Instructional Videos and Others on Youtube

Similarities and Differences in Comments

Hugo Silva and Isabel Azevedo

Games, Interaction and Learning Technologies (GILT) Research Group, Instituto Superior de Engenharia do Porto,

P. Porto, Porto, Portugal

Keywords: Informal Learning, Instructional Videos, YouTube.

Abstract: YouTube is a video sharing platform and its resources have been used for formal and informal learning.

Users can add comments, as well as sign that they like a given video. The work described in this paper is

mainly devoted to the comments provided by users and how they differ (or not) depending on the type of

videos: those that are used to support learning versus those that does not. An application was developed to

collect data available on the YouTube platform. The analysis of comments extracted from YouTube was

performed using natural language processing techniques and differences in writing were also analyzed. Two

major groups of videos were considered: technical, or instructional videos, and non-technical videos. The

former usually have an educational nature and are watched by people that aim to improve their knowledge

and skills, while the others are more devoted to entertainment. The similarities and differences found

between these different types of videos are discussed.

1 INTRODUCTION

Formal and informal learning are not straightforward

concepts. Different dimensions have been used to

distinguish them. The physical space where learning

occurs is one of them, where the main distinction is

related to the use of in- or out-of-school learning

environments (Ramey-Gassert, 1997), e.g. a

classroom.

Marsick and Watkins compared formal and

informal learning: “Formal learning is typically

institutionally sponsored, classroom-based, and

highly structured. Informal learning, a category that

includes incidental learning, may occur in

institutions, but it is not typically classroom-based or

highly structured, and control of learning rests

primarily in the hands of the learner. (…) Informal

learning can be deliberately encouraged by an

organization or it can take place despite an

environment not highly conducive to learning.”

(Marsick and Watkins, 1990).

Another often used characteristic is who mainly

manages the experience, e.g. the learner or a teacher.

Smith explores informal learning as an

administrative notion (Smith, 2008).

Incidental learning, on the other hand, is defined

as “a byproduct of some other activity, such as task

accomplishment, interpersonal interaction, sensing

the organizational culture, trial-and-error

experimentation, or even formal learning” (Marsick

and Watkins, 1990).

As many professionals need to continuously

acquire new skills to keep up with their usual

activities, the number of online resources, such as

instructional videos, forums, technical blogs, FAQ

websites and others, has been increasing.

YouTube videos are available to a great number

of people, including self-learners. They have been

used for formal and informal learning. Watching an

instructional video demands an intention to learn

about a given subject. Thus, incidental learning may

occur secondarily, but for others topics.

YouTube resources have been utilized in

academic contexts as part of the adopted

pedagogical strategies. Tan and Pearce described

their experience in using YouTube videos in a ten-

week introductory sociology course (Tan and

Pearce, 2011). The students’ insights on the use of

YouTube were generally positive but the learners

emphasised the importance of watching the videos in

the classroom with teachers’ support and wide

discussion.

Azevedo et al. reported about 7,600,000 results

obtained with ‘tutorial’ as search term (Azevedo et

418

Silva, H. and Azevedo, I.

Instructional Videos and Others on Youtube - Similarities and Differences in Comments.

DOI: 10.5220/0006333104180425

In Proceedings of the 9th International Conference on Computer Supported Education (CSEDU 2017) - Volume 1, pages 418-425

ISBN: 978-989-758-239-4

al., 2014). As of December 2016, there are about

169,000,000 results for the same search

specification, including one entitled “Java

Programming Tutorial - 1 - Installing the JDK” with

more than 4,600,000 views. This video has more

than 5,000 comments and one of them is “…What is

Java actually used for, in terms of programming?

...”. It had some responses, including this one: “As

far as I know: Desktop apps/games and android

apps”.

Another comment on a video entitled “REST

API concepts and examples”, starts this way: “Your

content is absolutely excellent. I very much

appreciate your videos and am learning a lot from

them. I just wanted to give you some tips, as a web

developer myself, but someone who spent 5 years as

a teacher. Your videos are very well structured,

however I feel that you, at some points, are speaking

too fast. …”.

Despite the large number of papers about

YouTube platform, little is known about the

comments provided to educational videos freely

available on YouTube. The main research question

of this work is: What kind of information do these

videos’ comments give when compared to others,

and how can this information be retrieved?

In this research, we analyzed the following

characteristics: the use of emoticons, linguistic

characteristics, number of response to comments,

among others.

The rest of this paper is organized as follows.

This introduction provides detailed information

about the context and the research question of this

work. The second section mainly examines the

YouTube platform as an online social network and

the allowed interactions. The next section discusses

how data was extracted, characterizes the data and

provides a detailed data analysis. The penultimate

section summarises the results, discusses the most

relevant contributions and includes some final

remarks. Finally, the last section states future

research in the field, mainly based on the limitations

of this work.

2 YouTube

Online social networks are among the most visited

websites. An online social network allows a direct

interaction between users. Facebook, Twitter,

Google Plus and YouTube are some examples of

social networks. In most of these social networks,

the users establish connections between them in a

user-user interaction (Wattenhofer et al., 2012). In

contrast, social networks like Facebook and Google

Plus provide a different experience, with users

sharing personal thoughts, URLs, videos and

creating connections with other users (friendship).

The interaction between users is perhaps the

utmost of social networks, since users make

appreciations about contents published or shared by

other users. Alternatively, users can express

themselves also by commenting publications of

other users.

YouTube is a somewhat different social

network, in which the main goal is the video sharing.

