CV Parser Benchmark – Sprint CV vs Textkernel

Accurately parsing CVs is essential for recruiters and HR professionals who want to save time and evaluate candidates effectively. At Sprint CV, we integrate two powerful CV parsing solutions into our web interface: the native AI CV Parser and Textkernel (formerly Sovren). In this CV Parser Benchmark, we compare both solutions side by side using five different CV formats. From highly technical DIGIT-TM II templates to Europass CVs, LinkedIn profiles, branded company templates, and long non-IT résumés, we tested how each parser performs in real-world scenarios. Let’s take a look at the results!

DIGIT-TM II CV: What the CV Parser Benchmark Shows

The DIGIT-TM II format is known for its technical complexity and length. That makes it one of the most challenging CVs to parse. This format is primarily used in the European Union context and enables candidates to apply for different projects within it.

When parsing this CV with the AI CV Parser and Textkernel, we noticed that the latter failed to capture the professional experiences and projects. Aditionally, it didn’t parse the tech stack correctly. It appears that the parser could not interpret the full extent of the candidate’s career, as it assumed the career began in 2025. In addition, it did not extract the CV Summary included in the uploaded document.

By contrast, the AI CV Parser correctly calculated the years of professional experience. Moreover, it parsed the number of projects, the actual career start date. It also generated a complete tech stack based on the original CV.

Regarding the education and training details, we noticed that Textkernel classified the Master’s degree as a training course. This resulted in an inaccurate representation of the CV. By comparison, the AI CV Parser processed this section correctly. Both parsers performed well in extracting the language details.

Finally, we can see that the AI CV Parser generated a general description and a tech stack that remained faithful to the original content in the professional experience section. Differently, Textkernel collapsed the candidate’s entire career into a single entry, making it appear as though there was only one professional experience.

DIGIT-TM II Result

We believe that Sprint CV’s AI CV Parser is a far more robust tool for handling the DIGIT-TM II CV format. It not only parses languages, training, and education details, but also provides a clear picture of the consultant’s overall career, expanding on each professional experience section with a complete tech stack and a general description. Textkernel, by contrast, fails to parse the candidate’s professional experience accurately, collapsing the entire career into a single entry and thus producing a confusing result.

Europass CV Results in the CV Parser Benchmark

The Europass CV is widely used across the European Union, primarily by individual candidates. It is also recommended by educational and public institutions in Europe. This format is safe and comprehensive for people either entering the job market or seeking new opportunities.

When it comes to calculating the years of professional experience, we found that Textkernel provided a more accurate result. On the other side, the AI CV Parser missed some years from the original CV. The same applies to the number of projects. Nevertheless, Sprint CV’s AI CV Parser captured more skills in the tech stack, thus performing better in that aspect.

Regarding languages, the AI CV Parser excelled at capturing the proficiency levels for each language, while Textkernel fell short. In terms of trainings and educational details, both parsers performed accurately.

Lastly, even though the AI CV Parser did not capture all professional experiences and grouped the initial ones under a single entry, it still provides solid general descriptions and an accurate tech stack for each role. Textkernel delivers similar results in this area, although the writing style of its entries differs slightly.

Europass CV Result

In the Europass case, we consider this a draw. Both parsers performed satisfactorily. The AI CV Parser missed some professional experiences. Textkernel failed to capture the language proficiency levels and some skills in the tech stack. However, both did a great job with trainings, education, and the general descriptions under individual experiences.

How the Company CV Result Stacked Up in the CV Parser Benchmark

Many recruiting agencies rely on branded CV templates to present candidates consistently to clients and projects. That’s why, when using HR software to parse CVs, they need a solution that ensures consistency and can accurately extract all the information from branded templates.

In the initial sections, we can see that Textkernel failed to parse the professional experience and projects correctly. The AI CV Parser captured everything accurately.

Similarly, while the AI CV Parser captured every detail related to languages, education, and training, Textkernel confused some of the education entries, failed to identify language proficiency levels, and incorrectly generated certifications from the tech stack, duplicating skills and placing them in the wrong category.

Finally, Textkernel was unable to create a consistent professional experience section from the information in the branded CV. Conversely, the Sprint CV AI CV Parser captured every experience and, as usual, generated a clear general description and corresponding tech stack for each one.

Company CV Result

Our third example shows that Textkernel is not strong at parsing CV templates of this kind. It failed to provide a clear view of the candidate’s experience and therefore could not generate an accurate career description. It was also unable to attribute correct proficiency levels for the consultant’s languages. However, in this scenario, we believe it performed slightly better than the AI CV Parser on the tech stack, as it produced a somewhat larger set of skills.

