An examination of automatic video retrieval : technology on access to the contents of an historical video archive

Purpose

To provide a first understanding on the constraints historical video collections pose to video retrieval technology and the potential that an online access offers to both archive and users.

Design/methodology/approach

A small and unique collection of videos on customs and folklore was used as case study. Multiple methods were used to investigate the effectiveness of technology and the modality of user access. Automatic keyframe extraction was tested on the visual content while the audio stream was used to automatic classification of speech and music clips. The user access (search vs. browse) was assessed in a controlled user evaluation. A focus group and a survey provided insight on the actual use of the analogue archive. The results of these many studies was then compared and integrated (triangulation).

Findings

The amateur material challenged automatic techniques for video and audio indexing thus suggesting that the technology must be tested against the material before deciding on a digitization strategy. Two user interaction modalities, browsing vs. searching, were tested in a user evaluation. Results show users preferred searching but browsing becomes essential when the search engine fails in matching query and indexed words. Browsing was also valued for serendipitous discovery; however the organization of the archive was judged cryptic and therefore of limited use. This indicates that the categorization of an online archive has to be thought in terms of users who might no understand the current classification. The focus group and the survey showed clearly the advantage of an online access even when the quality of the video surrogate is poor. The evidence gathered suggests that the creation of a digital version of a video archive requires a re-thinking of the collection in terms of the new medium: a new archive should be specially designed to exploit the potential the digital medium offers. Similarly user’s needs have to be considered before designing the digital library interface as needs are likely to be different than those imagined.

Research limitations/implications

The study is limited to a single archive; other institutions provided the technology used on which we had no influence.

Practical implications

The guidelines drawn as result of this study could impact on the strategy and policy of digitization projects that include video archives.

Originality/value

This paper is the first attempt to understand the advantages offered and limitations hold by video retrieval technology for small video archives like those often found in special collections.