|
Computer-Assisted Content Analysis -- Page
Germany and the United States Mirrored:
An Exploration of Computer-Assisted
Content Analysis
Robert L. Stevenson
Sabine Stiemerling
Antje Brockmann
Institut f r Kommunikationswissenschaft
Ludwig Maximilians-Universit t M nchen
Haoming Denis Wu
University of North Carolina at Chapel Hill
Contact:
Robert L. Stevenson
UNC-CH School of Journalism
CB 3365 Howell Hall
Chapel Hill NC 27599-3365
Voice: +1.919.962.4082
Fax: +1.919.962.0620
[log in to unmask]
Submitted to the Communication Theory and Methodology Division, Association for
Education in Journalism and Mass Communication, meeting in Baltimore, August,
1998
Abstract
Germany and the United States Mirrored: An
Exploration of Computer-Assisted Content Analysis
Robert L. Stevenson
Sabine Stiemerling
Antje Brockmann
Institut f r Kommunikationswissenschaft
Ludwig Maximilians-Universit t M nchen
Haoming Denis Wu
University of North Carolina at Chapel Hill
Content analysis, used in one-quarter to one-third of recently published
communication research articles, has been influenced, albeit modestly to date,
by rapid growth in availability of electronic news archives and simple computer
programs that expand the range of material examined and, to some degree, emulate
traditional coding techniques. This study addressed the question of the image
of the United States and Germany as reflected in two major daily newspapers, the
New York Times and S ddeutsche Zeitung of Munich, as a case study of
computer-assisted content analysis. Capabilities and limitations of the method
are discussed, and several widely available programs are described.
Germany and the United States Mirrored:
An Exploration of Computer-Assisted Content Analysis
Introduction
Content analysis remains a staple of communication research on both sides
of the Atlantic. In 1995-96, Journalism and Mass Communication Quarterly
contained 40 (out of 128, 31%) research articles based on content analysis as
methodology, while Publizistik, the leading German communications research
journal, published seven (out of 31, 23%) content analysis-based articles during
the same period. At least two new books are devoted to the method.[1] A
methodology that occupies a quarter to a third of the published literature must
be taken seriously even though, more than other standard social science
methodologies, content analysis is difficult.
A major difficulty with content analysis, aside from the tedium and
drudgery inherent in the method, is interpretation of results. On one hand, we
want to link media content to something else, usually to some characteristic of
the sender or to some effect on the receiver. We want to explain variance in the
content as a function of the source or to use variance in the content to account
for effects on the receiver. However, linkages between the message and either
source or receiver are tenuous at best; the traditional conservative academic
position is that inferences of cause and effect cannot be drawn from content
alone.
On the other hand, if we use content analysis descriptively, we lack a
basis for explaining results. Numbers by themselves are neither big or small,
good nor bad. Comparisons of numbers are interesting but rarely provide a
satisfying basis for interpretation. Content analysis does not tell us how much
foreign news should be included in the front page mix or how much violence on
primetime television is a lot.
Nevertheless, as numbers from the two journals suggest, interest in content
analysis remains high, and new data sources and new analytical tools are
changing the method itself and opening up new possibilities for developing and
testing theories. In this paper, we consider how new technologies have
influenced the traditional use of content analysis and why new data resources
and new analytical toolsDeven if most are of the "dumb clerk"[2] varietyDare
important. As a substantive issue, we address a familiar but important question
of national images as they are reflected in leading newspapers of Germany and
the United States. The two nations as mirrored in the media of the other
represent a case study in the use of computer-assisted content analysis.
Background
Methodological Issues. Content analysis, on the whole, has lagged behind
recent changes in the mass media, which typically are the raw materials for
studies utilizing the method. The biggest change in the media is the explosive
growth of current and archived media content in electronic form. Via the
internet itself, a handful of commercial services such as Nexis, or a
fast-growing collection of CDs in almost every library, the researcher has
instant access to literally hundreds or thousands of media sources, many of
which now go back ten to 20 years. While some of the traditional bibliographic
and summary indices have been transferred to useful electronic format, the
revolutionary change is searchable full-text.
