Building20 Memories

 small screenshot Building20 Memories was an idea for using technology to help a community to tell its collective story without reporters or editors. The general approach is to collect small memories from members of the community into a database and have the system present and link these memories along several dimensions — who, when, where, etc.

The idea was that the automatic organization of individual memories will allow a large database of them to be presented in coherent ways, and that the presentation would trigger associative memory in the readers, who would then feel encouraged to contribute new memories.

 historical aerial photo of MIT's Building 20 MIT's Building 20 was a humble structure with a proud history. Some of this history is documented in MIT's Building 20: The Magical Incubator, which was edited by Prof. Paul Penfield, Jr.. (Prof. Penfield graciously allowed the reminiscenses on that site to be used to prime the Building20 Memories system. I later obtained permission from most of the individual contributors to make their contributions publicly available in the Building20 Memories system.) Building 20 has been razed since this project was done, to make way for the new Stata Center.

Building20 Memories was my final project for a class in Spring 1998. The prototype is implemented in Lisp-based Web CGI, and was used by over a dozen people who worked in Building 20.

Most of the semantic tagging of memories was done by a human operator. Examination of the tagging thus far suggested that the majority of it could be done by computer algorithms with a sufficient degree of accuracy. I implemented a tagging parser for 'fuzzy' dates and date ranges, and designed (but did not have time to implement) some algorithms for tagging people and places.

Preliminary results with manual semantic tagging of the reminiscenses for Building20 Memories included such gems as the system grouping together the memories of a woman seeing a little girl riding her tricycle in the hallways with the memories of another woman, who was that little girl.

If you are an MIT undergraduate would would like to pursue outgrowths of this as a UROP, please contact me. I'm also open to collaborating with researchers at other institutions.

© Copyright Neil Van Dyke      Contact