First Monday

Evaluation of Web access to historical sheet music collections and music-related iconography by Maurice B. Wheeler and Mary Jo Venetis

Abstract
Previous research within Music Information Retrieval (MIR) has examined audio and textual facets in attempts to retrieve information about the music itself, including humming melodies, encoding of audio for transmission, extracting bibliographic data as well as melodies and harmonies. An area lacking within MIR relates to the retrieval of images and illustrations that often accompany printed music. Addressing that deficiency, this paper will briefly discuss historical American sheet music and report results from research indicating whether researchers can retrieve sheet music imagery from digital music collections, using basic Internet search engines. The findings are expected to advance our understanding of the complexities of retrieving digital music collections and music–related iconography.

Contents

Introduction
Background
Research problem
Research design and methodology
Preliminary findings
Implications
Conclusion

 


 

Introduction

The visual imagery in the popular music of any culture captures not only the imagination of the creators, but also the social, political and cultural realities of the times. American popular music is no exception, and the popular music (songs) of past eras provides a unique opportunity to gain a greater understanding of the origins and development of many social issues in American society. Because ethnicity and race have historically been topics of political and social debate and controversy, the topics have remained a constant in American popular music for over two centuries.

Historical music sheet music and its accompanying illustrated cover art present a rich source of social commentary, providing researchers with the societal norms and values of a particular time in history. Highlighting the importance of access to digitized song collections and the historical social record captured by the cover sheet iconography, this paper places the discussion of imagery and music information retrieval (MIR) within the context of historical American popular sheet music.

This paper reports findings from a study that sought to determine if researchers can retrieve historical sheet music and accompanying iconography from digital music collections using basic Internet search engines. These findings are expected to advance our understanding of retrieving digitized music collections and music–related iconography.

 

++++++++++

Background

Because there was a significant increase in the inclusion of cover illustrations with human subjects on sheet music starting with the mid–1800s, popular songs of the mid–nineteenth and early twentieth centuries are uniquely suited for a visual examination of social attitudes and perceptions regarding race in post–slavery America (Wheeler, in press). Technology has made it possible for sheet music collections, including lyrics, manuscripts and illustrated cover art, to be digitized and available for easy access. Digital audio recordings are also available in many other formats including CDs, MIDI, and MP3 files.

Historical song collections and blackface minstrelsy

For the past few decades digitization projects in public and academic libraries have preserved many historical songs, making them available online to researchers. Some examples of these collections are the Library of Congress’ American Sheet Music Web sites [1], which includes a portion of Brown University’s collections, Harvard Theatre Collection, the Chopin Early Editions Collection at the University of Chicago [2], Duke University’s Special Collections [3], the Yale Music Library Collections [4], New York Public Library’s digital collections [5], and Michigan State University’s Library Africana Collection [6], among many other similar collections.

While a few of these American sheet music collections may be well known, or moderately well known, others are not, e.g., the Keffer Collection of Sheet Music housed at the University of Pennsylvania Library [7], the University of Michigan’s Corning Collection and Montgomery Collection of Popular American Sheet Music, the Irish Music Center at Boston College [8], the E. Azalia Hackley Collection at the Detroit Public Library [9], and the San Francisco Public Library’s Dorothy Starr Sheet Music Collection [10].

A large portion of the songs preserved within these American sheet music collections, from the mid–nineteenth and early twentieth century, are from the predominant blackface minstrelsy shows in America. Americans enjoyed the blackface minstrelsy show as their most popular form of entertainment. Unfortunately, its negative portrayal of African Americans continued throughout the twentieth century and into the twentieth–first century. According to Wheeler (in press), these images, combined with the song titles and lyrics, were "used to entice the consumer not only to make a purchase, but also seduced the willing and gullible into accepting the portrayals of African Americans as at least believable, if not authentic." Despite their distasteful and offensive depictions of African Americans, the images provide documentation and an undisputable record that allows for the exploration of explicit and implied meaning in the context of musical, textual, and visual composition. Ultimately, they become documentation on the development and dissemination of cultural knowledge.

Music iconography

Music iconography, whether illustrated art or moving images, also allows researchers to understand thoughts and ideals of individuals, as well as their social, cultural and political values by connecting the past to the present (DeTemple, 2001; Mitchell, 1986). Describing the visual arts as well as seeking meaning in the visual arts is not a new idea (Panofsky, 1974). Yet, it is a challenge to describe them by form, color, and pattern as literal representations (Markey, 1983; 1988). There are also underlying meanings or subjective/abstract descriptions (Shatford, 1986). Clearly, visual arts are difficult to describe, catalog and index, especially when an image can invoke literal and abstract meanings. While we suggest that music–related iconography should play an important role in MIR, it also poses a unique set of problems.

Referring back to traditional information retrieval, the term "aboutness" comes into play (Ingwersen, 1992). What is the image about? How does a person describe an image? Imagery can invoke literal meanings as well as abstract meanings as shown in O’Connor and O’Connor’s research (1999) when users were asked to provide verbal descriptions for specific photographs.

Along these same lines, an image can be "iconic, indexical and symbolic" depending on the context of the image itself [11]. As an example, Santini (2001) uses the image of a female which could represent womanhood, motherhood or a specific person.