Consider, for example, a typical study of the "image of Germany" in U.S.
media. Not long ago, one might have begun with the printed New York Times Index
or Guide to Current Periodicals to identify material relevant to major topics,
such as unification or neo-Nazi violence. Coding would proceed once the stories
were identified. Or one could select a sample of major U.S. newspapers and from
them construct a week or set of weeks over a reasonable number of years. One
would then sift through the sample for stories about Germany and project
findings from the narrow study to U.S. news coverage in general. The sample was
always limited in range of media and in time.
Now, in contrast, a few keystrokes can identify every reference to German
or Germany in ten or more years of major publications or include several hundred
daily newspapers or even more magazines and news services. If the material is
then subjected to traditional content analysis, the computer has at least served
to find the appropriate needles in very large haystacksDan important practical
contribution to research in itself. In addition, the depth of traditional
content analysis can be sacrificed to take advantage of the volume and breadth
of computer-identified data. This opens the door to new analytical approaches,
such as time-series analysis, and to use of the conceptually simple mechanical
routines or search programs to approximate traditional coding on large data
bases. The latter application is of interest here where we want to use the
computer to help us answer simple questions about frequency of coverage and
general description of content.
Substantive Issues. The lack of international news in U.S. media is a
familiar complaint and part of a broader criticism of Western media, which
continue to dominate global news flow.[3] The specific situation of U.S.
coverage of Germany and German coverage of the United States contains elements
of broader issues and elements unique to the special relationship between the
two allies.
The broader issue is the general complaint, originally made by Third World
countries against Western media as part of the debate over the New World
Information and Communication Order (NWICO) in the 1970s, of a quantitative
imbalance of coverage and emphasis on disruption, often represented by the
catchphrase "coups and earthquakes."[4] The original NWICO complaints grew out
of Western coverage (or lack of coverage) of the Third World, but they have
their counterpart in the images of Germany and the United States in the media of
the other.
Few Germans would recognize their own country from the fleeting
imagesDusually confined to the few surviving old Nazis or their young would-be
followersDthat intermittently flash across American TV screens. The problem
includes both news and entertainment. A German writer, newly resident in
Washington in 1996, complained of the image of the country that came through the
entertainment side of the medium:
Turn on your TV set on almost any evening in America, and on one of the many
cable channels you will likely find a German. He is blond; sometimes, he is
incredibly handsome. But he has these cold blue eyes, wears a brown or black
uniform, clicks his heels and shouts: 'Right away, Herr Kommandant!'[5]
On the other hand, unlike the United States where the image of Germany
seems to reflect old movies and the occasional old newsreel more than today's
prosperous democracy, the German picture of the United States is a product of
massive regular news coverage as well as pervasive popular culture. From an
American perspective, German coverage of the United States seems obsessed with
idiosyncrasies and catastrophes, of which crime and violence are usually a key
element. The complaints about two sophisticated Western media systems reporting
two allies echo the Third World complaints in the NWICO debate.
The specific complaint is the gross imbalance between the minimal coverage
of Germany in the U.S. news media and almost obsessive attention to the United
States in German media. There is no shortage of coverage of the United States
in German media; if anything, attention to the minutiae of American affairs
occupies more time and space in German media than seems justified by the usual
news values. On the other side, coverage of Germany in U.S. mediaDexcept for
handful of major newspapersDis fleeting and inconsistent.
In a large-scale, cooperative study of foreign and international news in
40-plus countries in late 1995, the United States emerged as a unique global
news superpower in virtually all countries. While regional and national
self-interest remained dominant news values universally, news about the United
States towered above all other individual countries. The two-week analysis of
five German dailies (four serious, plus the tabloid Bild) and four TV networks
showed that the United States was represented in 21.6% of all foreign or
international stories. In contrast, in the American media for the same period
(New York Times, plus a regional paper, ABC evening news and NPR's "Morning
Edition"), Germany was mentioned in only 1.2% of the foreign or international
stories and did not appear once in the ABC coverage.[6] A study based on all
three networks' coverage over the crucial three years that included the opening
of the Berlin Wall, 1988-1990, found a total of 1,191 stories running almost
exactly 40 hours.[7] This produces an average of about 2.5 stories per week per
network during one of the most dramatic and important episodes in modern
history. Even during 1989, the 556 stories reduced to about 2.5 stories per
network per week. Like other countries, Germany emerged suddenly and briefly in
the limited U.S. media picture of the world, then disappeared, except for the
occasional violent episode, usually framed around Nazi images.
