appendices | bibliography
back to rolandht

RolandHT – Appendices

Appendix A. List of primary sources in RolandHT
Appendix B. Expressing Semantic Information in HTML
Appendix C. Cleaning Up First-Pass XML Encoding
Appendix D. Theme Statistics
Appendix E. XML
Appendix F. Semantic Code Structure and Interface
Appendix G. Sample Teaching Modules

Appendices in PDF


Appendix A. List of primary sources in RolandHT

The following is a list of the primary sources encoded in RolandHT. Most of them are text-based; works composed in other media are designated as such in the "Title/Name" column (see, for example, Angoulême Carvings).

Year Author Title/Name Geo. Origin
~817-836 Einhard The Life of Charlemagne France
1095-99 unk. Song of Roland France
~1100 Pseudo-Turpin History of Charles the Great and Orlando France
1100s unk. Angoulême Carvings, stonework France
~1150 Pfaffe Konrad Rolandslied Germany
~1180 Bertrand de Bar-Sur-Aube The Song of Girart of Vienne France
XII-XIIIc unk. Roland and Oliver, Chartres Cathedral statues France
XII-XIVc unk. Firumbras England
XII-XIVc unk. Reims Triptych, stone carving France
XIIIc unk. Karlamagnús Saga Norway
XIVc unk. The Middle English Song of Roland England
early XIVc unk. Otuel and Roland England
1300s unk. Rouland and Vernagu England
1308-21 Dante Alighieri Divine Comedy Italy
1300-1600 unk. Cân Rolant Wales
XVc unk. Dubrovnik Roland, statue Croatia
1471-86 Matteo Maria Boiardo Orlando Innamorato Italy
1532 Ludovico Ariosto Orlando Furioso Italy
1572 unk. The Tale of Ralph the Collier Scotland
1605-15 M. de Cervantes Saavedra Don Quixote Spain
1810 William Sotheby Constance de Castille England
1820 Thomas Campbell The Brave Roland England
1831 Laetitia Elizabeth Landon Roland's Tower. A Legend of the Rhine. England
1839 Emmeline Stuart-Wortley The Tower of Roland England
1849 Henry B. Hirst The Penance of Roland USA
1849 William Motherwell Roland and Rosabelle Scotland
1855 Robert Browning Childe Roland to the Dark Tower Came England
1860 William Caldwell Roscoe Eliduke, Count of Yveloc England
1868 Thomas Westwood An Angler's Dream Under Rolandseck England
1875 Albert B. Barrows Roland of Algernon USA
1901 Robert Williams Buchanan The Death of Roland England
1903 John Warren Roland at Roncesvalles England
1911 Maurice H. Hewlett style="font-style:italic;"The Birth of Roland England
1930 Benjamin Low Roland, A Symphonic Poem USA
~1941-45 League of Roland Roland: Country First England
1942 Adair Forrester The Children's Story of Roland England
1949 Peter Racine Fricker Rollant et Oliver England
1975 L. Sprague de Camp The Compleat Enchanter USA
1978 Warren Zevon Roland the Headless Thompson Gunner USA
1982-2004 Stephen King The Dark Tower I-VII USA
1994 Gianni Celati Orlando Innamorato Italy
1995 Greg Roach The Madness of Roland USA
1999 Shayne Amaya et al. Roland: Days of Wrath USA/Brazil

Table 1. Primary sources encoded in RolandHT.

Notably absent from the above list is Shel Silverstein's 1973 "Roland the Roadie and Gertrude the Groupie," written for the rock group Dr. Hook and the Medicine Show. Although parts of the song are encoded, they are commented out because permission to use the work online has been denied by the copyright holder. Encoding data from Silverstein are included in the statistics presented in this dissertation.
back to top

Appendix B. Expressing Semantic Information in HTML

In the first – HTML-based – version of RolandHT, semantic information was encoded into the project by way of toggling (using JavaScript) visibility of such information when the user clicks on a given link. For example, the Song of Roland excerpt titled "knight, ambassador, coward" has a hyperlink around Roland's words to his stepfather Ganelon: "I love you not a bit":

Figure 1
Figure 1. "Knight, ambassador coward" hyperlink.