LinkedIn Profile CV Performance in the CV Parser Benchmark

Our fourth example is a CV format commonly used by individual consultants: the LinkedIn CV. This format is shorter and simpler than, for example, a Europass CV. However, because it is downloaded directly from LinkedIn, where many people regularly update their professional experience, it is widely used and offers a very practical way to have a CV ready to share. For this reason, parsing precision is especially important, as many recruiting firms receive a large number of CVs in this format.

At first glance, both parsers handled this CV format well, producing very similar results that matched the content of the original document. In addition, both generated a CV summary identical to the one in the original CV.

In any case, it is clear that Textkernel underperformed compared to the AI CV Parser in the areas of education, training, and languages. First, it parsed incorrect dates in the education section (for example, January 2006 instead of September 2006). Second, it assigned the wrong proficiency levels in the language section, indicating the candidate had only an A1 level, which contradicts the original CV. Finally, it duplicated the two training entries.

In the Professional Experience section, the general descriptions in the Textkernel result are more confusing than those produced by the AI CV Parser. To be fair, however, Textkernel did manage to parse all experiences, titles, dates, and the tech stack.

LinkedIn Profile CV Result

While the AI CV Parser performed better on the LinkedIn CV, Textkernel’s result was not as poor as with the DIGIT-TM II format. It captured every experience, training, education, language, and technical skill, but at times it failed to categorize them correctly, duplicated entries, or produced more confusing descriptions. The AI CV Parser, on the other hand, delivered consistent results for this CV format.

CV Parser Benchmark: Evaluating the output from Non-IT CVs

Our final example is an individual non-IT CV. With this test, we want to evaluate the versatility of each parser and see how well they handle CVs from industries outside of IT, which can be just as complex.

The first impression when reviewing the parsing results is that this is a very large CV, spanning multiple decades of experience and dozens of projects. While Textkernel did capture some of this information, it failed to parse a significant part of the candidate’s career, projects, and tech stack.

In the educational details, Textkernel did not create an entry, which is somewhat understandable since it refers to a high school degree. Regarding languages, although it failed to capture one proficiency level, its overall performance was similar to the AI CV Parser, providing a satisfactory result. For the training details, as with the previous CV tested, it captured the entries but duplicated some of them. Otherwise, the AI CV Parser created good enough sections in this case.

The professional experience section is where Textkernel fell short. It failed to parse the last 12 years of the candidate’s experience and did not generate general descriptions. In contrast, the AI CV Parser created an entry for every individual experience, each accompanied by a general description and a tech stack.

Non-IT CV Result

This last example is very complex, and although Textkernel failed to capture some parts of the career, it still did a reasonable job considering the CV’s complexity. The AI CV Parser, on the other hand, performed flawlessly, demonstrating its ability to capture all details and interpret the parsed information to generate new sections and accurate descriptions.

Overall Results of the CV Parser Benchmark

The results of this CV Parser Benchmark show a clear pattern:

• The AI CV Parser outperforms Textkernel in most cases, delivering accurate results in education, certifications, languages, skills, and professional experiences. It is able to handle large and complex CVs fairly well, producing consistent results.
• Textkernel performs on par when it comes to Europass CVs, even parsing projects and trainings more effectively in some cases. Overall, it’s not that Textkernel performs very poorly across most CVs, but rather that its results can be inconsistent and sometimes confusing, particularly with language proficiency levels, duplicated trainings, and the general descriptions of professional experiences.

Why Choose Between Parsers When You Can Use Both

Sprint CV allows companies to use the AI CV Parser, Textkernel, or both. In your company settings, you can select one parser or configure both to work together. This ensures that you benefit from the strengths of each solution and always have a reliable option available.

Conclusion and Key Takeaways

This CV Parser Benchmark confirms that the Sprint CV AI Parser is the most versatile solution. It consistently delivers more accurate parsing results across multiple formats and industries. Textkernel remains useful for Europass CVs, but the AI CV Parser is better overall at:

• Capturing complete professional experience
• Recognizing education, training, and certifications
• Parsing IT and non-IT skills with years of experience
• Identifying language proficiency levels

With Sprint CV, you do not need to choose one parser over the other. You can leverage both to maximize accuracy and efficiency in your recruitment process.

If you want to try it yourself, simply log in to your Sprint CV account, configure your preferred parser settings to your liking and start testing