Youtube coach

a website that acts under the service of YouTube, called YouTube Coach, oriented towards providing the user with procedural instructional videos (Eg. like YouTube Kids, functioning as an independent version of YouTube).

Year

2022

Category

Academic

Duration

1 Week

Platform

Web

Team: Amit Patjoshi, Soumyoprabho Mukherjee, Jribh Shandilya, Shraddha Prabhu

Problem statement

There are potentially billions of instructional ‘how-to’ videos on the internet, majority of them being present on YouTube. The problem that persists with the existing videos is that they are unstructured or random.This reduces the opportunity of YouTube being an excellent instructional platform too. Additionally, there are many instances when a certain step followed in one YouTube video might be better than the same step from some other video, but the subsequent steps might be better somewhere else. ‍

Our goal is to introduce a system that can extract the most relevant steps from instructional videos on YouTube categorised based on a certain topic, and compile these steps to get a final instructional sequence for that topic. ‍

What is YouTube Coach?

YouTube Coach is an online tutorial compiler for YouTube that, based on the query, searches through all available videos, takes snippets from some of the most helpful ones and combines those snippets to provide an instructional compilation. The snippets are relevant steps belonging to different videos.

How we solved the problem

As an outcome of the following study for the project we had proposed a website called YouTube Coach, that acts under the service of YouTube, oriented towards providing the user with procedural instructional videos (Eg. like YouTube Kids, functioning as an independent version of YouTube). YouTube Coach, after going through millions of videos available on YouTube for that query, provides a final high quality compilation divided into steps. These steps are procured from different videos, based on their individual relevance and quality.

Being a subset to YouTube, the product only works with YouTube videos.Videos should be either in English or totally silent. For the scope of this project, we are not processing regional languages currently as a limitation, as that would require extensive filtering to be able to avoid multi-language compilations.

How do people learn using videos?

The trend of learning from videos has grown tremendously over the past few decades. Video training is the preferred method especially for Micro learning i.e.receiving information in small, highly focused chunks. Videos are one of the best ways to share knowledge, people are 75% more likely to watch a video than to read written documents or web articles. (Forrester Research) When both audio and visual inputs are available to the consumers on the same page, 72% of them would preferably watch a video to learn about a product or service. (Wyzowl Benchmark Report, 2019)

Videos boost viewer learning by providing relatable narratives in a visual format micro learning can use videos to provide the viewer 'just-in-time' learning, that can be part of a bigger learning path. Knowledge derived from micro learning can help the viewer exactly in the moment of need Knowledge when provided in small bits has a better chance of retention by the viewer (Raman, 2010), which is exactly what micro learning provides. Microlearning videos can be watched on the go, consuming bits of knowledge when needed.

What makes YouTube a great learning tool?

YouTube allows for effective microlearning: YouTube videos, due to their nature of being relatively short and precise, ensure an effective microlearning environment.

It allows easy to learn on the go: Multi platform support, especially the ability to be able to play videos on mobile and consume information.

It generates and promotes online discussions: Learning is hampered in an isolated environment, and YouTube ensures just the opposite, providing a social and discursive space in the form of comments.

It allows the students to be teachers: Giving everyone an equal opportunity of contribution ensures the formation of strong online.

What makes a ‘how-to’ video good?

Instructional videos are of many types, out of those one category of videos help people to answer certain questions or to achieve a certain goal by completing a certain task.

Knowledge is divided into
• Declarative- This type of knowledge comprises of factual information.
• Procedural- This type of knowledge helps people in developing certain skills to do a specific task.
• Causal- This type of knowledge comprises of laws and principles
• Conditional- Knowlegde of when to use declarative or procedural knowledge
• Relational- This type of knowledge is central to basic human reasoning, analogy and planning For this paper we have focussed on videos that provide procedural knowledge (like tutorials and instructional 'how-to' videos).

What constitutes ‘procedural knowledge?

A procedural knowledge component skill has seven content elements:
- An object or situation to be modified by the procedure (the task).
- The name of the procedure.
- A list of the steps or activities for executing the procedure.
- The sequence for execution of those steps.
- A demonstration of the task illustrating the individual activities or steps required.
- The consequence of each step and the whole procedure.

Demonstration for procedural knowledge:
For procedural knowledge component skills, it is best to demonstrate (Merrill, 1983)
- A specific instance of the task .
- The execution of each step should be demonstrated and the consequence of executing each step should be shown.
- The defining and ordering properties of each step should be described.
- Guidance directing learners’ attention to the name of the step being executed
- The action that is taking place, and the consequence of the action.

How is Procedural Knowledge taught effectively?

Procedural knowledge can be well taught by:-
• Using descriptions of the sequence of actions and decisions necessary for achieving goals.
• Providing the opportunity to the learners to construct on their prior knowledge.
• Providing a conceptual elaboration of the declarative knowledge base in the form of concepts, processes, and principles that explain why the procedure works.
• Complex procedures should be chunked into segments of seven to nine new steps, to avoid cognitive overload.

