Yoyogi seminar achieves lectures that are easier to understand and more effectively learned, and enhances service quality by leveraging "BlendVision AiM."

Yoyogi Seminar, Takamiya Gakuen Educational Corporation

Trigger for Use

• Content analysis and improvement to provide better services.
• Reduction of resources for managing a vast amount of video data.

Impressions After Use

• Significant advancements in both accuracy and speed.
• Discovery of new possibilities in an unprecedented approach targeting video information.

Future Prospects

• Use for video tagging tasks to reduce work time.
• Improved channel video assets for further service improvements.

Yoyogi Seminar, a university entrance prep school with six locations across Japan, provides online lecture recordings to students and partners. To efficiently manage and utilize their vast archive of 20,000 videos, Yoyogi Seminar started using "BlendVision AiM." We spoke with Mr. Hajime Ito, General Manager of Corporate Planning Office's Sateline Education Planning Department, about the potential of "BlendVision AiM" for video content planning and production.

The creation of approximately 8,000 90-minute videos each year presents ongoing challenges to maintaining and improving quality.

How does Yoyogi Seminar produce and distribute its video content?

Mr. Hajime Ito from the Corporate Planning Office, Sateline Education Planning Department (hereinafter referred to as Mr. Ito) explains:

"Since the test preparation textbooks are revised every year, the lecture videos are updated daily to keep pace with these changes. We produce about 8,000 videos, each approximately 90 minutes long, annually. We conduct filming almost every day. After recording, we check the footage for any problematic remarks, and we begin distribution within just a few days of recording, operating with a rapid turnaround."

In managing such an enormous amount of content, what challenges did you encounter?

Mr. Ito:

"The recorded videos are labeled, categorized and itemized—by alumni students of Yoyogi Seminar according to the lecture content. One of the challenges is reducing the labor and time required for this labeling work. Another challenge is making effective use of the archive. Because we are producing content at a very rapid pace, we simply don't have enough time to analyze each piece of content in detail to further improve the lectures and videos. It was at that point that we heard about 'BlendVision AiM' and thought it might provide a solution to these challenges, so we decided to trial it."

‍

Extracting the targeted videos from an enormous archive.

Yoyogi Seminar has experienced significant improvements in both accuracy and speed.

How are you currently using "BlendVision AiM"?

Mr. Ito:

‍
AiM is being tested by uploading approximately 300 videos of around 30 minutes each. These videos cover a wide range of subjects, including English, mathematics, Japanese, science, and history. The goal is to evaluate AiM's ability to accurately recognize exam-related terminology and correctly identify and classify subjects, topics, and units.

What are your impressions of using "BlendVision AiM"?

Mr. Ito:

The speech recognition was a significant improvement over traditional systems, especially in accurately transcribing audio to text. However, it did struggle with specialized terminology and mathematical formulas, and occasionally failed to return results even when queried via chat. Despite this, the ability to instantly suggest related videos and extract information based on classification categories after uploading a video was impressive, showcasing significant technological advancement.

Additionally, the speed of development and improvement is notably fast, and their responsiveness is very adaptable. Our requests were met with near immediate feedback, such as  'We can implement that at this speed,' or 'This feature will be added in the future.' This aspect was also very impressive.

Improved operations and lecture quality are anticipated with the implementation of BlendVision AiM as an assistant tool for lecturers and administrators.

Based on its current functionality, what potential do you see in "BlendVision AiM"?

Mr. Ito:

AiM could potentially reduce the time spent on tasks that were previously done manually, such as tagging and classifying video content. Additionally, it could be used to enhance the quality of lectures by allowing lecturers to easily search for and review past lectures on a particular topic. With BlendVision AiM, lecturers can extract and review only specific parts of a lecture, without having to watch the entire video.

While AiM's responses are not yet 100% accurate, the current level of precision is sufficient for knowledgeable users, such as lecturers, who can discern the correct parts and disregard the inaccuracies. AiM could also be used by the video content sales department to suggest relevant videos to partners. Furthermore, students could use AiM to search for and access specific sections of lectures, which could contribute to more efficient learning.

Maximizing AiM's Potential Through Strategic Content Creation

Does the potential of "BlendVision AiM" depend on how it is used?

Mr. Ito:

"Absolutely, if a user possesses substantial knowledge of the video content, AiM can function as a supplementary tool for various applications. However, my experience with BlendVision AiM has shown that vague keywords or questions result in ambiguous responses. Therefore, specificity is essential when interacting with AiM. For instance, 'Tell me about participial constructions' yields more accurate results than a broad query like 'Tell me about English study methods.'"

How do you think "BlendVision AiM" can be utilized in future video content creation?

Mr. Ito:

In order to optimize BlendVision AiM's potential, it is essential to modify lecture content accordingly. AiM operates by transcribing audio to text, which means it cannot analyze anything that is not spoken. For example, if information is written on a board without being verbally explained, it will not be recognized. Therefore, instructors should verbalize all key points and summaries, and explicitly mention relevant textbook page numbers. This will ensure that all essential information is captured by AiM for analysis and suggestions.

These adjustments will not only benefit educational settings, but also businesses who utilize BlendVision AiM. In situations where verbal explanations are limited or not possible, adding narration to supplement visual content can enhance AiM's effectiveness.

What do you expect from AiM in the future?

Mr. Ito:

The digitization and utilization of video information has presented a new direction for the video industry beyond entertainment, which is a departure from the traditional approach. BlendVision AiM has  opened up new possibilities, and I hope it will continue to evolve by proactively adding new features and updates.

‍