In both studies, the same problems arise. First, there is the question of
definitions. To what degree are results a function of various ad hoc decisions
and operationalizations?[8] Second, there is the common problem of
interpretation. It doesn't take a formal content analysis to see that German
media give more attention to the United States than vice versa, but there are
obvious and defensible explanations for the newsflow imbalance that range from
the difference in size of the two countries to the difference in relative
importance of each to the other. The largerDand more difficultDissue is
interpreting the numbers. The results may be predictable, but are the numbers
big or small? Too big or too small? And, of course, how can one explain them
or link them to some effect?
Keil blames TV coverage for general negative public opinion toward Germany
and Germans in the United States without providing evidence of either
anti-German public attitudes or any link to news coverage. The Berlin daily
paper, taz,[9] linked lack of coverage in American media to a lack of interest
in Germany among Washington's movers and shakers. Germany was clearly off the
radar screen in both newsrooms and the halls of government.
In contrast, Marc Fischer, Washington Post correspondent in Bonn and Berlin
during unification, reported that his contacts in the Foreign Ministry in Bonn
seemed to know immediately the details of his dispatchesDand to suspect him of
an obsession with reporting racial violence and hints of lingering
anti-Semitism.[10] The image of Germany in the U.S. media, however fleeting and
distorted in the German view, is overwhelmed by German media attention to all
things American, from the latest Washington scandal and Hollywood rumor to
informed and generally sympathetic coverage of history and culture as well as
current politics and economics.
Without attempting to address the interesting questions of links between
coverage of the two countries and reporting style or public opinion, we can
illuminate part of the question that precedes any discussion of explanation and
effect: What is the image of each country in the media of the other? AndDmore
to the point of this studyDcan we illuminate the question more thoroughly and
more efficiently by invoking relatively new electronic data sources and new
techniques of computer-assisted analysis?
Method
More than in other methods, results of content analysis are a function of
the usually pragmatic and often arbitrary decisions made along the way. A
typical study uses a relatively small availability sample of an arbitrary group
of media and a set of descriptive categories that are created to address
specific research questions. These decisions are often made without
consideration of their impact on results.
The Frankfurter Allgemeine Zeitung and New York Times are usually included
in a general analysis because they are assumed to influence other media as well
as to represent elite journalism, but Bild Zeitung, with ten times the
circulation, and a bland regional U.S. paper might, in some manner, be more
typical of the media in both countries than the elite papers. "The news" is a
difficult concept to define, let alone to sample randomly.
Data
News content in electronic form is now available from a number of sources.
Many newspapersDmostly the large and influential papers in many
countriesDpublish their own CD-ROMs with a built-in search program. As a rule,
content and software must be used together, but most software programs include
similar options, and search results can usually be exported for further
analysis.
The Nexis system in the United States, of course, is a unique resource with
several hundred daily papersDmostly American and virtually all in EnglishDbut a
handful of other large data bases are also available. The single largest source
of electronic news is now the Internet, and many web sites include an archive.
The biggest question with any electronic source is whether it is complete. CDs
usually indicate what parts of the paper, if any, are excluded, but not always.
CDs, while expensive to purchase, are probably the most reliable sources. Some
early electronic archives of smaller papers, in contrast, included only locally
written material. With Nexis and on-line archives, the question is often also
pertinent because print and electronic editions may differ, and problems of
completeness also arise.
