TOOLBOXBROWSE TOPICS
RESOURCESABOUT THIS SITEpmwiki.org |
Henry Lowood, Ph.D. Conversation with Reid Smith (30-Mar-2007) Silicon Genesis projectAround 2001, we were interested in constructing an oral history of Silicon Valley and give users a number of options for accessing the material: watch a video, read the transcript, search, etc. After a couple of attempts with other companies (including Dragon), we eventually selected Virage software (now called VideoLogger) for video indexing. The software used an IBM speech recognition package that was claimed to be 95% accurate. Our early experiments showed that the speech recognition was probably better than 95% accurate. However, it failed on names (e.g., people, technology). This meant that it wasn't good enough to allow people to read transcripts. In addition, we expected names to feature heavily in the searches people were likely to attempt. As a result, for each interview, we captured the audio from the point in the workflow where Virage poured in its audio transcript and sent tape cassettes to a human transcriber. Then we asked the interview subject to look over the resulting transcript. Finally, we put the text back into Virage. The time codes are sometimes a bit off, but but the "composite" system works well for searching. See Silicon Genesis Interviews Search to see how the system works. The software was purchased by Stanford's streaming service. We found it difficult for the library staff to use (e.g., to construct the search page). As a result, we had to use the streaming service to provide technical assistance. By now, they have had a lot of experience with Virage. We did not do any subject tagging. We assumed that the videos, transcripts and search would be sufficient to enable people to find information of interest. Virage constructs metadata (in what is called a "metadata database"). Henry is involved with another video project: the Machinima archive (animated movies made using game engines). Like this AAAI Video archive, Machinima enables individuals to contribute entries. Henry lightly edits the entries, especially the tags. See Browse by Subject / Keywords. Suggestion re tagging: include the name of the contributor of an entry as a tag Suggestion re archive.org: it is possible to have different views constructed automatically. RGS to follow up on this. We discussed scraping information off the source archives so that AAAI people would not have to manually type in information that exists on other sites for this archive.
We discussed Ed Feigenbaum's Self-Archiving Legacy Toolkit project. The idea is to enable a professor to input his info directly into an archive, with software that takes a first cut at tags, indexing, involves the archivist, etc. This started 20 years ago with the first accession of Ed's papers. Henry taped Ed talking about the boxes and folders of papers. The second accession began in 2005. Ed wants to do something like the original audio commentary again, but this time thinking about the semantic web and using expert systems to assist. Note to Bruce Buchanan (28-Mar-2007): "You should also know that we are currently engaged in several projects that utilize social software (wikis, for example) in archival or historical projects. I've been working very closely with Michael Shanks in this area, who is Professor of Classics (Archaeology), director of the Metamedia Lab and co-director (with me and Jeffrey Schnapp) of the Stanford Humanities Laboratory. A little down the line, after the media conversion aspects are worked out, we might think about whether there is a fit with a suite of projects we are putting together around future archives (which we call "Archives 3.o" or "Animating the Archive")." |

