Available Technologies

    Innovation Sphere



Project TitleDetecting and Localizing Internationalization Failures in Websites
Track Code2016-063
Short Description
TagsInternationalization, i18N, fault localization, fault detection, algorithms
Posted DateAug 30, 2017 8:51 PM

Market Opportunity

Internationalization of websites refers to developing websites that can be used in different parts of the world in different languages. Oftentimes, when a web page is translated from one language to another language, the translated web page may contain visual errors (e.g. unreadable or corrupted text that runs off the page or changed image locations). Human web page testers spend time manually detecting these visual errors and determining how to fix them. The testers use checklists to ensure that errors are detected. The results, however, are far from perfect. Testers typically detect only about 70% of the errors on a translated web page. These testers can spend more than 6 minutes per web page to return these dismal results.

In today’s robust global economy, there needs to be a process that accurately detects errors in a short amount of time. Such a technique will save money and allow companies to spend time on other problems.

USC Solution

USC researchers have developed a technique that assists web page developers with automatically detecting when internationalization efforts have caused a page’s appearance to become disoriented and helps them identify the elements responsible for the problem. The metric for evaluating this technique is referred to as recall and precision. Recall measures how well the technique finds "all" of something and precision measures how many false positives are produced. USC’s technique has 100% recall and 91% precision. Within 2 to 3 seconds, the technique accurately identifies 100% of the errors on a web page and only about 1/10 of the results are false positives. The precision number is impressive because automated detection and localization techniques that are used for other kinds of problems are known for producing numerous false positives.

Value Proposition

  • 100% accurate in detecting visual faults in web page
  • Performs fault detection and locates responsible elements in 2 to 3 seconds
  • No need for human testers


  • Internationalization efforts in web applications

Stage of Development

  • Experimentally validated

  • Available for exclusive and non-exclusive license

Intellectual Property


Provisional Number: 62/483,823

Key Publications:

Detecting and Localizing Internationalization Presentation Failures in Web Applications, IEEE 2016 (Winner of Best Paper Award – ICST)

Contact Information

Denise McKenzie, Licensing Associate

USC Stevens Center for Innovation

(213) 821-0945



File Name Description
NCD 2016-063 - Detecting and Localizing Internationalization Failures in Websites.pdf None Download