Love in general, and its lack between Roland and Ganelon in particular, is a recurrent theme in the corpus. An interesting exception to this is the Karlamagnús Saga episode in which the two meet, become kin, and swear fellowship to each other. To highlight this, clicking on the link brings up an overlaid comment, "Ah, but they loved each other once," complete with a link to the excerpt titled "how rollant met guenelun":

Figure 2
Figure 2. "Knight, ambassador coward": hyperlink clicked.

This approach proved frustrating for most of the twenty or so informal beta testers of the project in 2001: every link required two clicks instead of the customary one, considerably slowing down the reading process. In addition, although the prose blurbs conveyed thematic information, there was no categorized theme set. Consequently there was no way to see, for example, a list of all the excerpts treating the subject of love.
back to top

Appendix C. Expressing Semantic Information in HTML

Having encoded themes and imagery that appeared interesting at the time of encoding, I analyzed the statistics of their occurrence in the corpus. The following themes were deleted (number of occurrences in parentheses):

Deleted themes
army (4) pity (1)
ceremony (3) prudishness (1)
conspiracy (1) ritual (2)
disguise (1) spite (3)
denial (2) stubbornness (3)
gift (3) summons (1)
hostility (2) temper (4)
jealousy (2) threefold repetition (2)
mastery (1) writing (2)
Table 2. Deleted themes and imagery.

I also consolidated some rarely occurring theme elements and attributes (marked with an @ symbol before the attribute name) and semantically similar or more appropriate ones:

Original theme/language Folded into...
arrogance overconfidence
adultery kinship
council counsel
cruelty violence
crusade violence
gore violence
incest kinship
invulnerability unconquerable
martyrdom sacrifice
massacre violence
prophecy magic
prudishness (1) chastity
prudishness (1) chivalry
ritual (1) chivalry
rolands priorities chivalry
theft treachery
trust fellowship
family ties kinship
@imagined['yes'] @realized['no']
Table 3. Consolidated themes and imagery.

Working on encoded places and character names (which, being recurrent and/or echoes of each other, are one of the elements that tie the corpus together) led to regularization of their names to their presently most widely used forms. Aix and Aix-la-Chapelle became Aachen. Among characters, Basile was changed to Basil, Balsan – to Basin, Blancandrin – to Blancadrin; Geluvis to Geluviz, Gilem to Gille, and Olivieri to Oliver. Problems arose during regularization, some of them of a surprisingly political nature. For example: do I encode a river as located in Taiwan, or in China? (The answer to this was ultimately China, because both of the following conditions are true: the primary source had explicitly placed it there; and the river was in China at the time of the book's writing.)

The other significant challenge was posed by the name of Roland's mother. Her names are sometimes wildly disparate and clearly not variations upon one another. In the corpus, she is called in turn Gille, Gilem and Bertha. Bertha occurs only once, but this is a common name for her in most of the Mediterranean, and so it seems unwise to fold Bertha into Gille. Gilem, however, was folded into Gille.

Places acquired a new attribute – @where – which denotes the country (if working with a city, church, etc.) or the continent (if it is a country, mountain, etc.) within which it is located. This will allow users to search not only for specific places, but for what is located in, for example, Asia.
back to top

Appendix D. Theme Statistics

To arrive at the following statistics, I have broken the corpus works down into four temporal groups: medieval (designated MVL below, works written in or before 1350); Renaissance (REN, 1351-1650), modern (MOD, 1651-1899) and contemporary (CON, 1900-present). For each theme, the number columns represent the number of times that theme is encoded in works from the relevant time period. The percentages were calculated using the total number of <theme> element occurrences in the works from that period. Note that the same theme was, as likely as not, encoded multiple times in the same excerpt, so the numbers below do not represent the number of excerpts in which these themes occur.

