Professional Documents
Culture Documents
ICHIM2007 - Tate Online
ICHIM2007 - Tate Online
Abstract
As the Internet extends its global reach, language can remain a barrier
preventing people from being able to fully explore material of interest.
Tate Online, the Web site for Britain’s Tate art galleries, serves as a
good case study for exploring the issue of meeting international site
visitors’ multilingual needs. The site contains a great deal of material
that is of international interest; however, much of this is currently
accessible only in English. The present study used a variety of methods
to gather a set of requirements and recommendations for providing
enhanced multilingual content on Tate Online. These included a
competitor analysis, on-line survey of 457 Tate Online visitors, log file
analysis, machine translation resource evaluation, and basic user test.
Findings from this preliminary study provided information about users’
main activities on Tate Online. They indicate that many individuals
would appreciate having more content available in their own language,
either due to necessity or out of preference. However, the best means of
providing this content depends on a variety of factors, including the
pragmatic consideration of resources available for translation. Insights
gathered here can also apply to other cultural heritage organizations
looking to expand the amount of multilingual material on their own Web
sites. The means by which this is accomplished may involve striking a
balance between that which fulfills site user needs and that which is
feasible for the organisation to implement.
Keywords: cultural heritage Web sites, multilingualism, localisation, user
needs, translation, Tate Online
Introduction
Many international organisations with a web presence recognise the need to offer
multilingual content in order to reach a wider and more diverse audience. For cultural
heritage Web sites in particular, “the audience for the material goes beyond linguistic
and national boundaries,” so the greater the audience reached, the greater the value of
The World Wide Web is a useful tool for the dissemination of cultural heritage-
related material because it frees users from the constraints of physical location, allowing
them to access information from all over the world. However, associated with this wide
range of sources are language barriers which may prevent people from being able to
To help address this issue, the MultiMatch project aims to create a vertical search
engine focused on the cultural heritage domain that will, among other things, provide
automatic query translation and cross-language retrieval tools to help users search for
and access content across language boundaries. However, little is currently known
about the multilingual abilities and needs of individuals who are likely to access cultural
Tate Online provides an excellent case study for expanding knowledge on this
topic because, as the official Web site for the UK’s Tate art galleries, it is a popular and
heavily-visited on-line resource with global appeal. However, most of the site’s content
is only available in English, thus potentially excluding part of its would-be audience.
2
The aim of this study was to gather a set of requirements for increasing
multilingual access to Tate Online. Expanding the multilingual content available on this
(or any cultural heritage organisation’s) Web site has the potential to benefit both site
visitors and the organisation itself. By considering both of these parties’ needs and
desires, an effective and mutually acceptable strategy for enhancing multilingual access
may be developed.
person from any location at any time, in reality the number of Internet users (and the
languages spoken by these users) is not equally distributed across the globe. This is a
continuously evolving area: for example, in 2000, 51% of all Internet users spoke
English as a first language (Graddol, 2006), but by 2007 this figure had decreased to
29% (http://www.internetworldstats.com/stats7.htm).
research indicates that they will still prefer to operate within their native language if the
users prefer to be addressed in their native language (Eurescom Project, 2000), either
survey found that 97% of cultural Web sites in the UK were monolingual English sites
(Minerva Project, 2006). Therefore, any organisation that offers multilingual content
3
Translation Tools
Making multilingual content available on Tate Online will inherently involve
translation. However, the nature of the content to be translated will determine the
approach taken. There are a variety of tools and resources that the Tate could use for its
translation needs.
responsible for translating permanent content that will appear on the top-level pages
detract from the overall prestige and image of a site (Guyon, 2003). However, machine
translation can be employed in instances where it is considered acceptable for the reader
to get the main idea of the text, or for frequently-changing and updated content.
retrieve information in a language other than that which is used to conduct the search. It
information retrieval techniques (Clough & Sanderson, 2006), and there are a variety of
4
Tate Online is designed to entice people to physically visit the galleries but it is
also becoming an on-line destination in its own right. In 2005, more people visited Tate
Online than visited all four of the offline galleries combined (Rellie, 2006).
Two primary goals for UK museums on-line in 2006 were increasing audiences
and improving access (ibid). One important way of achieving this is to make Web site
content available in many languages. The benefits of doing so would extend to the Web
site visitor but also to the Tate, as “the localization of products and services based on an
In its present state, Tate Online offers basic information about planning a visit to
each of its offline galleries in 12 different languages. This information includes opening
hours, directions to the galleries, and other details that would be useful for prospective
visitors to know.