Computer programs
Computers are good at finding needles in haystacksDnot an insignificant
skillDand at doing a few other simple tasks of the "dumb clerk" sort that can
have great practical value in projects such as the "national images" study
outlined earlier. If the key search termsD"German or Germany," "Amerika oder
USA oder Vereinigten Staaten"Dare relatively straight-forward and unambiguous,
the first sweep usually identifies the appropriate raw material with reasonable
efficiency. If the key concept is abstractDsymbols of war and peaceDor has
multiple definitionsDTurkey and ChinaDthe search may pick up a lot of chaff with
the wheat. Defining "the news" is no easier, but drawing a larger, broader
sample of it is a lot easier.
Some search programs (the Nexis software, for example) have a KWIC (key
word in context) provision that allows the researcher to identify a coding or
context unit. The stories or context units retrieved by the search can be run
through qualitative content analysis programs such as AskSam, CatPac, and
In-text.[11] The KWIC command can produce a specified number of words either
side of the key term, a line, sentence, or paragraph.
At this point, the data file typically consists of thousands of units,
rather than the hundreds in traditional content analyses. Often the researcher
must take overDto check for validity of the material, to weed out minor or
irrelevant hits, or to do traditional coding. In theory, it is possible to
download the material directly into one of several data base programs. EndLink
(an attachment to the EndNote bibliography software) and Nota Bene advertise
this capability. The logic is that the user establishes a data base file with
fields that match the format of the news archiveDBYLINE, DATELINE, LENGTH, DATE,
TEXT, etc. The software matches the two sets of fields to produce a data base
file that can be examined manually (but very efficiently). Traditional codes
can be added and the entire data set then exported to a statistical program.
In our experience, these programs seldom work as advertised. A major
problem is that news stories rarely contain the same set of fields. Some
stories have no datelines; others lack bylines; special fields such as PICTURE
are rarely used. An early version of the now-defunct Notebook software noted
wanly that in such cases the program did the best it could. Usually it is not
good enough. AskSam, however, is one of several programs that are good at
separating a long file of retrieved stories into individual stories and loading
them into the database program as separate documents.
At this point, the researcher should have a file with the key term(s)
within some specified context, such as story, sentence, or paragraph. One
direction is then to continue with traditional manual coding for appropriate
descriptive content factorsDvalence, general topics, or themes, or theoretically
derived attributes such as "frames." Sometimes, these same factors can be
inferred from output of simple content analysis programs.
Another direction that is useful if the data file is unusually large is to
let concordance-creation software go one step beyond finding the relevant
content and isolating it within an appropriate context. Simple
search-and-replace instructions can replicate traditional recoding. In this
way, variants of key wordsDGerman, Germans, Germany, and various hyphenated
formsDcan be reduced to a common root that can be searched and counted.
Results
Our test of computer-assisted content analysis to assess the image of
Germany and the United States in each other's media involves the New York Times
and S ddeutsche Zeitung (Munich). Both are serious and influential newspapers;
the content of both is available on CD-ROM. We used the year 1996 because it
was the latest year available at the time. The software that comes with the CD
is inextricably linked to the initial search. While similar, programs vary in
the details. For example, in the S ddeutsche CD, it was possible to limit the
search to the "Titelbereich," or headline and lead paragraph. That option is
available when searching the New York Times with Nexis but not with the CD. The
nearest approximation was to search the headline. The academic version of Nexis
is limited to printing (electronically or on paper) a single screen; it is
possible but tedious to call up each story in a search of the headline and lead
and save it to disk.
Stories retrieved from the S ddeutsche appear on the computer screen as
they appeared in the paper; New York Times stories are retrieved only in text
format and with additional cataloging fields added. Added fields typically
include topics, persons, and a short abstract of the article, which, if not
eliminated, can distort the results because they represent material not in the
original content. The original headline is retained, along with useful
information about position in the paper and prominence. In the Times, picture
cutlines are included; a separate field gives the number of words and
characterization of the story as short (roughly fewer than 100 words), medium,
or long (more than 700 words). S ddeutsche stories also contain a number of
additional fields, including length, section or special page of the paper, and
general subject categories, among others.
Frequency of coverage. How much coverage of the United States is in the
German media? And vice versa? As usual, simple questions have complex answers.
