FIAT/IFTA MMC Seminar 21 - 22 May 2015
Transition from One MAM to the Next and from the Traditional Archivist to Media Manager
Ingrid Veenstra and Cor van Veen - Netherlands Institute for Sound and Vision
Exploring protein-protein interactions by Weak Affinity Chromatography (WAC) ...
FIAT/IFTA MMC Seminar May 2015. Transition from MAM to the Next and From Traditional Archivist to Media Manager. NISV
1. From one MAM to the next
&
From traditional archivist to Media Manager
How the two transitions reinforce each other
Ingrid Veenstra iveenstra@beeldengeluid.nl Cor van Veen cvveen@beeldengeluid.nl
Netherlands Institute for Sound and Vision
5. Once upon a time
No computer, only pen and paper on my desk
No metadata
No internet
Essence: copy items (tape to tape)
6. 2006: MAM 1 in production
Integrate video & audio
We know best what we want
Easy to make changes
7. 2006: MAM 1 in production
Main advantages:
Part of the production chain
Watch video & keyframes in the
MAM
Itemize episodes in smaller
portions
8. Source 1
• Basic metadata on episode level from
scheduling system
Manual
• Divide episode into items with logging tool
Manual
• Annotate items
Archivists work: 2006 - 2012
12. Source 1
• Basic metadata on episode level from
scheduling system
Source 2
• Auto update with item descriptions from
online publication system
Subtitles
• Auto attach subtitles to the items &
automatically generating metadata
Media Management: automation projects
16. Challenges Media management
Monitoring: automated metadata &
new techniques
Communication: give us the metadata
Knowledge (technical skills,
production chain, complex mappings
and annotation techniques)
24. Transitioning from one MAM to the next
&
From traditional archivist to Media Manager
Ingrid Veenstra: iveenstra@beeldengeluid.nl
Cor van Veen: cvveen@beeldengeluid.nl
Netherlands Institute for Sound and Vision
Editor's Notes
Intro/outline:
- In 2012 Sound and Vision started the Media Management programme, a series of projects focusing on:
A proactive instead of reactive approach to archiving / improving quality of metadata earlier on in the production environment;
Getting more metadata by means of automation: mapping to more than one metadata source + automatiocally generating of metadata.
- Simultaneously the process of acquiring a new MAM system started.
- Both processes influenced each other along the way.
- Media Management helped define the requirements for a future MAM.
- Requirements for the new MAM system & Media Management also had great impact on how we fully exploit the possibilities of our current MAM till it’s replaced.
- What you see here = hourglass model of Brecht Declercq (presented on last MMC Seminar) in Amsterdam. On top you see 4 different ways of creating metadata
- Before 2012/media management Sound & Vision focussed solely on full manual metadata, created post ingest.
- This was timeconsuming
- And out of date:
- currently there is so much metadata created earlier on in the production chain.
- And evolution of automated annotation techniques, outgrown it’s experimental R&D phase and ready to go into production.
Nog inkorten.
- In 2012 we started series of (20+) projects focused on proactive and technologically driven approach to archiving
- Goal: increase findability without manual annotation post ingest. Examples
- Projects focused on techniques:
- Embedding thesaurus in production environment to enhance label quality in an earlier stage in the production chain;
- Pulling metadata from more than 1 source system. Cor will talk more about this.
- Clever use of subtitles. Cor will talk more about this.
- Projects focussed on a culture shift:
- Archivists stepping away from their desk to participate in the production environment / to go and actively communicate with the editorial staff about improvement of metadata creation.
- Media Management from 2012 till 2014 was a series of projects. Now it is the reality. Since beginning 2015 we don’t annotate manually anymore and we don’t have documentalist, we have media managers.
- Cor now will tell you more about the transition from traditional archivist to media manager.
Hello, my name is Cor. I am ingest coordinator at Sound and Vision. In the media management programme I led a project focused on clever use of subtitles.