• Following some steps with a “whole task” practice where procedural chunks are gradually assembled into larger “wholes” to show the bigger picture. This feedback should be used to close the gap between current and required performance.

After receiving the procedural knowledge the learner should be able to Recognize and identify the key steps, apply the correct series of steps and evaluate the consequence for the whole procedure of making changes to a given step.

Qualities of a Good Instructional Video

A good instructional video is the one that has a well-structured title
• Has well-defined steps in sequence.
• Provide possible alternatives to make the steps more relatable to the masses.
• Has a clear video and audio.
• The vocals should semantically match to the subsequent part of the video.
• Instructions should be delivered at a comfortable pace
• Steps manually written down in the video description

• If an instructor is in the video watchers tend to learn more effectively than the videos which don't have an instructor.

Previous works suggest that users learn effectively from youtube and other instructional videos when videos are divided into multiple sub steps(procedural instructions).

Product Positioning- Nature of procedural knowledge targeted by YouTube Coach

YouTube Coach targets procedural videos on YouTube that provide tutorials to the viewer to accomplish a certain task or reach a certain goal by following a set of clearly defined steps. In other words, the videos should teach the user how to do something with a set of given ingredients/materials by following a sequence of steps one after the other. ‍

The tutorials can be for relatively simple and small tasks like.
• How to change a bike tire?How to make a strawberry milkshake?
• How to delete Instagram account ‍

Or they can be for something a bit more complex such as
• How to apply for a driving license?
• How to assemble a water cooled PC?
• How to apply for US Visa?

Low fidelity wireframes

Coach Home Page.

Shows the recommended tutorials, tailored just for you.

Shows steps of the tutorial where you left off last, at the top.

Search for something

See the top compilation, along with some other suggestions- sorted by relevance.

See the starting steps, rating and other important details on the search page itself.

All in One Video Player

The video is divided in steps on the player itself. Easily check and modify the source video for each step.

Can rate the compilation, comment on it, or submit some suggestions.

Become the Top Coach

Submit suggestions for compilation sources and the sequence. Coach levels can be increased with more contributions.

Youtube coach backend

Compiling the tutorials

Selection of the best and the most relevant YouTube videos to take individual steps from and then stitching them together to present the user with a compilation is done by an algorithm we would be referring here to as the Parent Algorithm. The Parent Algorithm uses four children algorithms for various specialized tasks. The steps followed by it are as follows:

Taking the best steps from relevant YouTube videos and then stitching them together to make the compilation is done by the Parent Algorithm.


The Parent Algorithm uses 4 children algorithms for various specialized tasks.

For videos with vocal instructions, Algorithm A1 is used to generate written transcript from audio

Algorithm B1 is used to

generate written transcript from visual activities based on actions happening in the vide

Algorithm C divides transcript into individual steps and assigns confidence scores, to sort the best steps.

In Algorithm D, visual detectors match transcript steps with video screenshots to score the video based on visual relevance. Thus, we have a collection of steps from all YouTube videos, ranked according to confidence scores based on their quality of instruction.

Processing User Feedback

User feedback processed by the YouTube Coach system can be either passive or active. Passive Feedback includes- Selecting the next video from the suggestions for a certain step. (Gives the algorithm an idea about the generalized behavioral patterns of users to determine better videos). Active feedback includes- Editor allowing for people to manually submit improved versions of an algorithmically suggested compilation. Star rating system to determine which compilations are working out and which are not Comments by users.

Persona 1 user journey

Arun, a XII class student, wants to make a milkshake, but can’t find a ‘perfect’ tutorial. He uses YouTube Coach to search for the compiled tutorial

User journey 1, Vimala (active contributor)

Vimala, a CS Student, wants to assemble a water cooled PC. She searches YT Coach for the tutorial, and gets one decided by the algorithm. She doesn't like it, and wants to suggest some edits.

Future Scope

Some of the features that could be implemented in future updates to the application are :
• An insider program similar to YouTube Creator Insider, acting as a social media featuring best coaches, videos with regional languages to create regional language-specific compilations,

• Training programs for the coaches to use more formal instructional methods and contribute to the algorithmic ranking system to make it more qualitative.
• The points shared in this report can be made available to the coaches on the website as tips on how to upload and compile better instructional videos.

Although not currently covered in the scope of this particular project, it can be done as part of a Coach Improvement program YouTube Coaches can formally enroll into, serving as a guide on how to give qualitatively better compilation suggestions.

Conclusion

The idea for YouTube Coach was conceived after realizing how big of a learning industry YouTube is currently supporting. Unfortunately, this wealth of learning information highly lacks structure, and the viewer is at the mercy of recommendations, a problem which we sought to address. We believe that, over time, our platform would see an exponential increase in its potential of being a good tutorial website, as the algorithm recommendations get better and coaches give more and more submissions and feedback.