As the Tate aims to increase the scope of its multilingual content, the Collection
section of the site has been identified by Tate staff as the area of greatest interest for
translation. It is composed of various subsections: those which are most often used
include Artist A-Z (an alphabetical listing of all artists in the Collection along with
biographical information) and an illustrated glossary of art terms. Not only does it
represent the core of the galleries (the artworks), but it receives roughly one-third of all
site traffic.
5
Analysis of Multilingual Functionality of Tate Online and
Similar Sites
Tate Online is unique in that it is the only site dedicated solely to providing
information about the Tate galleries and their collections. However, there are similar
sites that can be classified into three groups: other museums and galleries in London,
other popular modern art museums worldwide, and general art and culture-related sites.
Appendix A.
localisation approaches taken by the various sites. Tate Online is similar to many of its
offered. However, the question is whether it can or should expand upon this basic
languages should be identified as priorities, and these will most likely relate to the
the audience that stands to benefit from increased multilingual content, it was therefore
logical to consider their needs and involve them in the planning process. As a result, a
user-centred design approach was employed, which involved consulting actual site
6
visitors and using their input to evaluate a series of prototypes (Petrelli et al., 2006).
This approach can help create a design for an efficient, effective multilingual Web site
Different data collection techniques were used at each stage to gather both
advantageous because it enables one to build a holistic view of the myriad issues at
hand. Each technique yields information from a certain perspective which can help to
Therefore, the first step in the process was to assess user and usage needs or
questionnaire. Tate log file analysis also helped to give an idea of the types of basic
tasks and searches carried out on Tate Online. An evaluation of machine translation
ideas, following which a basic usability study was conducted with native and non-native
English speakers. This was all that was possible to accomplish in the time available;
however, results and observations from all stages influenced and were incorporated into
to Tate Online, it was important to consult actual end-users of the Web site. Areas of
7
• User characteristics (where are international visitors located and what
However, it must also be noted that the end-users are not the only ones
concept of ‘user’ should be defined to include everybody whose work is affected by the
product in some way” (Nielsen, 1993). Therefore, the Tate’s needs should also be taken
• Demand and feasibility (is there really a need for more multilingual
To provide insight into these and other questions, a questionnaire was designed
to gather information regarding languages spoken by Tate Online users and to assess
their preferences in order to prioritise the aspects and areas of the Collection site that
Questionnaire results
A total of 457 people responded to the questionnaire, which was made available
through two links on the Tate Web site. Based on IP address, 31.2% of respondents
Language skills
8
When users were asked which language they preferred to use when searching
and browsing the Internet, the top ten results were as follows:
Language Percent
Spanish 34.4
English 26.4
Italian 18.2
French 13.1
Portuguese 2.7
Dutch 1.3
German 0.9
Japanese 0.7
Turkish 0.4
Welsh 0.4
given that these were the languages in which the survey was available. Regarding self-
reported secondary language skills, the percentages of people who said they could
9
English Reading 39.6
Writing 33.7
Writing 17.9
Writing 9.2
Writing 8.5
English language skills is not surprising, given that Tate Online’s content is
predominantly in English. What is unknown is the number of people who would like to
visit Tate Online but have been hindered from doing so by a lack of English ability.
53% of respondents had visited the Collection site before. Of these, 69% had
done so out of personal interest, while 27% were motivated by academic or research-
related objectives.
Regarding primary objectives for using the Collection site, roughly 60% of
respondents used the site to browse the Tate’s holdings without looking for something
specific, or to prepare for a visit. The relative frequencies of all tasks were:
10
Task Percentage
Other 4.3
Respondents whose preferred language was not English rated the amount of
Response Percentage
11
Overall, 76.4% of those who did not prefer to view Web sites in English stated that they
would be more likely to visit the Collection site if it were translated into their preferred
language.
Area Percentage
Artist Biographies
35.9
General instructions
(i.e., how to use the 22.4
subject search)
Glossary 18.1
Artist names
7.6
Most of the above areas (besides “Search” are static content and thus could be translated
Other revealing attitudes were expressed in the section for general comments and
suggestions. In general, people praised the English version of Tate Online, while
indicating they would appreciate the ability to view the pages in different languages.
12
Justification for Translation
The results of the initial survey suggest that providing more multilingual content
offering enhanced multilingual content would attract more visitors to the Collection site
reinforced; however, the specifics of how and what to translate must be ascertained in
The high degree to which the Collection site is used out of personal interest
satisfied with machine translations that allow them to get the gist of the information.
However, one issue to consider regarding reported use of the Collection site is
the time period in which the data was collected (during school summer holidays.) It is
possible that this affected the results, as presumably fewer people would be using the
site for academic purposes. It is possible that machine translation may not be acceptable
instructions, and search,) the means of translating should be considered. Given the vast
number of biographies (approximately 3,000) and the fact that the pages are all
13
One possible compromise could be to provide professional translations for biographies
of the most popular artists and then to provide a means for automatic translation of the
rest.
