Improving problem-oriented mailing list archives with MCS

摘要

Developers often use electronic mailing lists when seeking assistance with a particular software application. The archives of these mailing lists provide a rich repository of problem-solving knowledge. Developers seeking a quick answer to a problem find these archives inconvenient, because they lack efficient searching mechanisms, and retain the structure of the original conversational threads which are rarely relevant to the knowledge seeker.

We present a system called MCS which improves mailing list archives through a process called condensation. Condensation involves several tasks: extracting only messages of longer-term relevance, adding metadata to those messages to improve searching, and potentially editing the content of the messages when appropriate to clarify. The condensation process is performed by a human editor (assisted by a tool), rather than by an artificial intelligence (AI) system.

We describe the design and implementation of MCS, and compare it torlated systems. We also present our experiences condensing a 1428 message mailing list archive to an archive containing only 177 messages (an 88% reduction). The condensation required only 1.5 minutes of editor effort per message. The condensed archive was adopted by the users of the mailing list.