Jorgensen (2003) reminds the reader that imagery has been part of human communication for centuries even before the development of the phonetic alphabet. Immediately, the thought of cave art and Egyptian hieroglyphics is invoked. These images represent an era long gone. An image can and does provide information, including the ideas and representation of an historical era. Weedman (2002) refers to images as "carriers of information" allowing researchers to use image collections in different ways.

Because of its literal and abstract meanings, iconography within image databases or collections is not easily retrievable as indicated in many studies within the past decade. Problems highlighted have related to access points, indexing, describing subject content, user queries and evaluation of image retrieval systems (Armitage and Enser, 1997; Brunskill and Jorgensen, 2002; Enser, 1995; Fidel, 1997; Jorgensen, 1998, 1999; Hastings, 1999).

Compounding the problem further, musical works are more difficult to classify because of their nature, e.g., multiple song titles and instruments. Classification authority control, however, has been established in the form of uniform titles and other metadata tags (Smiraglia, 1989). Nevertheless, visual components of musical works, such as illustrated cover art, have not been described adequately.

Because of advances in technology, visual information has become more prevalent, whether it is on a Web page or pictorial holdings within an online database (Choi and Rasmussen, 2003; Kherfi, et al., 2004). There are few search engines designed to locate imagery, and there is a need to access these images as a "source of information or for illustration" [12]. Therefore, illustrated cover art or music iconography becomes valuable to scholars in interdisciplinary fields, such as history and sociology, allowing them to retrieve music–related images relevant to their discipline.

Music information retrieval

Present research within MIR regarding retrieval of music information falls into several categories, for example, musical/audio and textual representations. The first category, musical or audio representation, involves retrieval of music by means of vocal methods such as humming melodies (Downie, 2004; Lippincott, 2002) and efforts to encode audio for transmission, storage and conversion to text (Foote, 1999; Goodrum and Rasmussen, 2000). The second category, textual representation, is based on traditional information retrieval techniques by locating bibliographic information, e.g., song title, composer, and publication information (Downie, 2003); content descriptors (Melucci and Orio, 2004); and exploring technologies to extract harmonies and melodies (Uitdenbogerd and Zobel, 2004), etc.

However, music manuscripts with cover illustrations, as well as video recordings of live performances and entertainment such as MTV videos, are not currently included within the visual information area of MIR.

 

++++++++++

Research problem

As mentioned previously, there are many well known digital music collections, such as the Library of Congress’ American Sheet Music Collection entitled "Music for the Nation." This particular collection contains digital reproductions of works published between 1820 and 1885. There are many other music collections that may not be as well known, such as the E. Azalia Hackley Collection which contains a digital collection of songs focused specifically on African American "themes." Because the Hackley Collection Web site is fairly new, it is not yet well known. Its existence has been passed verbally among colleagues.

A discussion of the various music collections led to the question: are there other historical sheet music collections of similar relevance which would be useful for scholars in other interdisciplinary fields? Further, could scholars, including those not trained in the music field, retrieve historical sheet music along with music iconography relevant to their own fields of study? Also, could novices locate a specific music–related image needed from digital music collections by using basic Internet search engines, without relying on personal knowledge of such collections? Along these same lines, still other questions remain. How would researchers rate the value of these digital music collections after accessing them through various Web sites?

The impetus of this study was the desire to identify iconography that depicts how African Americans were portrayed during the nineteenth and twentieth centuries in historical popular American sheet music. The premise of the study is based on the following research questions:

  1. What characteristics, such as background information, are needed to determine whether a digitized music collection is appropriate for inclusion as a research site, enabling scholars to access and analyze music iconography and gain insight regarding societal and cultural practices, norms and perceptions?
  2. Can researchers retrieve relevant digital music collections by using basic Internet search engines?
  3. By using controlled subject queries to retrieve relevant Web sites, how would basic Internet search engines rank these digital music collections?
  4. What attributes do these music collections have in common?
  5. Which subject queries would retrieve more hits within the keyword search or browsing links?
  6. How would researchers assess the value of the collection in order to determine whether the Web site is relevant?
  7. How often does the scholar need to revise the subject query to fit the historical context?

 

++++++++++

Research design and methodology

This section reviews the parameters of this exploratory study. Because the primary focus was to locate iconography that depicts the portrayal of African Americans during the nineteenth and twentieth centuries in historical popular American sheet music, the six digital music collections chosen for this study had the following attributes:

Six digital music collections

As noted, the Hackley Collection, because its focus on African American themes, was chosen as a "control" Web site along with five other music collections with attributes that would be matched for comparison. The E. Azalia Hackley Collection is part of the Detroit Public Library’s Music and Performing Arts Department in Detroit, Michigan. The collection’s "19th and 20th Century Songs on Negro themes" consists of over 600 works of sheet music published between 1799 and 1922.

Although Michigan State University Libraries has a renowned Africana Collection, its Ethnic Studies Collection, which includes 200 pieces of sheet music, was not available online for comparison.

The well known Library of Congress (LC) American Sheet Music Collection (commonly referred as "Music for the Nation") and its separate African–American Sheet Music 1850–1920 Collection, at http://lcweb2.loc.gov/ammem/award97/rpbhtml/aasmhome.html, were chosen as two separate Web sites to be evaluated for this study. The "Music for the Nation" collection is a more comprehensive collection which has minstrelsy music and related iconography. On the other hand, the African–American Sheet Music Collection focuses on African–American themes, which is similar to the Hackley Collection’s focus.