The short answer is, it depends. Is the image of a country built up from every
reference in the media, or from the major stories that only occasionally
dominate the front page? In this case, traditional indexes that recognize only
major stories are probably a reasonable guide, but it is important to remember
that any definition is arbitrary. And that results, as always, are a function
of the definitions.
Table 1 shows the various measures of frequency.
--------------------
Table 1 about here
--------------------
The relative imbalance between German coverage of the United States and
U.S. coverage of GermanyDas indicated by two major newspapers in one yearDis
clear, although we lack a criterion for deciding which of the various sets of
numbers is most appropriate.
Or even most valid. An obvious step is to screen out the false positives.
Because the two concepts are generally unambiguous, this is a minor problem (but
would be worse if we were comparing China and Turkey). The wildcard format[12]
in GermanDUS*--picks up references to Peter Ustinov, Usbekistan (German
spelling), and a few other words, but stories about Lateinamerika and S damerika
are screened out because of spelling conventions. Searching Times headlines for
"German or Germany" without a wildcard eliminates "Germantown" but also
"German-American" and other similar compound words. Searching for "German and
restaurant" and "German and automobile" was a guess that a few strictly domestic
stories might be included, but this appeared not to be the case. A story about
the prominence of Spanish as a foreign language in Bismark, North Dakota,
included enough of the city's founding and early development by German
immigrants that we decided to leave it in the data set.
As long as more or less equivalent criteria are usedDheadline only,
headline and lead, entire textDand due attention is given to the operational
definitions, probably any measure is reasonable. It would not be reasonable to
argue that the German paper really carried more than 14,000 stories about the
United States or that the Times coverage of Germany was really less than 300
stories in a year. As long as the same definition is used, the ratio of about
five to seven stories in the S ddeutsche to one in the Times is maintained and
seems to be a realistic finding that mirrors more traditional analysis.
National image. At this point, the computer-assisted content analyst has
several choices. One is to load the stories identified by the initial search
into some kind of database program so that each story becomes a separate record
or unit of analysis. A second is to use a KWIC procedure to identify a new unit
of analysis. The new unit typically would be a sentence that included the key
word, or the key word with a specified number of words on either side of it.
Both procedures continue computer-assisted analysis. A third choice, now that
the relevant material has been found, is to continue traditional analysis with
the researcher making various coding decisions. With luck, the new variables
can be added as fields to a database program so that "topic," "direction," etc.,
can be added to each story and then exported to a statistical program. Finally,
one can use the additional coding data included in the CD as a surrogate topic
card one finds in a traditional library subject index.
To limit the analysis to major stories, we did the following: a search of
the Times for "German or Germany" in the headline produced 293 stories. A
selection of "major" stories in the headline and lead paragraph of the
S ddeutsche coded on the CD during the first three months of the year produced a
sample of 541 stories.
Identifying major stories was a subjective decision, but it was relatively
easy to screen out stories in which the reference to the United States, even in
the headline or lead, was peripheral. For our exploratory purposes, the two
samples were considered roughly equivalent.
In both cases, a certain amount of manual editing was required to get the
data into a usable format. With a standard word-processing program, we
eliminated the various fields added to the electronic versionDsubjects, people,
abstractDthat could influence the computer-assisted analysis. Then we loaded
the data file into an AskSam program that put each story into a separate
document in the database program. Some of the other qualitative analysis and
text-searching programs include similar routines, but AskSam was available, is
relatively popular in academic research, and is easy to use. It is flexible in
searching for and counting words and character-strings. Since the news story is
the typical unit of analysis in media-based studies, we decided to concentrate
on various words contained in major stories about Germany and the United States
rather than sentences or phrases containing these terms.
Table 2 shows the results of AskSam searches for various key terms in the
two data files and separate use of the page/section field in the S ddeutsche to
identify various broad topics. To follow the bulk of previous national image
studies, we focused on sources of information, geographic elements, and key
topics.
