• How do I access AMI corpus?
  Just login or register here for free and go to "Access AMI corpus" section of the menu.
• Why do I need to register?
  We need to be able to argue that creating the data was time and money well-spent, and so we need enough information to make that case.
• Where is the documentation?
  Full documentation of the corpus is available from Documentation section of the menu.
• Can I browse the data without downloading it?
  Yes. First you need to login or register here for free and go to Browse section to browse the AMI corpus online.
• What annotations have been performed?
  All the documentation about the annotation are visible in the Documentation section.
• Can I download data?
  Yes. First you need to login or register here for free and go to Download section. This page can be used to download everything except high resolution videos - that is, audio, reduced video, transcription and annotations. The chooser allows you to specify what parts of the data you want, to reduce the download size.
• Is there a manifest of the download files?
  You can find out what files should be in your download by looking at the wget script you used to obtain them - it lists each file separately. View it in any text editor. Some files that you might be expecting for a complete corpus might not be present; the HTML page you reach after asking for the download wget script explains any omissions.
• Can I download everything there is?
  No. The full corpus contains high resolution videos that can only be obtained by requesting a firewire hard disk drive. Information about this is available from here for which you need to login or register here for free.
• I can't find the transcriptions - where are they?
  They're in with the annotations, in a zip file on the Download page.
• What consent form have meeting participants signed?
  The original text of the AMI consent form is in 041119-AMI_Meetings_Corpus_Consent_Formjp.pdf.
• What is the best paper to reference if I use the corpus?
  Unless there is a specific publication detailing the aspects of it of you are using, the best compromise for being easy to find and high-level is: Carletta, J. (2007) Unleashing the killer corpus: experiences in creating the multi-everything AMI Meeting Corpus. Language Resources and Evaluation Journal 41(2): 181-190.
• How can I get a printable transcription?

One way is to call up an NXT interface that has a transcription view that you like - you can print from NXT.  To get a simpler transcription view as tab-delimited data, use NXT's FunctionQuery utility at the command line.  After setting up your Java CLASSPATH to find NXT's programs, you can run, for example,

java FunctionQuery -o IB4010 -c AMI-metadata.xml -q '($s segment)' -atts '$s@starttime' '$s@who' '$s'

to get very simple transcription for meetings that only have transcription and no annotation, and, for example,

 java FunctionQuery -o IS1004c -c AMI-metadata.xml -q '($s dact)' -atts '$s@starttime' '$s@who' '@extract(($t da-type):$s >"da-aspect" $t, gloss)' '$s'

to get slightly facier transcription with one line per dialogue act, for coded meetings.  For help with FunctionQuery, see the NXT documentation.
 

• How can I comment or ask questions?
  The best way is to use the forum for registered users, but there is also a contact form.