The African–American Sheet Music collection is a joint project between the LC and Brown University. The original sheet music collection is housed at the John Hay Library located at the Brown University in Rhode Island. The African–American Sheet Music collection contains 1,305 pieces of sheet music published from 1850 to 1920.

The larger music collection within the "Music for the Nation" site has 62,500 pieces of digital reproductions, covering the period of 1820 to 1860, and 1870 to 1885. The songs included were popular during that era, including secular choral, band and operatic music as well.

Two additional collections, Duke University’s Historic American Sheet Music Collection and the Lester S. Levy Collection of Sheet Music at the John Hopkins University [13], were chosen because of their focus on sheet music from the nineteenth and twentieth centuries, including minstrel music. Both music collections are also referred to in the external links within the Hackley Collection Web site.

The Historic American Sheet Music, which is referenced as the Duke Collection in this paper, provides digital access to 3,042 popular songs published between 1850 and 1920. The collection’s wide range of music includes minstrel, patriotic, plantation, sentimental, and other popular songs.

The Lester S. Levy Collection of Sheet Music includes 29,000 pieces of sheet music and bound monographs about popular music and its history. Like the Duke Collection, the music types are varied. The time period is more extensive, covering musical works from 1780 to 1960.

The last Web site, "Parlor Songs" at http://parlorsongs.com/, was chosen because of our awareness that it contained a significant number of minstrel songs. The site is not a library collection in a pure sense, in that there is no song database within this collection, nor is there an indexing to assist efforts in retrieving specific songs.

On the other hand, "Parlor Songs" provides access to popular songs not covered by the other music library collections included in the study. Similar to the original Levy Collection, this is a personal collection of songs owned by Rick Reublin who inherited the sheet music from his parents. The beauty of the illustrated covers led Reublin to digitize the collection, giving birth to the Parlor Songs Collection. As of January 2004, 1,300 works have been digitized, although the personal collection contains over 5,000 sheets. By choosing these six musical collections, a wide variety of results is expected within the Internet search engine results and rankings.

Three Internet search engines

Of particular interest was whether these six digital music collections could be retrieved by scholars, notably those not trained in the music field, by a basic Internet search engine.

A recent Internet study demonstrated the popularity of Google with a 35 percent share of all Web searches in the United States as of December 2004, with Yahoo having a 32 percent share, and MSN coming in third with 16 percent (Sullivan, 2005). Yahoo also owns AltaVista, one of the first Internet search engine to provide audio, image and video capabilities. These Internet search engines are well known not only to United States but also worldwide. For this reason, Google, Alta Vista and Yahoo were selected as the search engines for this exploratory study.

Procedures

Five controlled subject queries were developed based on the topic of "American historical sheet music" with the narrower topic of "African American sheet music." Also, because of the historical usage of terminology from the nineteenth and twentieth centuries, the following terms were included in the controlled subject queries: ‘minstrel,’ ‘minstrelsy,’ and ‘black minstrel.’

An Internet search engine comparison tool was used to search the same subject queries within three Internet search engines concurrently to minimize the error of typing the subject queries incorrectly. This was also done to ensure that the rankings were correct at that particular time.

Ranks.NL (http://ranks.nl/) was the comparison tool selected to run the subject queries because the three Internet search engines were listed in its menus (see http://www.ranks.nl/tools/compare.html). Ranks.NL also allowed ranking of the six digital collections by scanning the vertical columns and recording the number of results retrieved for each controlled subject query.

In theory, a successful Internet search will result in ten Web sites or hits on one page at a time, sorted by relevance. Further, according to Jansen, et al. (2000), 58 percent of Web surfers would not review the results beyond the 10 hits on the first page. For the purpose of the study, it was decided that the first three pages of a search would be reviewed to record the rankings for the first 30 results. This method allowed for greater breadth and depth of all possibilities in retrieving the six music collections.

When the first stage of searching was completed, the six digital music collections’ Web sites were analyzed for similar attributes across the board. A sampling of the features included the following attributes:

A chart listing the attributes was developed as a guideline to compare all six digital music collections. Again, the Hackley Collection was chosen as a control Web site for the other five music collections’ attributes to be matched for comparison. The chart was designed to elicit "yes" or "no" responses, with room for additional comments. The final stage of the study was a two–part analysis in retrieving relevant music imagery utilizing two methods.

The first method was a controlled subject query using a keyword search. A variety of subject queries were established to retrieve the number of results based on modern and historical usage of terms as cited in the LC’s Web site under "Search Help: Modern Usage vs. Historical Usage Comparison Table" (see http://memory.loc.gov/ammem/help/search_help.html).

We used "African American" as the first controlled subject query as this term is currently in greater use. However, in the nineteenth and twentieth centuries, the term "African American" was not used. Instead, the terms "coon" (a derogatory term) and "Negro" were commonly used at that time. The term "Negro" continued to be used until the 1960s, when it was replaced by the terms "black" and "black American," which were later replaced with "Afro–American." Likewise, the term of "African American," while not a new term, did not appear in common use until the 1980s, as discussed in the usage note provided by Dictionary.com (http://dictionary.reference.com/) on the word "black." The terms selected for use in the study show the range of common usage from the past to the present, even though some of the terms are now considered offensive or derogatory.