__________________
Table 2 about here
As with the frequency measures, absolute results are strongly related to
the operational definitions. But they are remarkably consistent with studies
that have employed traditional methods and data sources, such as the foreign
news study cited earlier. News is largely politics and business and originates
from a handful of major cities. By themselves, the numbers give us some clues
for interpretation: the image of the United States emerges from Washington and
New York and is largely defined by politics and economics. If anything, the
image of Germany is broader but more fleeting and resonates with references to
past and present Nazis. In both cases, the most frequent and consistent frame
of the image is the exchange rate between dollar and mark and changes in the two
national stock exchanges.
A surprising finding is the heavy reliance on news agencies by these two
major international papers. To be sure, much of the agency-supplied material is
short or routine financial items. Even with correspondents in the other
country, their efforts wisely are put to pursuing material not available from
agency sources.
Are German foreign ministry concerns about American reporters' concern
(obsession?) with the Nazi past and right-extremist groups now justified?
Probably yes. However, the easy access to the full text of key stories allows
us to provide a qualitative explanation behind the 42 "Nazi" hits in the Times.
One aspect of the issue surfaced in March with an op-ed piece by Daniel J.
Goldhagen, author of Hitler's Willing Executioners, a controversial
reinterpretation of the Third Reich. The piece was followed by a long review of
the book itself and several letters to the editor. The Schneider piece at the
end of the year reflected the long-running nature of the issue. In contrast, a
search for key terms in the January-March S ddeutsche data and search in the
yearlong CD produced no reference at all to Goldhagen or the book. The handful
of hits for "Nazi" referred either to U.S. groups or were historical references.
Part of the explanation is the German preference for the term "right
extremism" to describe the groups operating in Germany today. However, the lack
of any reference throughout the year to Goldhagen and his book, which set off a
firestorm in Germany and became a bestseller in translation, is curious. With
the 1997 CD-ROM, we could trace the rise of the issue in Germany; with other
electronic sourcesDparticularly the influential newsweekly, Der SpiegelDwe could
probably demonstrate how the book and the controversy it generated moved from
U.S. media to agenda-setting German media, then to the general media and the
public. In the meantime, we are left with the puzzling nine-month silence in
one of Germany's most influential dailies on an issue that did resonate from one
side of the Atlantic to the other.
KWIC format. One way to test this selective search of key words and terms
is to compare results to an automated search of words physically close to the
key concept of German and Germany. For this exercise, we recoded Germany and
Germans to German, then used AskSam to find every occurrence of the key search
term in the New York Times sample and to list five words before and five words
after it. The result is a single document with the output of the search as a
continuous text file. In picking up the context words, the program does not
stop at periods or other end-of-sentence punctuation. As a result, some of the
cataloging and administrative information gets caught in the file.
Because AskSam concordance program lists the number of documents in which
each word occurs (and the KWIC file is a single document), we transferred the
KWIC file to another qualitative content analysis program, CatPac. CatPac also
has a default list of common excluded words, which is easy to expand. An
alternative is to create a list of words to be included.
The output of the KWIC search is shown in Table 3, after a first pass to
eliminate "German" itself plus a number of words related to indexing of the
Times file. The number of words included in the list can be specified.
--------------------
Table 3 about here
--------------------
The output can be used to produce a three-dimensional diagram of distances
among the words captured. Whether this automatically generated concordance of
terms physically proximate to the key term of Germany -- even after more pruning
of irrelevant terms -- is more useful than a search for specific terms that are
considered part of the image of the country is a matter for further
investigation and personal preference. The emphasis on business and economics
is apparent; the special concern for right-extremism and ghosts of World War II
does not emerge since the terms tend to be limited in frequency.
Conclusions
Content analysis is still a tedious business and vulnerable to arbitrary
decisions influenced by pragmatics of available resources and the unique
purposes of a study. Linking content to other aspects of the communication
process is still difficult. However, the method has changed. Raw resources are
enormously broader, and even intellectually trivial help from the computer can
be of significant practical consequence. Consider the ease with which we could
pursue questions about the Goldhagen book (and its author) and shed light on the
complaint about U.S. journalists' obsession with lingering effects of World War
II. Both explorations included a quantitative analysis combined with a
traditional qualitative analysis to add context and interpretation to the
numbers.