Statistics for the top five most often occurring themes for each time period are rendered in boldface.

Theme MVL # MVL % REN # REN % MOD # MOD % CON # CON %
accusation 44 4.64% 4 2.72% 22 4.24% 9 4.10%
anger 30 3.16% 5 3.40% 5 0.96% 3 1.36%
beauty 5 0.53% 0 0.00% 0 0.00% 0 0.00%
betrothal 3 0.32% 0 0.00% 4 0.77% 0 0.00%
chastity 2 0.21% 2 1.36% 3 0.58% 1 0.45%
chivalry 22 2.32% 4 2.72% 19 3.66% 7 3.18%
combat 105 11.06% 17 11.56% 40 7.71% 35 15.91%
conquest 6 0.63% 0 0.00% 7 1.35% 2 0.91%
counsel 3 0.32% 1 0.68% 0 0.00% 0 0.00%
courage 30 3.16% 6 4.08% 16 3.08% 1 0.45%
cowardice 9 0.95% 6 4.08% 2 0.39% 1 0.45%
death 65 6.85% 3 2.04% 80 15.41% 12 5.45%
deceit 11 1.16% 7 4.76% 0 0.00% 5 2.27%
defiance 1 0.11% 0 0.00% 2 0.39% 0 0.00%
diplomacy 8 0.84% 0 0.00% 0 0.00% 0 0.00%
dream 6 0.63% 2 1.36% 6 1.16% 1 0.45%
evil 2 0.21% 0 0.00% 4 0.77% 0 0.00%
fear 8 0.84% 1 0.68% 2 0.39% 0 0.00%
fellowship 40 4.21% 0 0.00% 25 4.82% 11 5.00%
glory 4 0.42% 0 0.00% 5 0.96% 0 0.00%
grief 31 3.27% 6 4.08% 25 4.82% 5 2.27%
honesty 1 0.11% 0 0.00% 4 0.77% 0 0.00%
honor 9 0.95% 0 0.00% 9 1.73% 0 0.00%
insult 5 0.53% 1 0.68% 3 0.58% 0 0.00%
journey 3 0.32% 6 4.08% 0 0.00% 0 0.00%
kinship 3 0.32% 0 0.00% 5 0.96% 0 0.00%
knighthood 13 1.37% 0 0.00% 20 3.85% 7 3.18%
lament 9 0.95% 0 0.00% 2 0.39% 0 0.00%
love 29 3.06% 15 10.20% 26 5.01% 16 7.28%
loyalty 19 2.00% 1 0.68% 15 2.89% 7 3.18%
madness 14 1.48% 13 8.84% 2 0.39% 14 6.36%
magic 8 0.84% 1 0.68% 3 0.58% 25 11.36%
marriage 10 1.05% 1 0.68% 10 1.93% 2 0.91%
monjoie 4 0.42% 0 0.00% 0 0.00% 0 0.00%
nobility 11 1.16% 0 0.00% 9 1.73% 2 0.91%
omen 20 2.11% 0 0.00% 1 0.19% 0 0.00%
overconfidence 3 0.32% 2 1.36% 0 0.00% 0 0.00%
pain 5 0.53% 1 0.68% 9 1.73% 0 0.00%
piety 25 2.63% 1 0.68% 0 0.00% 8 3.64%
pride 21 2.21% 1 0.68% 8 1.54% 0 0.00%
protection 13 1.37% 1 0.68% 5 0.96% 1 0.45%
quest 5 0.53% 2 1.36% 0 0.00% 0 0.00%
religion 127 13.38% 11 7.48% 43 8.29% 11 5.00%
revenge 9 0.95% 0 0.00% 3 0.58% 1 0.45%
sacrifice 4 0.42% 1 0.68% 2 0.39% 1 0.45%
shame 26 2.74% 2 1.36% 9 1.73% 0 0.00%
storytelling 1 0.11% 0 0.00% 1 0.19% 0 0.00%
strength 24 2.53% 8 5.44% 15 2.89% 4 1.82%
threat 22 2.32% 3 2.04% 8 1.54% 2 0.91%
tower 0 0.00% 0 0.00% 2 0.39% 0 0.00%
treachery 24 2.53% 3 2.04% 11 2.12% 4 1.82%
unconquerable 7 0.74% 3 2.04% 0 0.00% 0 0.00%
violence 30 3.16% 4 2.72% 16 3.08% 11 5.00%
virtue 3 0.32% 1 0.68% 4 0.77% 2 0.91%
weakness 3 0.32% 0 0.00% 7 1.35% 0 0.00%
wisdom 4 0.42% 1 0.68% 0 0.00% 0 0.00%
Total 949 100% 147 100% 488 100% 178 100%
Table 4. Theme statistics.