The second method was a browsing link supplied by each musical collection. The subject query would vary from site to site. The Duke Collection led a user to the subject query of "Legacies of racism and discrimination — Afro–Americans" and retrieved 262 hits. On the other hand, there is no such subject query of "Afro–Americans" within the Levy Collection. Yet, by using the browsing link for the term "minstrel," the same user retrieved 969 hits. Each subject query was listed and recorded with the number of results retrieved. Data was carefully gathered and analyzed for each research stage. Both the Hackley Collection and the LC’s African–American Sheet Music Collection, because of their thematic content, have broader subject headings which are pertinent to African Americans.

Data analysis — Rankings

For the first stage of the study Ranks.NL was used to analyze the rankings of the first controlled subject query, "American historical sheet music." Google recorded an approximate 1,040,000 hits, while AltaVista had 647,000 hits and 721,000 hits for Yahoo. The Hackley Collection and Parlor Songs were not retrieved by the first subject query. However, the LC’s African–American Sheet music was retrieved and ranked as number 7 under Google. Yet, the site was retrieved neither under AltaVista nor Yahoo. The Duke Collection and LC’s "Music for the Nation" were ranked in the top four in all three Internet search engines. The Levy Collection was listed as number 5 by two Internet search engines. The only exception was that the Levy Collection was not found by Google. The results are shown in Table 1.

 

Table 1: American Historical Sheet Music query results.
 
Google
AltaVista
Yahoo
American historical sheet music 1,040,000 647,000 721,000
Hackley
0
0
0
Duke
3, 4
1, 4
1, 4
Levy
0
5
5
LC Music
1, 2
2, 3
2, 3
LC African American
7
0
0
Parlor Songs
0
0
0

 

The second controlled subject query, "African American historical sheet music," provided a different result. This time, the Hackley Collection was retrieved as shown in Table 2, along with the LC’s African–American Collection, only after searching beyond the first 10 hits.

 

Table 2: African American Historical Sheet Music query results.
 
Google
AltaVista
Yahoo
African American historical sheet music 376,000 213,000 217,000
Hackley
17, 22
28
23
Duke
7, 24
10, 17
6, 13
Levy
0
0
0
LC Music
0
0
0
LC African American
1, 2
5, 7
1, 3
Parlor Songs
0
0
0

 

The Levy Collection was omitted from the rankings because its focus was not on African American historical sheet music, similar to the LC’s "Music for the Nation." Once again, the Parlor Songs Collection was also omitted from the rankings. Perhaps the Parlor Songs omission was because it is not a music collection in a pure sense and is not organized and indexed as other library collections.

The next two subject controlled queries of ‘minstrel’ and ‘minstrelsy’ retrieved none of these collections. This begs the question: do any of these collections have minstrel and minstrelsy songs and related iconography? The answer is yes, but these collections were not found by these subject queries. Instead of retrieving music collections, the various Web sites listed were, in general, related to minstrels and minstrelsy.

The final subject query was "black minstrel." Only the Hackley Collection was retrieved with this query, ranked as number 16 under AltaVista and number 17 under Yahoo. It does not appear in the Google Web page results.

In all, these Internet search engines indicate that they do not return the expected results. The retrieval results depend on two key factors; the formulated subject query by the user, and the choice of the Internet search engine. With such varying results, the challenge is even greater in determining the value of a Web site in the research process.

Data analysis — Similar attributes

It is not expected that all digital music collections would have the same attributes. Nevertheless, in order to provide scholars and researchers a method of evaluating the value of searches, levels of commonality such as the following should be considered:

All six digital Web sites allow the user to search by keyword and browsing features. Another common feature is describing the work itself, i.e., the song title, publication information, subject headings or description. This is commonly referred to as metadata or "bibliographical facet" (Downie, 2003). A song database is often part of the overall metadata scheme, which allows a user to retrieve a song by title, composer, or subject queries.

After analyzing the Web sites it was determined that all provided basic metadata, such as song title, publisher information, and composer. In examining the Parlor Songs Collection, a different method in classifying its metadata was utilized. Parlor Songs is the only site that did not provide direct access to a song database. Instead, Parlor Songs takes the user to featured issues and special motifs in a separate menu. Each feature issue showcases one to six songs on a specific theme, such as "Coon Songs" which was a special issue in April 2000. When a song is listed in a featured/special issue the user can view the image, where bibliographical data will also appear, but there are no subject headings. In its place, a lengthy description is provided for each song.

Through the use of essays, the Parlor Songs Collection provides a glimpse of perceptions and attitudes about race in the nineteenth and twentieth centuries, including how blackface minstrelsy evolved during that era. The other five Web sites also provide a social context for understanding the era in which the songs were composed. By guiding the user to the appropriate links, explanations were also found showing how the collection was developed, and materials chosen. Only a few Web sites included biographies and references to assist the user in conducting further research.

Another useful feature examined was the provision of external links to other music collections. Oddly enough, the LC’s "Music for the Nation" did not provide external links to outside sources, only to its own American Memory collections. Even the Parlor Songs, which is unconventional in its approach, provides external links to the Duke, Levy and LC collections. The Levy Collection, on the other hand, also had no external links to any other resources.

As revealed through the analysis of visual representation, all of the six Web sites provide both black and white and color images, including thumbnail–sized versions. All except the Parlor Songs Web site allows the viewer to enlarge the image and access the lyrics and/or musical composition. However, the Levy Collection does not always provide images for each song because of their interpretation of copyright law. If the song’s image is copyrighted before 1923, that image falls into public domain. After 1923, there is no image associated in the Web site with a listed song.