Simple measures such as purposeful frequencies can be useful, especially
when they are drawn from broad sources. Their value depends to a large degree
on the specificity of the search. If the word is unambiguous, finding and
counting it are relatively easy; knowing whether it occurs in one unit in ten or
one unit in a thousand is useful. The technique is crude, however; the joint
occurrence of "crime" and "United States" in the same sentence or within a
specified numbers of words from each other in a sentence provides no real clue
about how the two are linked in meaning. Computers will have to get a lot
smarter before they can generate clever ideas, produce insightful
interpretation, or replace the old-fashioned researcher.
Table 1. Frequency of stories about Germany and the United States in the New
York Times and the S ddeutsche Zeitung in 1996.
S ddeutsche Zeitung Total Head/Lead
Amerika 1515 84
USA 6809 846
Amerika* 7567 521
US* 11053 1811
"Vereinigten Staaten" 9 7
US*/Amerika*/"Vereinigten Staaten" 14864 2210
January-March 5411
New York Times Nexis Total Head/Lead
Germany 2913 925
German 2971 1155
German* 4683 1744
New York Times CD Total Head
German or Germans or Germany 4493 2931
*The Asterisk is a wild-card search format; the slash is the equivalent of the
Boolian OR.
1Used in further analysis.Table 2. Content characteristics of S ddeutsche
Zeitung coverage of United States and New York Times coverage of Germany in
1996.
Sources in S ddeutsche Zeitung:
Reuters 67
AP 47
DPA 71
Own Correspondent 42
S ddeutsche Zeitung 30
Bloomberg 7
Geographic references:
Washington 167
New York 69
Hollywood 10
Los Angeles 15
Miami 8
Boston 8
Chicago 13
Israel 10
Japan 20
Asia/Asian 10
Europ* 68
Deutsch* 81
Bonn 9
Berlin 8
Topic references:
Crime/murder/violence 40
Goldhagen 0
Kohl 0
Nazi* 5
Dollar 104
Clinton 82
Trade 61
Politic* 46
Business 130
Cultur* 14
Sport* 32
EU 11
NATO 8
OECD 0
UN 12
Sources in New York Times:
Reuters 51
AP 38
Bloomberg 46
UPI 1
DPA 1
Geographic references:
Bonn 54
Berlin 54
Munich 18
Hamburg 17
Cologne 13
Bavaria* 12
Topic references:
Kohl 30
Kinkel 7
Unemploy* 44
Inflation 25
Crime 12
Violence 9
Extrem* 10
*Nazi* 42
Goldhagen 9
Mark 54
Dollar 38
Wall Street 5
Business 90
Bank 67
America* 72
Europe* 106
NATO 19
EU or OECD 0
Table 3. CatPac concordance of KWIC output from New York Times ("Germany" plus
and minus five words)
TOTAL WORDS 1077 THRESHOLD 0.000
TOTAL UNIQUE WORDS 25 RESTORING FORCE 0.100
TOTAL EPISODES 1071 CYCLES 1
TOTAL LINES 4762 FUNCTION Sigmoid (-1 - +1)
CLAMPING Yes
DESCENDING FREQUENCY LIST ALPHABETICALLY SORTED LIST
CASE CASE CASE CASE
WORD FREQ PCNT FREQ PCNT WORD FREQ PCNT FREQ PCNT
--------------- ---- ---- ---- ---- --------------- ---- ---- ---- ----
EAST 125 11.6 409 38.2 AGAINST 41 3.8 245 22.9
RATES 61 5.7 292 27.3 BANK 42 3.9 226 21.1
INTERNATIONAL 57 5.3 336 31.4 BUSINESS 34 3.2 198 18.5
WEST 54 5.0 265 24.7 CENTRAL 32 3.0 192 17.9
CUT 44 4.1 233 21.8 COMPANY 31 2.9 173 16.2
EASTERN 43 4.0 214 20.0 CUT 44 4.1 233 21.8
FORMER 43 4.0 248 23.2 DOLLAR 32 3.0 198 18.5
BANK 42 3.9 226 21.1 EAST 125 11.6 409 38.2
GOVERNMENT 42 3.9 228 21.3 EASTERN 43 4.0 214 20.0