But first, talking about the changing role of the archivist and to put things in perspective…
A long, long, long time ago when I started working as a radio-archivist <click>
I had no computer, just pen and paper on my desk. There was almost no metadata, in fact, I never ever heard of the word metadata. And, nowadays hard to imagine, internet wasn’t available.
<click>
And as for the essence: due to the cost we only archived items instead of the entire programme. We copied items from tape to tape.
<pauze>
Talking about the changing role as an archivist…
For those who are wondering when I started working as a radio archivist…. It was the summer of 89
<click>
Our current MAM iMMix is developed by ourselves in 2006 (twenty-six).
Why did we make our own MAM back then???
There were several reasons:
- Ambition to integrate video and audio within the MAM.
- Self-convidence that we know best what we want.
- Easy to make changes if they’re needed.
We faced a lot of different challenges while developing our own MAM:
For starters we had to learn to communicate with the IT-people who are going to build the MAM. Indeed, that proved to be a challenge. In the early stage one of the people who built our MAM said:
Imagine you are going to buy a new car and you can pick any car you want. You just have to tell us which one.
So we did <click>
We told him this is the car (eeh MAM) we want <pauze>
He said ok and after some time he proudly presented our new homemade MAM <click> <pauze>
When you look closely… Even the colour was wrong?!?! <pauze>
To make it clear. I’m not talking about the quality of the MAM! But of the Lesssons Learned that we had to develop our technical skills (then you get what you say you want to have).
A changing role for the archivists.
In the end our new MAM functioned really well… ok
<click>
Sometimes the MAM was slow, very slow and collegues were yelling and screaming behind there computer
<click>
Sometimes it wasn’t all that stable. But in the end
<click>
We work with iMMix for about 10 years now and I am really proud of it!! People tend to forget the good things, but it was a really big step forward.
<click>
The main advantages were:
MAM was part of the production chain, so we could use a basic set of metadata to start with
Being able to listen or watch video and keyframes in the MAM (for both archivists and (!) customers)
Being able to itemize episodes in smaller portions (for archivists and (!) customers)
<Pauze>
- Let me now take you through the workprocess of an archivist as we knew it befor media management <click>
This is an overview of the process during the period 2006 (twenty-six) – 2012 (twenty-twelve)
<pauze>
<pauze>
Let me illustrate this with a concrete (but fake) example
The example is especially created in English for this occasion <click>
Source one: a basic set of metadata ingested from the broadcasting scheduling system
In this stage only information like: title, broadcasting company and a, sort of, high level summary
<click>
Then archivists enrich the metadata:
At first by dividing episodes in smaller portions with the help of a logging tool and keyframes or video.
After this:
<click>
(logging tool = internationale term voor VSE of aanmaken selecties m.b.v. de player)
Further manually annotating by the archivist such as:
Adding a detailed description (f.e. summary and shots)
Labeling persons (king Willem-Alexander and queen Maxima), labeling locations (Germany) etc.
<pauze>
<pauze>
Now let me show you the same example, but then with automatically ingested metadata
<click>
<Pauze>
Step 1 is the same as before: a basic set of metadata is ingested.
Step 2 is the result of Media Management. Now we get metadata updates from a second source, used by the Dutch Broadcasting Association for online publication of episodes and items. Before we annotated these items manually
Step 3 we are adding subtitle texts to the items that were automatically ingested in step 2. I was involved in this project.
Let me shortly visualize these steps:
<click>
Step one: Basic metadata set, the same as before
<click>
Step two: items are automatically ingested. As you see it contains less information than the manually annotated items
<click>
But in step three the episodes subtitles that match the time-in and time-out point of the item are added to the item.
After this we can enrich the item with detailed information using the subtitles and new techniques.
In this example king Willem-Alexander and queen Maxima are automatically labeled as persons.
<pauze>
Besides that we use new techniques like speaker-labeling. Our collegue Josefien Schuurman will provide more details about these kind of techniques tomorrow.
<click>
With these new techniques we face some new challenges <click>
Firstly we have to monitor the automated metadata flows and make sure they reinforce each other;
<click>
Secondly, communication and communication with the broadcasting companies about the metadata they have to provide us with.