The leader in providing digital audio recordings for each song is the Parlor Songs Collection. The Web site asks the user to download a software application called "Sibelius Scorch" in order to listen to the audio recordings. The Hackley collection provides the user the ability to listen to only seven songs, while the LC’s "Music for the Nation" only provides some audio recordings from their "In Performance Choral Works from the Collection."

This completed the first stage of the study by analyzing the rankings of the historical sheet music collections using basic Internet search engines. The six digital music collections were also compared to ascertain whether they have similar attributes. Overall, they contain bibliographic data such as song title, description, imagery, and external links. In these collections, imagery are also used as the visual analysis of music, however, the descriptions of these illustrated cover art are lacking.

Several subject queries were used, indicating how indexing is used in these digital collections. Ironically, although all six collections have minstrel and minstrelsy songs, the results indicated no hits, which could lead a novice astray in researching musical iconography. This leads to the second stage of the study in analyzing how relevant music imagery could be retrieved by keyword and browsing methods.

Data analysis — Keyword and browsing searches

The second stage of the research was a two–part examination of the use of keyword and browsing searches. Each music collection retrieved hits by using eight keyword searches. Out of these eight keyword searches, three keywords were also used in combination with other words, for example, "racism" with "African–Americans." Although the keyword searches within the Parlor Songs Collection resulted in hits, they were discarded because each result did not take the user directly to the song.

The term "Afro–Americans" was considered a classified subject heading by LC, even though the prefix ‘Afro’ is not currently used by the majority of African Americans. The term "Afro–Americans" retrieved 2,054 results for five Web sites, a 0.9 percent difference from the 2,035 results from using the term "African Americans."

The Levy Collection had the highest number of records using the phrase, "African Americans" while the LC’s African–American Sheet Music site had the highest number of records using "Afro–Americans" as shown in Table 3.

 

Table 3: Keyword query results.
Keyword search
"African Americans"
"Black Americans"
"Afro–Americans"
"Negro"
"Coon"
Hackley
46
44
47
13
44
Duke
14
0
854
32
36
Levy
1,617
0
33
98
167
LC Music
367
0
401
12
86
LC African American
19
158
1,169
48
205
Total
2,035
202
2,054
203
538

 

On the other hand, the Levy Collection had the highest number of hits for the subject query of "Negro," resulting in 98 hits out of 203 records located. As for the usage of derogatory term "coon," 538 records from all five music collections were retrieved, with the LC African–American Sheet Music collection leading in the usage of this term, recording 205 hits.

In considering other terms, it is important to remember that several of the collections are already African American thematic and have a broader range of subject headings pertaining to race and African Americans. Under the subject query of "stereotype," whether by itself or used in conjunction with African Americans and Afro–Americans, only the Hackley collection retrieved 47 records.

In the Duke Collection, if the browsing link was used instead of a keyword search, the user would have retrieved 262 records within the browsing link called "Legacies of racism and discrimination — Stereotypes — Afro–Americans." The user could also retrieve 155 records with a similar subject query by locating a category called "HASM illustrations." HASM is the acronym for the Duke Collection’s official name, Historic American Sheet Music, and is used to indicate that the record has an illustration. Yet, by using the keyword search of "stereotype" by itself or with combined ethnic groups such as "African Americans," the user would not retrieve any relevant records. Apparently within the Duke Collection the word "stereotype" was not considered a keyword for searchers related to African Americans.

The Levy Collection also has a similar challenge in that the keyword search of "minstrel" retrieved only 143 records, far less than the collection actually contains. The viewer could locate seven boxes under "minstrel" which would retrieve 969 records. For this investigation, only every fourth record in the first box, Box 7, was sampled. All but one record in the first 40 records were songs and images related to African Americans. Within this collection, there are 38 topical categories.

By using the "Guide to the Lester S. Levy" menu (at http://levysheetmusic.mse.jhu.edu/levyguide/levycove.html), it leads the viewer to interactive subject headings. There, another browsing link within the Levy Collection is located. A subject query of "minstrels" resulted in 1,079 records. Unfortunately, the browsing link was not apparent until the "Guide" link was activated. Therefore, a researcher would have to use both methods, the keyword search and the browsing links, to retrieve all relevant songs and music iconography.

As shown in these results, which reinforce the ongoing studies in indexing, the question of whether to choose a keyword method over browsing search is not easily resolved.

 

++++++++++

Preliminary findings

Music iconography is an invaluable component within the field of MIR. Imagery allows scholars, in the social sciences and humanities, an opportunity to study and analyze the cultural and social norms of various historical periods in our national history.

This exploratory study’s findings, based on controlled subject queries to retrieve imagery within the six digital collections, have supported the ongoing concern regarding indexing and retrieving music iconography. The findings support the need to establish useful terms in describing images within musical collections, especially if they provide illustrated cover art.

Overall, the study provided a better understanding of digital music collections’ attributes when comparing six different Web sites. There are common attributes in these Web sites such as images, lyrics, and metadata. It should be noted that it cannot be said there are common searching methods as shown earlier in this paper.

There is no apparent strength in choosing one method, such as keyword searching, over another, such as the browsing method. The results for both methods match the ongoing issues with indexing and retrieving due to the formulation of subject queries in the music collections.

