Sherlock Holmes on the Radio Airwaves

Corpus Preparation

The radio scripts were sourced from the Generic Radio Workshop, where there were downloadable plain text (.txt) files that includes the body contents of the scripts. For the purpose of this research, metadata that was excluded from the plain text files at the top of the webpage display were also included into the files. There was a regular pattern of "series", "show", "date", and "cast", which led to the decision of strictly limiting the Relax NG schema code in this specific order. There were some discrepancies within the files, most notably being with "Murder in Casbah".

• There were some major differences in the structure of the files in the radio script corpus, that being of:
  • some scripts included a sound effect and special effects list
  • some scripts have notes that indicate the author or special notes
  • the order of the elements in the metadata columns would be different depending on whether there was a sound effect list, special effect list, and notes
  • most scripts have the regular pattern of [SOUND/MUSIC:] to indicate the start of a sound effect or musical tune, some have [SOUND FX] or [music]
  • there were minor typos such as ; and FC
  • some lines do not have a speaker attached, most likely being part of a larger paragraph of speech that was broken into separate lines in the script

The content of the radio scripts had to be regularized during mark-up so that queries could be performed easily on them in the future, taking into consideration that any changes would not affect the contextual or structural analysis of the project.

• The following changes to the corpus were made:
  • a <lineGrp> element was added to wrap around the spilt lines
  • a <line> element was added to wrap around the lines in <lineGrp> to preserve the original structure of the script
  • all lines were wrapped in <ln> for consistency
  • all metadata elements were formatted to be in the same order, regardless of the order it was in the original text
  • SOUND/MUSIC tags from the original text were removed since <sound> and <music> served the same purpose.

Stitchwork

The first step of analysis for the corpus was to mark the changes to the story from Arthur Conan Doyle's publication for the radio scripts. Text from Arthur Conan Doyle's original publication was sourced from Project Gutenberg, which allowed me to copy the text needed for the specific chapters into an XML file. For the initial stage of the analysis portion, I focused on the story "A Scandal in Bohemia", where there was a radio recording, the radio script, and original publication are available online for comparison purposes. XSLT was used to plant <xml:id>s in the Arthur Conan Doyle texts.

<xsl:template match="p">
        <p xml:id="{ancestor::xml/@xml:id}-p{count(preceding::p) + 1}">
            <xsl:apply-templates/>
        </p>
</xsl:template>

These <xml:id> tags were used to stitch together the <ln> elements in the radio script to show the correlation between the two files. The XSLT allowed automatic tagging for the Arthur Conan Doyle texts, which then gave a pointer for me to use to manually tag segments of specifc paragraphs from the text

Manual tagging was used for the segmented portions of the story where the changes between the two source files were not similar enough to be stitched together by the original paragraphs.

• In the Arthur Conan Doyle text, the <xml:id>'s are structured as followed:

            <p xml:id="SIB-p4">His manner was not effusive. It seldom was; but he was glad, I think,
                to see me. With hardly a word spoken, but with a kindly eye, he waved me to an
                armchair, threw across his case of cigars, and indicated a spirit case and a
                gasogene in the corner. Then he stood before the fire and looked me over in his
                singular introspective fashion.</p>

• With manual tagging, the <seg> tags are structured as followed:

                <p xml:id="SIB-p7">“Indeed, I should have thought a little more. Just a trifle more, I
                fancy, Watson. 
                <seg xml:id="SIB-p7-s1">And in practice again, I observe. You did not tell me that you
                intended to go into harness.”</seg>
                </p>

• After the <seg> tags are put in place, I then stitch the radio script <ln> tags together manually after comparing the texts side-by-side, by using the @pull attribute, as followed:

                <ln pull="#SIB-p7-s1"><speaker>HOLMES</speaker> And in practice again, I see. You didn't tell me that you'd gone back into harness.</ln>

• For some <ln> elements, it matches to several <seg> elements. The multiple <seg> elements are separated by white spaces in the following structure:

 
                <ln pull="#SIB-p44-s1 #SIB-p44-s2 #SIB-p46-s1">
                <speaker>HOLMES</speaker>
                This is my friend and colleague, <mention ref="watson">Dr.Watson</mention>. You may say anything before him that you can say to me. Whom have I the honor to address?</ln>

• Additional attributes are added to the <ln> tags to signify the nature of the relationship between the two text files.
  • @change is used to show the nature of the change by using the values "wording", "interpel", and "sig".

  
                  <ln pull="#SIB-p55" change="wording">
                      <speaker>HOLMES</speaker> 
                      That had not escaped me either, sir. In fact, if you will state your case, I shall be better able to advise you -- your Majesty.</ln>

  • @type is used to show unmatched lines that were significant to the radio script but do not match up to any part of Arthur Conan Doyle's publication, by using the value "unmatched".

  <ln type="unmatched" change="interpel sig"><speaker>WATSON</speaker> How do you do, sir?</ln>