YESTERDAY 42 3.9 243 22.7 ECONOMIC 29 2.7 165 15.4
AGAINST 41 3.8 245 22.9 ECONOMY 39 3.6 215 20.1
MR 41 3.8 207 19.3 FORMER 43 4.0 248 23.2
TODAY 41 3.8 242 22.6 G 33 3.1 182 17.0
ECONOMY 39 3.6 215 20.1 GOVERNMENT 42 3.9 228 21.3
MARK 38 3.5 215 20.1 INTEREST 33 3.1 199 18.6
OFFICIALS 37 3.4 213 19.9 INTERNATIONAL 57 5.3 336 31.4
BUSINESS 34 3.2 198 18.5 MARK 38 3.5 215 20.1
G 33 3.1 182 17.0 MR 41 3.8 207 19.3
INTEREST 33 3.1 199 18.6 OFFICIALS 37 3.4 213 19.9
CENTRAL 32 3.0 192 17.9 OVER 32 3.0 186 17.4
DOLLAR 32 3.0 198 18.5 RATES 61 5.7 292 27.3
OVER 32 3.0 186 17.4 TODAY 41 3.8 242 22.6
COMPANY 31 2.9 173 16.2 WEST 54 5.0 265 24.7
YEAR 31 2.9 172 16.1 YEAR 31 2.9 172 16.1
ECONOMIC 29 2.7 165 15.4 YESTERDAY 42 3.9 243 22.7
[1] Mark D. West [Ed.], Computer Content Analysis: Theory, Methods,
Applications. Norwood NJ: Ablex Publishing Corp. (in press); Carl W. Roberts
[Ed.], Text Analysis for the Social Sciences. Mahwah NJ: Lawrence Erlbaum,
1996.
[2] Robert L. Stevenson, "In Praise of Dumb Clerks," in Mark D. West
[Ed.],Computer Content Analysis: Theory, Methods, Applications. Norwood NJ:
Ablex Publishing Corp. (in press).
[3] "The Decline of International News Coverage," a Freedom Forum Media
Studies Center panel discussion, reported at
<www.mediastudies.org/international/international.html>; Stephen Hess,
International News and Foreign Correspondents. Washington DC: Brookings
Institution, 1996.
[4] For a current interpretation, see Johan Galtung and Richard C.
Vincent, Global Glasnost; Toward a New World Information and Communication
Order? Creskill NJ: Hampton Press, 1992.
[5] Peter Schneider, "For Germans, Guilt Isn't Enough." New York Times,
Dec. 5, 1996, p. 35.
[6] Project website: http://sunsite.unc.edu/newsflow/. Data from an
early analysis.
[7] Hartmut Keil, Deutschlandberichterstattung in den Abendnachrichten der
Fernsehanstalten ABC, CBS und NBC 1988-1990 [Reporting on Germany in the Evening
News of the ABC, CBS and NBC Networks 1988-1990]. Research report for the
Herbert Quandt Foundation. Munich: Amerika-Institut der Universit_t M_nchen,
1992?. This study was carried out using an index/abstract published by the
Vanderbilt University TV news archive. The information is detailed, but the
index must be considered a secondary source.
[8] The newsflow study coded dateline, main country and up to two
additional countries; the TV study, of course, relied on the index.
[9] Reference to come.
[10] Marc Fisher, After the Wall: Germany, the Germans, and the Burden of
History. New York: Simon & Schuster, 1995.
[11] The best website of content analysis packages, including links to
several downloadable test versions, is maintained by a German sociologist,
Harald Klein: <www.soziologie.uni-jena.de/home/klein/textanal.html>. In-text
is his own program, available in German and English.
[12] Most search programs are similar. A xxx* format includes any string
of "xxx" and anything following: Amerika* -> Amerikaner, Amerikanerin,
Amerikanisch, etc. A x?x format usually allows any letter to be substituted for
the question mark. This can be useful in searching a data base that has both
American and British spellings.
|
