[Date Prev] [Date Next] [Thread Prev] [Thread Next] Indexes: Main | Date | Thread | Author

[ba-unrev-talk]: Tech Impediments? ... see Ka-Ping Yee's ZEST: AutomaticOutlining for E-mail Discussions (a proposal to incl. IBIS capabilities intoRoundup)




Ref. <http://www.cs.berkeley.edu/~pingster/vertex/zest.html>    (01)

Ka-Ping Yee's Proposal:    (02)

        [ ...]. "I propose a new summarization technique for e-mail
        discussions, based on a conversation with Terry Stanley in 1997.
        This technique does not require users to do any more work;
        rather, it extracts structural information from the way that
        messages are already written.    (03)

        [...]." The technique threads messages in sections by looking
        for paragraph breaks and quoted text, instead of threading at
        the granularity of entire messages. We can assume that a
        paragraph immediately following a quoted section is a reply to
        that quoted text. In the summary, a section is represented by
        the sentences in its first two lines. This provides a much more
        relevant overview of the content than simply repeating the
        subject line. To keep the summary concise, not all sections
        appear. Any section that replies to or is replied to by another
        section appears; also, the first section of each message not
        otherwise represented by some section appears. The result is a
        point-counterpoint outline of the discussion that reads like a
        conversation. Informal tests show that this technique can
        produce a reasonable outline from real e-mail written by authors
        with no knowledge of this tool.    (04)

        The basic technique is easily extended to allow users to easily
        turn their outline into a structured argument map using a simple
        typographical convention. By inserting a textual symbol (which
        we call a "criticon") at the beginning of any paragraph, users
        can tag the paragraph as one of four types: [?] marks a
        question, [#] marks an informative comment, [+] marks a
        supporting position, and [-] marks an opposing position. This
        can also be extended to accommodate the informal voting
        conventions that already exist on some mailing lists (for
        example, the Apache and Python software development groups
        already signal approval and disapproval in e-mail by writing
        "+1" or "-1")."    (05)


        IMPACT ANALYSIS    (06)

        [...]. "This technology has the potential to transform every
        mailing list into a cost-free reusable knowledge repository. The
        outline becomes a collaborative workspace to which everyone
        contributes. For example, in a design group it can serve as a
        design history; in a political discussion forum it can become a
        record of debate."    (07)


        INNOVATION    (08)

        "Current mailing list tools [Mailman] and bulletin-board systems
        [Slashdot] arrange messages in threads to show chains of reply.
        In the thread view, messages are typically presented by showing
        their subject line, yielding a view where the same subject line
        is repeated many times. Because users still have to visit and
        read each message in turn in order to understand what has been
        said, the only way they can gain context is to read an entire
        archive.    (09)

        Structured discussion frameworks have been previously proposed
        [Flores], and graphical tools for interacting with them have
        even been prototyped [Conklin]. However, they impose a rigid
        formalism that the users have to learn, and they require all
        interaction to take place within a special software system. This
        led to only limited use of the tools. The technology proposed
        here is designed to work with existing e-mail practices, which
        have an extremely broad audience. Using the techniques proposed
        here, we can bring the power of structured discussion systems
        like IBIS [Kunz] to a wide audience without the associated
        inconvenience and rigidity.    (010)

        Many other visualizations of online discussions have been
        developed [Donath, Rohall]. However, in all such visualizations,
        the units of representation are entire messages. Overviews in
        these visualizations show no message content, only metadata such
        as the subject line or author."    (011)


        DEVELOPMENT PLAN    (012)

        "A prototype tool has already been developed that works with
        existing e-mail folders in the standard Unix text format. It
        generates one HTML file for each message and one for the
        outline. The program flexibly handles a variety of quoting
        styles, including those generated by most popular e-mail clients
        (such as Pine, Outlook, and Emacs). Figure 1
        <http://lfw.org/ping/criticons/> shows an example of the output
        from running the prototype tool on mail from a real mailing list.    (013)

        I plan to develop this tool as an open-source package to be used
        with Mailman, the GNU mailing list manager. Informal feedback
        from the Mailman and Python communities has been positive and
        enthusiastic.    (014)

        User testing will be conducted to optimize design variables such
        as layout, the amount of content to include in the outline view,
        and the best order in which to present points in the outline."    (015)