The findings discussed have advanced our understanding in retrieving connected music collections, notably in recovering relevant songs and music iconography. Results indicate that researchers’ personal knowledge plays an important role in selecting the appropriate music collection as shown in the Hackley and Parlor Songs Collections.

This study also provided information on the use of basic Internet search engines to indicate whether digital music collections, especially music imagery and illustrations, could be retrieved. The results confirm that search engines primarily "use text to look for images, without considering image content" [14].

Research questions with findings

Although the study was limited in its scope, by including music iconography as part of MIR’s system, it has provided significant findings through the series of research questions. Findings are discussed within the posed question below.

  1. What characteristics, such as background information, are needed to determine whether a digitized music collection is appropriate for inclusion as a research site, enabling scholars to access and analyze music iconography and gain insight regarding societal and cultural practices, norms and perceptions?
    As discussed previously, researchers should be able to retrieve the appropriate music imagery associated with their specific inquiry, for example, an on–going research project on African American music iconography and the historical construction of racial stereotypes. Each of the six digital music collections provides an insight to the historical, cultural, and social norms, notably the attitudes of white Americans towards African Americans in the nineteeth and twentieth centuries.
  2. Could researchers retrieve relevant digital music collections by using basic Internet search engines?
    The answer depends on which Internet search engine is used. With AltaVista and Yahoo, a scholar had a higher chance of retrieving relevant music collections. On the other hand, it also depended on what subject query was used. Still, other important and valuable resources were not retrieved because search engines are primarily designed to locate text rather than images or audio files. As an example, the Parlor Songs site is a unique digital music collection which did not appear anywhere in the Internet search engines’ rankings, yet this Web site is the only one that allows viewers to listen to audio recordings for each song listed in that collection.
  3. By using controlled subject queries to retrieve relevant Web sites, how would basic Internet search engines rank these digital music collections?
    As shown in Table 1, researchers could retrieve four out of six music collections using the query of "American historical sheet music." Only three music collections would be retrieved using the query of "African American historical sheet music." Yet, only one, the Hackley Collection, could be retrieved by the query of "black minstrel." Scholars will need to formulate subject queries which will fit both the specific topical area and the timeframe.
  4. What attributes do these music collections have in common?
    All six music collections have the following attributes in common: bibliographical facet (metadata) and visual facet (images, both black/white and color, thumbnails availability). In the audio representation, only Parlor Songs enable viewers to listen to the recordings for each song. The Hackley and LC’s "Music for the Nation" provides limited access to audio recordings.
  5. Which subject queries would retrieve more hits within the keyword search or browsing links?
    Results are based on retrieving records from both the keyword searches as well as browsing links. The researcher would need to become familiar with the music collection in order to maximize the full potential of that collection. This study’s results show that both keyword and browsing searches should be used by users to retrieve the maximum number of retrieved records.
  6. How would researchers assess the value of the collection to determine whether the Web site is relevant?
    Each music collection has strengths and weaknesses, and relevance depends on a specific scholar’s research needs. The Hackley collection and the LC’s African–American Sheet Music are the only two sites focusing on African American "themes," and should be considered differently from the other four Web sites. These two collections have a broader range of topics which will enable the researchers to retrieve relevant music iconography within topical areas instead of searching through hundreds of records. It would be the researcher’s judgment call in deciding which collection suits the specific needs of the research project. It is likely that the researcher would use more than one music collection. From this study, it appears that a scholar will also require some personal knowledge of lesser known music collections in order to do an effective search.
  7. How often does the scholar need to revise the subject query to fit the historical context?
    The key is to review the subject terminology within the appropriate time period before applying them to the search queries. The LC Web site alerts users that there are words that are used differently, such as "African American" versus "Afro–American," in modern and historical times.

 

++++++++++

Implications

Further research

Since the study was exploratory, it would be appropriate to expand it to include areas not yet fully investigated. To add to the value of the results surveying users to determine more of their specific needs and usage of the resource materials would be beneficial. It will be vital to obtain participants’ thoughts and observations through usability testing.

Moving beyond mere replication of the characteristics of ranking digital music collections in this study, future research could extend beyond the limitation of six specific music collections and three Internet search engines. It is essential to study the behavior of Internet search engines to determine the impact of specific indexing and retrieval systems. By determining their impact, all digitized music collections could be exposed to harvesters for Web crawling purposes. Also, the study of using additional Web mining is needed to determine how music collections can be accessed and studied for common attributes.

 

++++++++++

Conclusion

The value of popular songs of the nineteenth and twentieth centuries are significant in helping our understanding of the musical, social, and political times in which they were created. It is also clear that music–related imagery played an integral role in helping communicate the intent of the music and lyrics of a song.

This study highlights the significance of historical popular music and its accompanying iconography as a rich source of social commentary, reflecting societal norms and beliefs, opinions and perceptions of individuals at any given time in history. It also highlights the challenges that researchers face in attempting to access these resources and suggests the benefit of including music iconography or imagery as one of the access points within MIR.

Currently, digitized music iconography seems to be an ancillary benefit of accessing historical sheet music. The lack of a concerted effort in providing access to this important musicological and social science resource adds unnecessary steps to the research process, and potentially limits its effectiveness. It is recommended that visual components of musical works, including illustrated cover art and moving images, as in popular music videos, be considered for its transmission of cultural and societal values.