Another way to obtain information about what visitors to Tate Online do whilst
visiting the site is by examining the server log files. An analysis of log files for one 24-
hour period was able to yield some basic data regarding the types of search queries
entered and the aspects of the Collection site which were the most visited. Although
more in-depth and long-term analysis of log files would be necessary to provide a more
accurate picture of overall use patterns, this 24-hour sample gave some insight into what
Using the log file, it was also possible to extract search queries entered into the
Tate Online main page site search. Although these were not specifically from the
Collection site, they give a general idea of some typical queries. Overall, 2010 queries
were submitted in the 24 hour period, 933 of which were unique. Of these, only three
were in a foreign language. 85 of these unique queries were submitted five times or
14
Table 6 – Search queries by category
conducted (i.e. most artist names and the single words) would be unproblematic to
because all users are able to formulate queries in English, because people assume they
cannot even try to enter foreign language terms, or because foreign language speakers
access their information via a different route (i.e. navigating through the site pages.) In
any case, it does not appear as though creating a CLIR-based search functionality for
Since not all content can be manually translated, one option would to be to give
not available. It may be that when it comes to non-English content, having something is
better than nothing, and people will be happy with simply getting the gist of an
automatic translation.
is to use an on-line resource to translate an entire URL. Clearly, the advantage of this
approach is that it is fast and can accommodate several languages. Perhaps most
importantly in the context of the Collection, this method can be used to translate
15
dynamic web pages whereas other approaches (i.e. human translation or a software
program) cannot.
the system’s Web site. However, this requires extra effort on the part of the user and
instant translations through a language toolbar that can be added onto all desired pages.
participants were given a packet of texts and questionnaires to read and complete. The
texts were randomly selected from the Tate Collection site: one was a painting
description and the other was an artist biography. These texts were translated into each
participant’s native language using one of two freely available MT systems (SYSTRAN
system and WorldLingo.) Also included in the packet were the original English texts;
participants were able to consult these if they wished but it was not mandatory.
Participants were asked to read the first translated text and then to answer some
questions evaluating its comprehensibility, its acceptability, and its fidelity to the
original (if applicable.) They were also asked if they were able to learn something from
the text and if it would be a useful resource for writing an academic paper. This
16
SYSTRAN 4.00 2.80 3.61
Table 7 – Mean ratings for the painting text (1 is the most positive, 5 is the most negative.)
Table 8 – Mean ratings for the artist text (1 is the most positive, 5 is the most negative.)
valuable, then responses for categories (except acceptability) did not meet this criterion.
For both the painting and the biographical text, acceptability received a lower (more
must still be noted that on average, respondents felt that both texts required “a definite
effort to understand the meaning.” Whether the use of free online systems is an
possible, however, that paid, “professional” machine translation systems would provide
17
This brief exploratory study seems to correspond with findings by Guyon (2003),
content on a Web site: if readers perceive the texts as not being very serious, then the
feasible, other forms of computer support are worth considering” (Connolly, 1996).
These could include giving users with moderately good language skills links to on-line
dictionaries or thesauri, which could provide fast and convenient means to help clarify
unknown words. The potential benefit of using “other” forms of support may be
User Testing
method is international user testing,” having real users perform real tasks with the
system without receiving help. User testing was conducted to compare the performance
of non-native and native English speakers on the existing version of Tate Online.
In total, 14 participants aged between 16 and 35 took part in the study. Seven of
these were non-native English speakers who were visiting Tate Modern. Seven native
English speakers were chosen to serve as a control group: three of them were recruited at
Tate Modern and the other four were British university students. Overall, this group had
roughly the same age and gender balance as the non-native English speaking group.
Participants were brought to a public access computer room on the ground floor
of Tate Modern. None of the participants were frequent visitors to Tate Online; all had
18
Following an initial site familiarisation period, participants were given three
tasks to perform, each of which was related to a specific scenario. The first was to
locate five paintings currently on display at Tate Liverpool, the second was to find and
read a description of an artwork, and the third was to locate and read the biography of
Francis Bacon. These tasks were chosen to represent some typical browsing or free-
searching behaviours that site visitors might perform. Upon completing the three tasks,
participants filled out a short post-test questionnaire asking them to rate the difficulty of
using the Web site and also asking for any specific comments or suggestions they may
have had.
completed, the time elapsed before a participant either completed a task or gave up, and
Group Task 1 mean Standard Task 2 mean Standard Task 3 Standard Mean Standard
time deviation time deviation mean time deviation difficulty deviation
(minutes) (minutes) (minutes) rating
Non-
English 4.031 2.417 2.617 2.400 2.446 2.542 3.14 1.952
Although these results were not significantly different between the two groups at
a level of p<.05, the differences in time taken to complete the tasks is striking.