back to top

Appendix E. XML

The World Wide Web Consortium (W3C) describes XML documents as "made up of storage units called entities, which contain either parsed or unparsed data. Parsed data is made up of characters, some of which form character data, and some of which form markup. Markup encodes a description of the document's storage layout and logical structure. XML provides a mechanism to impose constraints on the storage layout and logical structure." (XML 1.0)

An element, the most widely used kind of XML entity, is enclosed in angle brackets. It consists either of an opening and a closing tag enclosing some content –

<element tag="opening">This is an element.</element tag="closing">

– or of a singleton tag "closed" by a forward slash:


Elements are allowed attributes, which function as sub-categories to the elements' categories; in the above example tag is an attribute, whose values are opening and closing. Elements can also contain other elements; however, contained elements must be wholly contained (beginning and ending tags properly nested, like concentric circles). There can be no overlapping of elements:

<sentence>This <verb>is</verb> a sentence.</sentence>

There are only a few other core rules to XML. Tags must be consistently capitalized (<name> and <Name> are different entities). Finally, XML must be well-formed: it must include a prolog (such as an XML declaration, for example <?xml version = "1.0"?>) and a root element "no part of which appears in the content of any other element" (XML 1.0) but which can, and usually does, surround all other content.

There may be one or more supplemental documents that define what elements, attributes and other entities may be used in a specific XML project, as well as where and how they may be placed. Such a document is called a DTD (Document Type Definition), or else a schema. If such a document is written and referenced in an XML file's prolog, the additional rules described in it apply to that XML file. Conformance to these rules makes the file valid. It is possible for an XML file to be well-formed but not valid if its general syntax is correct but it breaks the rules specified in the DTD or schema.
back to top

Appendix F. Semantic Code Structure and Interface

Semantic code structure

All excerpt texts and metadata, with the exception of image, video and/or sound files (stored separately), are contained in the file titled rolandht.xml. The basic encoding structure of the file (excepting information within an excerpt, for which see next section) follows; for information about XML see Appendix E.
  <work date="" geo="" lang="" name="" type="" timeperiod="">
      <author id=""></author>
      <title id="">
Song of Roland</title>
      <textnotes caption=""></textnotes>

    <excerpt id="" title="">
(see below)</txt>
[more excerpts, if and as needed]
[more works, as needed]
In conventional English, the above can be read as follows. This is a collection of excerpts from different works. Within each work are recorded the date of its creation (if known), its geographic origin, original language, name, type (prose, verse, drama), and time period (necessary for performing the statistical analysis presented in Appendix D). The author's name (if known), title and translator of the work (if any) are also noted. If general notes on the work are present, they are part of the header section. Each excerpt has a unique ID (necessary for processing for web presentation), a title, a short description of its context within the work, and the text of the excerpt itself. Each work contains one or more excerpts; the rolandht.xml document contains several works.