The research value and usage of digital musical collections will be greatly enhanced if additional descriptions are provided for all images associated with popular songs. It is highly recommended that renewed efforts should be made to include visual or image components in any future information of retrieval systems. End of article

 

About the authors

Dr. Maurice B. Wheeler is Associate Professor of Library and Information Sciences at the University of North Texas. He has earned graduate degrees from the University of Michigan in Music Performance and Library and Information Science, and a Ph.D. from the University of Pittsburgh. Wheeler began his career as a music librarian and archivist, and has held administrative positions in both public and academic libraries.
Direct comments to mwheeler [at] unt [dot] edu

Mary Jo Venetis is the manager of Acquisitions and Interlibrary Loan at Dallas Public Library, Dallas, Texas, and is currently a doctoral student in the School of Library and Information Sciences at the University of North Texas. Venetis holds a MLS degree from Texas Woman’s University and a BS degree in Journalism from Sam Houston State University. She is very active in Texas Library Association, serving on a wide variety of committees.
E–mail: mjvenetis [at] sbcglobal [dot] net

 

Notes

1. See, for example, "Music for the nation: American sheet music, 1870–1885," at http://memory.loc.gov/ammem/smhtml/smhome.html, accessed 23 September 2005.

2. "Chopin Early Editions," at http://chopin.lib.uchicago.edu/, accessed 23 September 2005.

3. "Sheet Music at Duke," at http://www.lib.duke.edu/music/fendshee.htm, accessed 23 September 2005.

4. "The Irving S. Gilmore Music Library, Yale University Library," at http://www.library.yale.edu/musiclib/muslib.htm, accessed 23 September 2005.

5. "NYPL Digital," at http://www.nypl.org/digital/index.htm, accessed 23 September 2005.

6. "Michigan State University, Library Africana Collection," at http://africa.msu.edu/africana.htm, accessed 23 September 2005.

7. "Keffer Collection of Sheet Music, ca. 1790–1895," at http://www.library.upenn.edu/collections/rbm/keffer/, accessed 23 September 2005.

8. "Irish Music Center, John J. Burns Library, Boston College," at http://www.bc.edu/libraries/centers/burns/services/irishmusic/, accessed 23 September 2005.

9. "The E. Azalia Hackley Collection," at http://www.thehackley.org/, accessed 23 September 2005.

10. "The Dorothy Starr Sheet Music Collection," at http://sflib1.sfpl.org:84/, accessed 23 September 2005.

11. Santini, 2001, p. 37.

12. Kherfi, et al., 2004, p. 36.

13. "The Lester S. Levy Collection of Sheet Music," at http://levysheetmusic.mse.jhu.edu/, accessed 23 September 2005.

14. Kherfi, et al., 2004, p. 36.

 

References

"African–American Sheet Music 1850–1920," The Library of Congress, American Memory and Brown University Library, at http://lcweb2.loc.gov/ammem/award97/rpbhtml/aasmhome.html, accessed 24 April 2005.

L.H. Armitage and Peter G.B. Enser, 1997. "Analysis of user need in image archives," Journal of Information Science, volume 23, number 4, pp. 287–299. http://dx.doi.org/10.1177/016555159702300403

Jeff Brunskill and Corinne Jorgensen, 2002. "Image attributes: A study of scientific diagrams," Proceedings of the American Society for Information Science and Technology, volume 39, number 1 (November), pp. 365–375. http://dx.doi.org/10.1002/meet.1450390140

Youngok Choi and Edie M. Rasmussen, 2003. "Searching for images: The analysis of users’ queries for image retrieval in American history," Journal of the American Society for Information Science and Technology, volume 54, number 6 (April), pp. 498–511. http://dx.doi.org/10.1002/asi.10237

Jill DeTemple, 2001. "Singing the Maine: The popular image of Cuba in sheet music of the Spanish–American War," The Historian, volume 63, number 4 (Summer), pp. 715–729. http://dx.doi.org/10.1111/j.1540-6563.2001.tb01942.x

J. Stephen Downie, 2004. "A sample of music information retrieval approaches," Journal of the American Society for Information Science and Technology, volume 55, number 12, (October), pp. 1033–1036. http://dx.doi.org/10.1002/asi.20054

J. Stephen Downie, 2003. "Music information retrieval." In: B. Cronin (editor). Annual Review of Information Science and Technology. Medford, N.J.: Information Today, pp. 295–340.

Peter G.B. Enser, 1995. "Pictorial information retrieval," Journal of Documentation, volume 51, number 2, pp. 126–170. http://dx.doi.org/10.1108/eb026946

Raya Fidel, 1997. "The image retrieval task: Implications for the design and evaluation of image databases," New Review of Hypermedia and Multimedia, volume 3, pp. 181–199. http://dx.doi.org/10.1080/13614569708914689

Jonathan Foote, 1999. "An overview of audio information retrieval," Multimedia Systems, volume 7, number 1 (January), pp. 2–11. http://dx.doi.org/10.1007/s005300050106

Abby Goodrum and Edie Rasmussen, 2000. "Sound and speech in information retrieval: An introduction," Bulletin of the American Society for Information Science and Technology, volume 26, number 5 (June/July), pp. 16–17. http://dx.doi.org/10.1002/bult.169

Samantha K. Hastings, 1999. "Evaluation of image retrieval systems: Role of user feedback," Library Trends, volume 48, number 2 (Fall), pp. 438–452.

"Historic American Sheet Music," The Rare Book, Manuscript and Special Collections Library, Duke University at http://scriptorium.lib.duke.edu/sheetmusic/, accessed 24 April 2005.