Nevertheless, some participants in both groups accomplished the tasks quickly and
easily, while others required more time. However, none of the native English speakers
19
gave up before completing any of the tasks, while some of the non-native English
speakers did.
comments they made. Some general trends and common patterns of difficulty became
evident (for example, using the site search was often problematic, as many of the top
search results were irrelevant.) Some of these problems affected all participants and
Preferences
At the beginning of the study, participants were asked what language they
preferred to use when viewing Web sites. Six of them mentioned their mother tongue
(Italian, Portuguese, French, German, Spanish/Galician, Spanish) and one chose English.
At the end, when asked, “Given what you have experienced here, would you
prefer to use Tate Online in English or in another language?” nearly half of these
participants (three out of seven) indicated that they would prefer to use Tate Online in
English. It should be noted that these individuals performed all of the tasks with very
little difficulty. Therefore, it makes sense that they would feel comfortable with the
English site. The other participants had more trouble with the tasks, and their preference
to use another language could indicate that they may have attributed part of their
20
Participant suggestions
The final question on the post-test survey was, “Do you have any other
suggestions or comments about how Tate Online can be improved for people who speak
• (2 people) – More visible language buttons on sub-sites too (not just the
engine.)
where they can find the information they are looking for (even if the links
The user study was a valuable way to gather information about paths people
typically take to navigate and find information on Tate Online, and to witness their
pitfalls and frustrations (or lack thereof.) However, it has its limitations due to the small
number of participants and the fact that Tate Modern visitors may not be a completely
Therefore, testing with a wider range of people, particularly those whose English
is not particularly strong, would be necessary to provide a more accurate picture of the
degree to which such individuals have difficulty using the Web site. Nonetheless, many
of the observations noted and suggestions made could help to make Tate Online more
21
Summary of Key Findings
several different elements that could play a role in influencing a strategy for enhancing
multilingual access. Relevant findings in these four areas will now be discussed.
Competitors
Many art and museum-related Web sites that are similar to the Tate in content
and scope have a roughly similar degree of foreign language material available, and this
is limited to visiting information. It is clear that providing more content beyond this can
be logistically difficult and potentially expensive. These reasons in part are probably
why many organisations do not offer extensively localised sites at this time, even if they
Clearly, some languages will be spoken by more site visitors than others and will
therefore be more of a priority than others. Therefore, in choosing languages for added
content, a good place to start would be to focus on what are known as the FIGS
languages: French, Italian, German, and Spanish. These are typically rated the easiest
languages for localisation (Yunker, 2003), they were the most frequently featured across
all the competitor sites, and these four languages were also chosen when the Tate offered
translated content related to a past exhibition. In addition, many site and gallery visitors
speak these languages, into which Tate has already translated some guidebook content.
The on-line survey, log file analysis, and 2004 visitor survey all reinforce the fact
that many people use Tate Online for general browsing or for visiting information. Log
22
file analysis and the user study suggest that people very rarely type in specific artwork
titles, so for this reason, implementing a query translation system is probably not a top
priority.
Translation Resources
However, the needs and interests of an organisation like the Tate are more business-
rather than research-based. The following resources are most likely to be useful to the
Tate:
• Human translators (for any permanent or top level content, i.e. on the
instantaneous machine translations outweigh the financial costs of doing so, and if the
Tate pages could include a link to an external automatic translation site, thus making it
clear to the visitor that the sub-par automatic translations are not directly associated with
Tate Online. This could then potentially reduce the risk of the translations having a
23
Conclusion
Overall, it is likely that no matter which course of action is pursued, the addition
international visitors to Tate Online. Some individuals will require such content in order
to use the site, while others will simply find it makes their experience more comfortable
and enjoyable.
The information obtained in the present study is a good preliminary step in the
foundation may include the design and testing of new interfaces, as well as a larger
The implications of this study also extend beyond the context of the Tate. For
domain can help to inform the design of effective focused translation systems. More
broadly, the findings of this study can be applied to the MultiMatch project (or any
cultural heritage Web site with a global scope) by reinforcing the notion that the
provision of multilingual access and support would be welcomed by people who wish to
interact with content in languages they do not speak fluently. It may also need to be
media types.
share with the world. Increasing multilingual access to this (or any other cultural
heritage) collection means that more people can take advantage of the extensive and
unique resources offered. However, as the present case study has revealed, the means by
24
and degree to which such multilingual support can be provided will ultimately depend
on the financial, technological, and human resources available to a Web site’s creators.
