‹ Go back to news
Extract! parsing upgrade – New language, speed improvements, accuracy improvements and more!
29 November 2018 • Categories: CV parsing, New release, News
Our powerful resume and vacancy parsing software, Extract!, has just been upgraded with new features and enhancements. In this release we’ve:
- added Japanese vacancy and resume parsing,
- improved French job parsing and Swedish resume parsing,
- enhanced parsing speed for all languages,
- and added support to deliver resume and vacancy parsing in JSON format.
Read the full details on each release item below.
New language – Japanese parsing
Extract! now supports parsing for Japanese vacancies as well as “rirekisho” resumes. This type of resume gives a brief overview of the applicant’s job and education history and is the most commonly used type of Japanese resume.
Customers that are interested in adding this new language can contact our Support team via firstname.lastname@example.org- who will be happy to help!
Accuracy boost – French vacancy parsing now powered by Deep Learning
In 2017 Textkernel achieved a major breakthrough by releasing the first resume parsing powered by Deep Learning, resulting in major accuracy improvement due to error rates decreasing by up to 60%. Since then 10 resume and 5 vacancy parsing languages have been upgraded with the power of Deep Learning, and now French vacancy parsing has been added to the list. As a result, the accuracy rate has been improved by 25%, which also directly improves Match!, as well as Jobfeed for both France and Canada.
Accuracy boost – Swedish resume parsing also gets Deep Learning power
Swedish resume parsing is another new entry to the list of parsing languages now powered by Deep Learning. This upgrade further improves the accuracy of extraction by 30% – 35%. Customers that have resume parsing for Swedish will automatically benefit from the new model. Do you wish to add Swedish resume parsing? Contact our Support team via email@example.com for more info.
Speed boost – processing time cut by up to 20%
Along with adding Japanese parsing and moving French and Swedish to Deep Learning power, we’ve been working on numerous other enhancements. We’ve improved extraction speed for all languages. The result: a typical resume now takes on average 10 to 20% less time to process. Also improved for all languages: better splitting of organisation names and places.
Languages specific improvements:
- For German cv parsing: better extraction of experience for Europass format
- For Spanish cv parsing: extraction of education levels for Spain has been improved
New feature – JSON templating now available
Last but not least we’ve added the ability to output resume and vacancy parsing in JSON format. This enables you and your development team to more easily integrate the output into your systems, though XML is also still supported.
Interested in getting your data output delivered in JSON? Please contact our Support team via firstname.lastname@example.org to discuss options.