<click>
We also have to keep our knowledge of the production chain, complex mappings and annotation techniques up to date <pauze>
And last but not least, we have to make sure that Media Management will still work properly in MAM number 2: a totally new environment for both media managers and our end users
Ingrid will now tell you more about the transition to MAM number 2
Background of MAM 1:
- It is a Cauliflower: We developed constantly. Different departments had different wishes and new not always consistent modules and functionalities were added.
- This lead to quality and perfo4mance issues.
Question is: how long can you keep on developing without creating a grotesque cauliflower.
The conclusions of research of the sustainability of our MAM were clear: the code is not consistent; a total re-write of our software is needed.
Because MAMs since 2006 evolved tremendously, in 2012 we decided to find a standard MAM solution.
Thanks to the MM programme the requirements for MAM2 started to become apparant.
- Because of the total cost of such an acquisition, we were bound by European rules. We had to follow the guidelines of a European tender process.
- Because we couldn’t choose our vendor, we had to carefully predefine our requirements so we would end up with the best solution.
- We did that in the RFI process: together with all departments we formulated requirements around our core processes.
- Worked together with a lot of people, spend a lot of time. RFI is a very large document.
- For the RFP we applied the pressure cooker methode: we worked very intensively together with a smaller project team.
- Interestingly enough in it’s concept phase the RFI was nothing more than a detailed description of the requirements our first MAM already fulfilled.
- Later on in the process we opened up more for different, new solutions.
- And also we included new insights as a result of the start of the Media Management programme in the RFP.
- That resulted in an RFP that was less focused on requirements on the annotation module (the core of MAM1) and more on automated metadata processing and workflows.
- Result of tender: a enormous contract and set of reqquirements. And: Vizrt. They offered an all in one solution
- Vizrt supplies us with a MAM for internal use (VizOne) and via their partner Xailabs also with an external portal for media professionals.
Not to forget: The workflow solutions offered by Vizrts partner Mayam, the company of our keynote speaker this day, have a crucial role.
Worflows are important part of the MAM solution we envisioned, e.g.:
- Metadata workflows and quality monitoring flows that support our media managers;
Auto updates of metadata from more than 1 source system;
Triggering of metadata analysis processes such as extraction of labels from subtitles.
- As said before the MM projects were important input for formulating the requirements for MAM 2
- Here you see a few oif them
- What is striking is that the requirements are formulated in a very general way if you compare them to the processes that Cor demonstrated.
- Reason for this is twofold:
- We wanted to give our vendor the opportunity to come up with different, possibly better solutions.
- The media management projects were not finished yet when we formulated the requirements, so we didn’t know all the details.
- Now we are much further with Media Management, partly thanks to the RFI and RFP process.
And here pops up a paradox as a result of the intertwining of the Media Management and future MAM projects:
Now we are much further with Media management and we concretized our requirements for 2016, we also know much better what we want with our current MAM to support Media Management in the two years we are still ‘stuck’ with it.
- Media Management as a programme ended in 2014 and was implemented, together with agile development methods (scrum), as a regurlar way of working in the beginning 2015;
- At the same time we knew in the beginning of 2015 that in the end of May (this month) we will effectuate a freeze on developments on our MAM, because we have to prepare for the migration to the next MAM.
- What happened: beginning this year up till now an enourmous development boost is taking place.
- We are implementing all the results of media managemment and develop even more fantastic feautures on top of this: speaker labeling, term extraction, visual support by clever use of keyframes, et cetera.
- Paradox is that now we have the luxurious position to develop a cauliflower on top of a cauliflower.
- Or a villa on top of an apartment building. The apartment building will be evacuated anyway in 2016, so this gives us a lot of space to experiment on using our current MAM to it’s full potential.
- This also comes with a risk:
- Can our current MAM in it’s last decadent state seemingly offer more than the future MAM (let’s hope it doesnt collapse before 2016!);
- How will our users respond to the rapid changes that are implemented now and the new changes they have to get used to again when switching to a totally different system and user interface?