Acknowledgements
References
Clough, P., & Sanderson, M. (2006). User experiments with the Eurovision Cross-
Language Image Retrieval System. JASIST 57(5): 697-708.
Connolly, J. (1996) “Problems in Designing the User Interface for Systems Supporting
International Human-Human Communication”. In: del Galdo, E., and Nielsen, J.
(eds.), International User Interfaces, pp. 20-41. New York: Wiley Computer
Publishing.
Eurescom Project Report (2000). Multi-Lingual Web Sites: Best Practice Guidelines
and Architecture (P923).
http://www.eurescom.de/Public/projectresults/P900series/923d1.asp
[Accessed 3 August 2006]
Graddol, D. (2006). English Next: Why global English may mean the end of ‘English as
a Foreign Language.’ Plymouth: Latimer Trend & Co; accessible on-line at
http://www.britishcouncil.org/files/documents/learning-research-english-next.pdf
[Accessed 2 July 2007]
Guyon, André (2003). Machine Translation and the Virtual Museum of Canada (VMC).
http://www.chin.gc.ca/English/Pdf/Digital_Content/Machine_Translation/
Machine_Translation.pdf [Accessed 2 July 2007]
Minerva (2005). Museo & Web: Kit for Planning a Quality Site for small museums.
25
http://www.minervaeurope.org/structure/workinggroups/userneeds/prototipo/museoweb
_e.html
[Accessed 2 July 2007]
Minerva Project (2006). Multilingual Access to the digital European cultural heritage.
http://www.mek.oszk.hu/minerva/survey/delir20060130. [Accessed 16 August 2006]
ibid. (1996). “International Usability Engineering”. In: del Galdo, E., and Nielsen, J.
(eds.), International User Interfaces, pp. 1-19. New York: Wiley Computer
Publishing.
Parry, R., and Arbach, N. (2005). The Localized Learner: Acknowledging Distance and
Situatedness in On-Line Museum Learning. In J. Trant and D. Bearman (eds.).
Museums and the Web 2005: Proceedings, Toronto: Archives and Museum
Informatics, published March 31, 2005 at
http://www.archimuse.com/mw2005/papers/parry/parry.html
Petrelli, D, Levin, S., Beaulieu, M., and Sanderson, M. (2006). Which user interaction
for cross-language information retrieval? Design issues and reflections. JASIST
57(5): 709-722.
Preece, J., Rogers, Y., and Sharp, H. (2002). Interaction design: Beyond human-
computer interaction. New York: Wiley.
Rellie, J. (2006). 10 Years On: Hopes, Fears, Predictions and Gambles for UK
Museums On-line, in J. Trant and D. Bearman (eds.). Museums and the Web 2006:
Proceedings, Toronto: Archives and Museum Informatics, published March 1, 2006 at
http://www.archimuse.com/mw2006/papers/rellie/rellie.html
26
Appendix A – Analysis of competitor sites
Language offerings of Tate Online and similar sites
PORTUGSE
JAPANESE
GERMAN
ENGLISH
RUSSIAN
CHINESE
SPANISH
ITALIAN
BASQUE
FRENCH
ARABIC
POLISH
WELSH
DUTCH
GREEK
BSL
Tate z { { { { { { { { { { { {
Online
British z
Museum
National z { { { { {
Gallery
V&A z { { { { { { { z
Museum
Natl. z { { { { { {
Portrait
Gallery
Louvre z z
Guggenheim z z z z
Bilbao
van Gogh z z
Museum
Rijks- z z
museum
Centre z z z
Pompidou
MoMA z { { { { {
Met, z { { { { { { { {
New York
Guggenheim z
New York
24 Hr z
Museum
Easyart.com z z z z z z
TOTAL 15 2 1 1 10 12 13 1 10 8 3 1 2 3 1 4
27
Functionalities offered by Tate Online and similar sites
Tate z { { {
Online
British z z { z {
Museum
National { { { z
Gallery
V&A z { { z z
Museum
Natl. Portrait z { { z
Gallery
Louvre z z Lafayette z- z z z
database
{ Atlas Kaleidoscope
database
Guggenheim z { { { { { {
Bilbao
van Gogh z z { z { z
Museum
Rijks- { z z { {
museum
Centre z z
Pompidou
MoMA z { {
Met z { { z z
New York
Guggenheim { {
New York
24 Hr { { z { z
Museum
Easyart.com z { z z z {
z - multilingual offering
- only in main language
{ - not offered
28