GitLab

The speech's text was duplicated, with a few very minor typographical
differences.  I went with the second copy.  This problem exists in the
source at the Miller Center site, so we should report it to them, but
in any case we might as well solve it here right away.

Thanks to Jason Levy for reporting the bug to me.

Also, remove the note in README.md about alternative sourcing for his
2019-09-25 speech before a press conference.  That speech's transcript
is now available from the Miller Center, who presumably got it from
the same White House web site page I got it from because it was
byte-for-byte the same.

As promised in commit e3312588, remove the remaining HTML elements
(mostly "<p>" and "<br>" and such) from the speech transcripts:

    $ pushd ~/src/html2text/

    $ git remote show origin | grep "Fetch URL:"
      Fetch URL: git@github.com:aaronsw/html2text.git

    $ git log -n1 | grep "^commit"
    commit 8ddc844b03042faf4875d465580ad122f9292978

    $ popd

    $ for name in *; do                                                    \
        ~/src/html2text/html2text.py --body-width=0 ${name} > ${name}.tmp; \
        mv ${name}.tmp ${name};                                            \
        git add ${name};                                                   \
      done

Thanks, Aaron.

Give all the speech transcripts Unix-style line endings:

    $ cd data

    $ dos2unix --version
    dos2unix 7.3.4 (2016-05-24)
    With Unicode UTF-16 support.
    With native language support.
    LOCALEDIR: /usr/share/locale
    http://waterlan.home.xs4all.nl/dos2unix.html

    $ for name in *; do dos2unix ${name}; done

Replace each HTML page with a text file (now with ".txt" extension)
that contains just the "President: <NAME>" line at the top and the
transcript of the speech.  There are some basic HTML formatting
elements still embedded in the speech texts, but they can be dealt
with in a subsequent commit; let's take this one diff at a time.

This commit reflects the output of a truly marvelous Emacs macro,
which unfortunately this margin is too small to contain.

In one, add the missing "Transcript" marker; in the other, add the
missing presidential identifier element.