An In-Context and Collaborative Software Localisation Model

Amel Fraisse, Christian Boitet, Valérie BellynckCOLING 2012

TL;DR

This paper introduces a novel "in-context and collaborative" model for software localization that leverages end-users and volunteer localizers. By allowing users to translate software strings directly within the application's interface as they use it, the model aims to overcome the limitations of traditional, out-of-context translation methods, leading to more efficient and contextually accurate localizations, especially for under-resourced languages.

What Problem Does This Paper Address?

The traditional software localization process, which relies heavily on professional translators and often translates strings in isolation from their actual use, faces significant challenges. These include high costs, scarcity of professional translators for many languages, and the inherent difficulty of producing accurate translations without understanding the context in which a string appears. This can lead to awkward or incorrect translations, particularly for user interface elements. The paper addresses the need for a more efficient, cost-effective, and contextually aware localization approach, especially for open-source software and under-resourced languages.

Methods

The core of the proposed model is an "in-context and collaborative" localization workflow. This involves:

  • In-Context Translation: Users, while actively using the software, can right-click on any UI string and initiate a translation or correction process. This provides immediate context for the translation task.
  • Collaborative Platform (SECTra_w): A collaborative platform is integrated to manage the translation process. This platform allows users to submit new translations, view existing ones, and potentially benefit from machine translation or translation memory suggestions.
  • Volunteer and End-User Involvement: The model actively recruits non-professional translators, including volunteer localizers and, crucially, end-users of the software. These users are familiar with the target language and the application's usage context.
  • Incremental and Continuous Localization: Instead of aiming for a perfect, final translation upfront, the model embraces an iterative approach. Rough translations are published and progressively improved over time through user contributions. This allows for continuous updates and faster availability of localized versions.
  • Two Interaction Scenarios:
    1. Direct Interaction: Users interact with the collaborative platform directly from within the application's interface.
  • Workspace Interaction: Users can access a dedicated "online workspace" on the collaborative platform to perform translations, after which the application interface is updated to reflect these changes.

Datasets

Not discussed in the paper. The paper focuses on the methodology and demonstrates its application on Notepad++.

Key Insights and Ideas

  • Context is King for Localization: The paper strongly emphasizes that translating UI strings without their contextual usage leads to suboptimal results. By bringing translation directly into the user's workflow, accuracy and relevance are significantly improved.
  • Democratizing Localization: Empowering end-users and volunteers to contribute to localization can significantly reduce costs and increase the availability of localized software, especially for languages that are not commercially viable for professional translation.
  • Embracing Imperfection for Speed: The shift from a "perfect translation" paradigm to an "incremental improvement" model allows for faster deployment of localized versions and continuous quality enhancement, mirroring successful community-driven translation efforts like Wikipedia.
  • Bridging the Gap Between Users and Developers: The collaborative platform acts as a crucial intermediary, facilitating communication and contribution between users and the localization project.

Key Results

The paper presents a demonstration of the model's functionality using Notepad++ as a case study. While specific quantitative results (e.g., translation accuracy improvements, time savings) are not detailed, the paper illustrates the practical implementation of the in-context and collaborative localization process through screenshots and descriptions of the user interaction scenarios. The qualitative outcome is a more integrated and potentially more efficient localization workflow.

Limitations and Future Work

  • Scalability and Quality Control: While the paper highlights the benefits of involving a large number of users, the challenges of managing quality control and ensuring consistency across a large, distributed group of contributors are implicitly present. The paper mentions professional translators and reviewers being asked to translate "crucial parts," suggesting a hybrid approach to manage quality.
  • User Motivation and Engagement: The success of this model relies heavily on user motivation to participate in the localization process. Strategies for sustained user engagement are not deeply explored.
  • Integration Complexity: Integrating a collaborative platform seamlessly into various software applications might present technical challenges.

Future work could involve more extensive empirical studies to quantify the benefits in terms of translation quality, speed, and cost-effectiveness compared to traditional methods. Further research into robust quality assurance mechanisms and effective user engagement strategies would also be valuable.

Summary updated

BibTeX
@inproceedings{fraisse2012context,
  author = {Amel Fraisse and Christian Boitet and Valérie Bellynck},
  title = {{An In-Context and Collaborative Software Localisation Model}},
  booktitle = {International Conference on Computational Linguistics},
  year = {2012},
}
Figure