Combining existing advancements in cross-language information retrieval (CLIR) with the new usercentered Web paradigm could allow tapping into Web-based multilingual clusters of language information that are rich, up-to-date in terms of language usage, that increase in size, and have the potential to cater for all languages. In this chapter, we set out to explore existing CLIR systems and their limitations, and we argue that in the current context of a widely adopted social Web, the future of large-scale CLIR and iCLIR systems is linked to the use of the Web as a lexical resource, as a distribution infrastructure, and as a channel of communication between users. Such a synergy will lead to systems that grow organically as more users with different linguistic skills join the network, and that improve in terms of language translations disambiguation and coverage.
TopIntroduction
In 1926, Bertolt Brecht was making the following suggestion about the utility of the radio: “a one sided” device “when it should be two“, “an apparatus, for mere sharing out” (Kaes et al., 1994, p. 616), that should be used not just for distribution, but also for communication. These visionary statements, apply about eighty years later to the new generation of services that transformed the web into a two-sided “device” that not only distributes content, but serves as the newfound communication medium on a social, cultural, and economic level. Researchers, software developers, and enterprises found themselves challenged to create these new web services in order to accommodate our communication needs in the dynamic context of digital technologies.
Our geographical boundaries disappear when connected to the Internet. Multilingual users meander through the web in search of facts, of answers to questions, in an attempt to discover new information, or just to keep alert on what goes on throughout the world. There are though, the language boundaries. They restrict access to the web in its entirety to users that have a good command of English, the predominant language of documents distributed on the web. Balance however is changing, and is very difficult to quantify accurately. In this context, the focus of the research field of Cross-Language Information Retrieval (CLIR) is to develop systems that support users in how they locate and present answers to their queries from resources in other languages, regardless of the querying language.
One of the pivotal aspects of CLIR is translation. This process entails mapping information encoded in the query-language to information encoded in the document-language. There are two approaches to translation: a) machine translation and b) user-assisted translation. The first is widely used and is supported by a variety of language resources from bilingual lists, to dictionaries, parallel corpora, wordnets, or interlingual indexes. Due to the quality of the language resources and the way they are used, current implementations of machine translation are far from perfect.
Let us compare for example any two bilingual dictionaries for English to French and English to Japanese. These two resources will differ in coverage, source style (human or machine readable), number of translations alternatives given, form of entries (root or surface words), etc. These differences in characteristics will then need to be handled by separate parts of the translations component, making scaling to other languages challenging.
Years of research, with mixed success, focused on developing methods that would perform well, independently of the pairs of languages the translations are run between (Levow et al., 2005), and one anticipated solution comes from interactive cross-language information retrieval (iCLIR). The iCLIR approach relies on the synergy between human and machine linguistic knowledge for improving the overall performance of a CLIR system. Currently, iCLIR systems have yet to fulfill their mission, since generally human-computer interaction systems take a very long time to tune and test.
At the same time, while iCLIR researchers are working on pinning down the best ways for users to help with the cross-language retrieval task, the web landscape has been flooded by a large number of web services that support the creation of online communities and collective knowledge pools (referred to as web 2.0 services). These communities are based on ad-hoc mechanisms for sharing information, communicating on events, stories, or things to-do, and overall facilitating each other to find and identify relevant resources. The latter is in fact the goal of any information retrieval task and the motivation for changing the setting for the users involved in assisting with a CLIR task and immersing them in the highly dynamic web community.
In other words, can users collaborate online to share their linguistic knowledge in the context of information retrieval and how can it be achieved? This sets the premise for the explorations in this chapter. We will assess the potential to get users actively involved in interactive cross-language information retrieval, specifically, on how the human users can contribute to a CLIR task by: a) creating multilingual resources, b) annotating web resources with metadata in different languages, c) mapping a query to its appropriate translation, or d) marking relevant results obtained from a cross-language query.