendimento umano. Perché i dati che raccogliamo qui sono unici. Si può registrare qualunque click, qualunque presentazione di compito qualunque post dei forum da decine di migliaia di studenti.

Da qui la necessità, per poter sfruttare questi dati prodotti dagli studenti, di ricollegarli all’identità di ciascuno: l’indirizzo IP del computer va bene per escludere studenti di interi paesi, ma non per identificare con precisione l’autore dei dati prodotti.

Perciò in senso stretto, è giusto che Coursera, essendo un’impresa commerciale statunitense, debba adempire alle sanzioni commerciali statunitensi.

Impegno per l’educazione globale?

Quelli di Coursera amano ribadire il loro impegno per rendere l’educazione accessibile a tutti nel mondo intero.  Daphne Koller lo faceva già in quella conferenza What we’re learning from online education del giugno 2012 e l’hanno puntualmente rifatto in quel post Update on Course Accessibility for Students in Cuba, Iran, Sudan, and Syria del 30 gennaio 2014. E ci crederanno pure,  perché questa globalizzazione fornisce loro tanti dati preziosi da sfruttare. Vedi sopra.

Però quelli di Coursera avevano anche fatto dichiarazioni roboanti sull’internazionalizzazione dei corsi tramite il “crowdsourcing” della traduzione dei sottotitoli delle lezioni video con Amara: vedi ad es. Coursera partners with Amara for crowdsourced captioning (Janko Roettgers, GigaOM, 27 agosto 2012). Avrebbe potuto funzionare se quelli di Coursera avessero optato per una partecipazione aperta, stile Wikipedia,  alla sottotitolazione con Amara. Invece per mesi (vedi la discussione Amara autocaptions for Coursera videos iniziata il 27 febbraio 2013), hanno rovinato quel “crowdsourcing” restringendone l’accesso, poi l’hanno abbandonato a fine marzo 2013.

Due mesi dopo, in Coursera Partnering with Top Global Organizations Supporting Translation Around the World sul blog di Coursera, hanno annunciato la nuova impostazione della traduzione dei sottotitoli: niente più crowdsourcing, ma un partenariato con organizzazioni accademiche “top” che avrebbero dovuto fornire, usando Transifex, “traduzioni della maggioranza dei corsi in russo portoghese, turco, giapponese, kazakh ed arabo” da fine settembre 2013. Ad oggi (6 febbraio 2014), la pagina https://www.coursera.org/courses mostra che su 603 corsi offerti, ce ne sono soltanto 113 con sottotitoli anche in una sola altra lingua che l’inglese , cioè  meno del 20% del totale: altro che maggioranza di corsi tradotti in sette lingue.

Insomma quelle dichiarazioni di impegno per l’educazione globale sanno di “window dressing”, o in italiano, di fuffa. Il ricorso iniziale al crowdsourcing con Amara, poi al programma libero online Transifex, per la traduzione di sottotitoli sono cose trendy che suonano bene. Possono anche funzionare – anche se l’uso di Transifex per la sottotitolazione non è il massimo, visto che non puoi controllare il risultato sul video – però ci vuole un minimo di impegno umano per l’impostazione e l’accompagnamento. E nell’ottica commerciale di Coursera, quell’impegno umano = una spesa nel bilancio. Appena supera una certa soglia, lasciano perdere.

La soluzione ci sarebbe

Come spiega una funzionaria dell’amministrazione US incaricata di far rispettare quelle sanzioni, in U.S. Government’s ‘bone-headed’ decision can be fixed with paperwork, official says (Donald Gilliland, PennLive.com, 30 gennaio 2014) c’è una politica favorevole alla concessione di licenze speciali per le attività a scopo educativo, però queste licenze vanno richieste. edX l’aveva fatto un anno fa e le ha ottenute, salvo per il Sudan, però edX ha lasciato aperto l’accesso  da quel paese ai materiali del corso. In effetti, per l’amministrazione US, è la chiusura dell’accesso tramite login che fa di un corso un servizio commerciale sottoposto a sanzioni. Senza login, i materiali diventano informazione non sanzionata.

Quelli di Coursera invece non avevano chiesto nessuna licenza e ci vogliono sette mesi per ottenerle, secondo la giurista di edX, citata da Gilliland. Però potrebbero togliere l’obbligo di login da Iran, Sudan e Cuba, come quelli di edX hanno fatto per il Sudan.

Visto  che non lo tolgono, c’è da dedurre che quella tesaurizzazione dei dati prodotti da ciascun utente loggato, descritta da Daphne Koller in quella conferenza  del 2012, conti di più per loro del loro conclamato impegno per l’educazione globale?

Certo, il buzz sulle ambizioni globali l’hanno già avuto: la prestigiosa rivista economica Forbes aveva addirittura dedicato un articolo alla traduzione “crowdsourced” dei sottotitoli dei corsi Coursera con Amara, invece ha educatamente taciuto quando quel progetto è andato in vacca. E per quanto riguarda il blocco dovuto alle sanzioni economiche, salvo per l’articolo di Gilliland su PennLive che spiega come sta veramente la situazione, tutti gli altri hanno presentato Coursera come sfortunata vittima di un’assurdità burocratica, anziché della propria incompetenza.

Quindi quanto ai media, quelli di Coursera non hanno da preoccuparsi più di tanto. Quanto allo scontento degli studenti tagliati fuori, beh, ce ne sono tanti altri, no?

Tuttavia ci sono le università partner. Non so le altre, ma nel caso dell’EPFL, P. Aebischer aveva preso 6 mesi di sabbatico dalla sua funzione di rettore per studiare i servizi MOOC esistenti – decidendo cautamente di provare prima sia con Coursera sia con edX. Per il futuro, dopo quel blocco caparbiamente mantenuto quando potrebbe essere tolto?