Peter Ingwersen, 1992. Information retrieval interaction. London: Taylor Graham.

Bernard J. Jansen, Amanda Spink, and Tefko Saracevic, 2000. "Real life, real users, and real needs: A study and analysis of user queries on the Web," Information Processing and Management, volume 36, number 2 (1 March), pp. 207–227.

Corinne Jorgensen, 2003. Image retrieval: Theory and research. Lanham, Md.: Scarecrow Press.

Corinne Jorgensen, 1999. "Access to pictorial material: A review of current research and future prospects," Computers and the Humanities, volume 33, number 4 (December), pp. 293–318. http://dx.doi.org/10.1023/A:1002065412222

Corinne Jorgensen, 1998. "Attributes of images in describing tasks," Information Processing and Management, volume 34, numbers 2/3 (March–May), pp. 161–174.

M.L. Kherfi, D. Ziou, and A. Bernardi, 2004. "Images retrieval from the World Wide Web: Issues, techniques, and systems," ACM Computing Surveys, volume 36, number 1 (March), pp. 35–67. http://dx.doi.org/10.1145/1013208.1013210

"The Lester S. Levy Collection of Sheet Music," Special Collections, Milton S. Eisenhower Library, The Johns Hopkins University, at http://levysheetmusic.mse.jhu.edu/, accessed 24 April 2005.

Aura Lippincott, 2002. "Issues in content–based music information retrieval," Journal of Information Science, volume 28, number 2, pp. 37–142. http://dx.doi.org/10.1177/016555150202800205

Karen Markey. 1988. "Access to iconographical research collections," Library Trends, volume 37, number 2 (Fall), pp. 154–174.

Karen Markey, 1983. "Computer–assisted construction of a thematic catalog of primary and secondary subject matter," Visual Resources, volume 3, pp. 16–49. http://dx.doi.org/10.1080/01973762.1983.9659063

Massimo Melucci and Nicola Orio, 2004. "Combining melody processing and information retrieval techniques: Methodology, evaluation, and system implementation," Journal of the American Society for Information Science and Technology, volume 55, number 12 (October), pp. 1058–1066. http://dx.doi.org/10.1002/asi.20058

W.J.T. Mitchell, 1986. Iconology: Image, text, ideology. Chicago: University of Chicago Press.

"Music for the Nation — American Sheet Music," Music Division, The Library of Congress, American Memory, at http://memory.loc.gov/ammem/mussmhtml/mussmhome.html, accessed 24 April 2005.

"Nineteenth and Twentieth Century Sheet Music of Negro Themes," The E. Azalia Hackley Collection, The Detroit Public Library, at http://www.thehackley.org/, accessed 24 April 2005.

Brian O’Connor and Mary K. O’Connor, 1999. "Categories, photographs and predicaments: Exploratory research on representing pictures for access," Bulletin of the American Society for Information Science, volume 25, number 6 (August/September), pp. 17–20. http://dx.doi.org/10.1002/bult.132

Erwin Panofsky, 1974. Meaning in the visual arts: Papers in and on art history. Woodstock N.Y.: Overlook Press.

"Parlor Songs 1800s–1920s, In Search of American Popular Song," at http://parlorsongs.com/, accessed 24 April 2005.

Simone Santini, 2001. Exploratory image databases: Content–based retrieval. San Diego, Calif.: Academic Press.

"Search Help: Modern Usage and Historical Usage," The Library of Congress, American Memory, at http://memory.loc.gov/ammem/help/search_help.html#p6, accessed 25 March 2005.

Sara Shatford, 1986. "Analyzing the Subject of a Picture: A Theoretical Approach," Cataloging and Classification Quarterly, volume 6, number 3, pp. 39–62. http://dx.doi.org/10.1300/J104v06n03_04

Richard P. Smiraglia, 1989. "Uniform titles for music: An exercise in collocating works," Cataloging and Classification Quarterly, volume 9, number 3, pp. 97–114. http://dx.doi.org/10.1300/J104v09n03_06

Danny Sullivan, 2005. "comScore Media Metrix search engine ratings," Searchenginewatch.com (11 February), at http://searchenginewatch.com/reports/article.php/2156431, accessed 26 March 2005.

Judith Weedman, 2002. "Thinking with images: An exploration into information retrieval and knowledge generation." Proceedings of the American Society for Information Science and Technology, volume 39, number 1 (November), pp. 376–382. http://dx.doi.org/10.1002/meet.1450390141

Alexandra L. Uitdenbogerd and Justin Zobel, 2004. "An architecture for effective music information retrieval," Journal of the American Society for Information Science and Technology, volume 55, number 12 (October), pp. 1053–1057. http://dx.doi.org/10.1002/asi.20057

Maurice B. Wheeler, in press. "Popular music imagery as an information system: Stereotypical depictions of African Americans in historical sheet music," International Journal of the Humanities.


Editorial history

Paper received 14 August 2005; accepted 18 September 2005.


Copyright ©2005, First Monday

Copyright ©2005, Maurice B. Wheeler and Mary Jo Venetis

Evaluation of Web access to historical sheet music collections and music–related iconography by Maurice B. Wheeler and Mary Jo Venetis
First Monday, Volume 10, Number 10 - 3 October 2005
https://firstmonday.org/ojs/index.php/fm/article/download/1283/1203