Semantic code structure within an excerpt

Besides the context and any structural encoding (paragraphs, line breaks for verse, etc.), an excerpt may contain the following elements60 and/or attributes:

<theme name="" who="" accused="" accuser="" charge="" realized="" simile="" metaphor="" tstart="" tend=""/>

<imagery name="" type="" realized="" magic="" called="" belongs="" whom="" simile="" metaphor="" tstart="" tend=""/>

<character name="" collective="" mention="" religion="" myth="" myth-origin="" historical="" tstart="" tend=""/>

<place name="" type="" where="" myth="" myth-origin=""/>

<speech who="" cont="" internal="" type="" according-to=""/>

<transl eng=""/>


A theme may have: a name that appears in the middle column of the website; a who designation, which attributes a theme (for example fear) to a specific character; a realized attribute, which specifies whether the action of the theme is performed or merely discussed. If the theme is accusation, the attributes accused, accuser and charge provide more information about the accusation. There are also attributes to categorize themes as similes or metaphors; and finally, tstart and tend refer to the time signatures between which the given theme occurs in a film clip. The four latter attributes may also be present in <imagery>; tstart and tend may also occur in <character>.

Imagery has a name (accessory, sound, nature or animal) and a type (weapon, hornblow, water, lion). An accessory may have a name of its own (be called Durendal, for example); belongs to someone, but may be in someone else's possession, which is noted in whom. An image may have magic properties, and may or may not be realized (see themes above).

A character has a name and perhaps a religion. The character may be collective (the Saxons); a myth (Saint Michael) – in which case it will have a myth-origin – or historical; and finally, the character may be merely mentioned but not present in the scene.

A place has a name (Paris) and a type (city); it may be a myth (in which case it may also have a myth-origin), and it may or may not be geographically contextualized using the where attribute.

Speech has a speaker (who); may be internal, such as a character talking to himself; and may have a type (for example lament). The speech of one character may be related by another; such cases are conveyed through the according-to attribute. Finally, the cont attribute is one of convenience: it permits me to designate a single speech instance spanning more than one structural element (such as a paragraph) without violating the XML hierarchy.

Translations from the Middle English, intended as reading aids, have only one attribute – eng for "English" – which contains the translation itself. The <transl> element surrounds the word or phrase being translated.

Finally, notes are in-line annotations (which appear as quill icons in the interface, see below) that draw attention to particularly interesting semantic connections within the corpus, reference translators' interpretations of obscure geographical locations, and/or provide further contextual information.

Web interface

The web interface is explained in the help file. The PDF file linked from the top of this page, however, includes information about the web interface.

back to top

Appendix G. Sample Teaching Modules

Module 1: Objective

Acquaint the student with critical and paratextual elements of RolandHT – information about the source, general text notes, translations, in-text notes and contextual information for each excerpt.

Module 1: Assignment

Load RolandHT in your browser. Find the excerpt titled "Forest [Battle at] Runcyvale." Answer the following queries, using the "Help" section if needed:

1. What text is this excerpt from? When was the text written?
2. Which character figures most prominently in the context for this excerpt?
3. Why was the Middle English left untranslated?
4. What do the following words mean: hende; y-slawe; ek; grethed; knyʒt?
5. Why is it remarkable that Roncesvalles (Runcyvale) is claimed to be a forest?

Module 2: Objective

Acquaint the student with a multilinear reading process guided by her own interests.

Module 2: Assignment

Load RolandHT in your browser. Do the following:

1. Read some excerpts – any from the long list on the left-hand side. Choose a theme or image that strikes you; click on that theme or image (listed in the middle column) to get a list of excerpts that contain it.
2. Record a path of at least five excerpts connected by the query term you selected; be sure to select excerpts from at least two or three different works. Note the title, author and date of the work's composition in addition to the excerpt title.
3. Examine closely the words or phrases highlighted in your chosen excerpts when you mouse over the theme or image in the middle column.
4. Write a paragraph or two on what role you think the theme or image plays in the corpus. What might it reveal about Roland? What contributions does this element make to individual plot lines? How are these contributions similar and different among excerpts from different works?
5. What is the most surprising piece of new information you have received in the course of this exercise? Explain in one or two sentences.

back to top

60 The elements are presented as singletons here in order to save space, but most of the time they are not. There are no restrictions on the nesting order of the semantic elements within an excerpt.  [back]