Created in 2005, YouTube raises about 137 million

users per month (Weaver et al., 2012). YouTube

platform is also accessible on mobile devices

through an optimized website or apps. In YouTube,

a user can create a channel, upload videos, comment

and express his opinion about a video using the

appreciation options (like or dislike). YouTube

videos are embedded in other websites and are

frequently shared in other digital social networks.

The usability and functionality of YouTube,

which allows users to easily create a channel and

post content that is shared almost instantaneously in

the internet, turns YouTube into an attractive

platform to content creators and media companies

(Susarla et al., 2012). The content of the videos is

key for the success of YouTube, as many videos

become viral when shared in social networks

(Weaver et al., 2012). A “most-viewed” category is

another indicator of the popularity of YouTube

videos.

In this social network, social interaction is

different from others (e.g. Facebook). On YouTube,

social interaction is more often a user-content-user

interaction (Wattenhofer et al., 2012) and thus users

do not establish a bidirectional connection with other

users unlike Facebook. On YouTube, connections

between users are based on channel subscriptions,

which were created by other users. The most

expected interactions are video appreciations (like or

dislike) and commenting videos.

3 CASE STUDY

YouTube is a web platform where users can

demonstrate a positive or negative opinion by

clicking the like or dislike buttons, respectively, or

can also express their opinion by writing a comment

on the video.

For example, the popular video Psy - Gangnam

Style (OFFICIALPSY, 2012) reached more than

two billion of views and more than four million

Instructional Videos and Others on Youtube - Similarities and Differences in Comments

419

comments. In January 2016, this video had more

than 13 million appreciations (88% likes and 12%

dislikes). However, some information remains

unknown:

 Does this happen in other types of videos?

 How do users comment videos?

With these questions in mind, our work aimed at

study and compare two different types of videos and

their users’ behaviour. Two different categories were

defined: technical videos and non-technical videos.

The technical videos selected for this analysis were

tutorials and learning video-classes related with

programming contents. The non-technical videos

consisted in movie trailers.

This work was divided in three main steps: data

collection, pre-processing and analysis with natural

language processing techniques.

3.1 Data Collection

An application to extract data from YouTube videos

was developed (see Figure 1).

Figure 1: Solution components.

Data extraction represents the data retrieved

from YouTube to be analyzed. This component is

responsible for communicating with YouTube API

and extract the videos and comments. Extracted

comments are stored in a database.

The YouTube API (version 3, REST) was used

to search for the videos according to the thematic

defined and to collect the data. As this work was

focused on analysing texts in English, it was

intended to retrieve content and comments mainly in

English. To achieve this, the parameter

relevanceLanguage provided by the API was used

with the value en.

For each type of video, 150 videos and

respective comments were collected. The search

terms were “movie trailers” for non-technical videos

and “javascript”, “java programming” and “php

web” for technical videos. The methods used were

Search.list for searching contents, Videos.list to get

video info, and CommentThreads.list to get

comments’ info. The API returned the data in JSON

format.

For each video, the following information was

kept: title, description, URL, video identification,

number of likes, number of dislikes, number of

visualizations, and channel identification.

3.2 Sample Characterization

In Table 1 and Table 2, the number of videos and the

respective YouTube category for technical and non-

technical collected videos is presented.

Table 1: Technical videos per category.

YouTube Category Number of videos

Howto & Style 60

Education 45

Science & Technology 36

People & Blogs 7

Movies 1

Classics 1

Table 2: Non-technical videos per category.

YouTube Category Number of videos

Film & Animation 87

Entertainment 50

People & Blogs 4

Comedy 3

Trailers 3

Shows 2

Music 1

For each video, the first 5000 comments, if

available, were collected in descending order by

date, as it has been used before (Cambria and White,

2014) and (Baccianella et al., 2010, p. 0). The data

collected for each comment was: comment text,

identification, date and time, author’s name, channel

identification and, if the comment is an answer to a

previous comment, the comment parent

identification.

Table 3 provides data about the number of

comments and their answers. Notice that

proportionally there are more comments provided as

response under the technical videos category (17%

of response-comments for technical videos and 14%

for non-technical videos), a point that is examined in

section 3.5.

CSEDU 2017 - 9th International Conference on Computer Supported Education

420

Table 3: Number of comments per type of video.

Technical

videos

Non-technical

videos

Number of

comments

73,901 657,014

Number of main

comments

61,486 565,370

Number of

comments with

responses

6,430 37,912

Number of

comments provided

as responses

12,415 91,644

Technical videos have an average of 91,4

characters per comment, while non-technical videos

tend to have longer comments. An average of 116,6

characters per comment was measured in our

sample.

Regarding the number of users represented in the

comments, we have 41,430 people for instructional

videos and 466,417 for the others.

According to the number of videos commented

by each user (see Table 4 and Table 5), users seem

to be prone to watch more than one instructional

video, perhaps aiming to improve learning.

Table 4: Percentage of users and number of videos

commented.

Number of commented videos

Percentage

of users

1 80,38

2 12.05

3 3.65

4 1.56

5 or more 2.36

Table 5: Percentage of users and number of videos

commented per category.

Number of

commented

videos

Percentage of

users (technical

videos)

Percentage of

users (non-

technical videos)

1 71.50 81.30

2 15.9 11.66

3 5.63 3.44

4 2.56 1.45

5 or more 4.41 2.15

On technical videos, 18.26% of comments have

(single) question marks, as usual in questions, a

number that decreases to 13.39% for non-technical

video comments. It was also verified that 90% of the

technical video comments with question mark (?) do

not have this punctuation signal repeated. In non-

technical videos, this percentage is 79.4%.

3.3 Pre-Processing

To be able to perform an analysis on the collected

comments, it was necessary to prepare the comments

considering the type of writing on YouTube. Writing

comments in platforms like YouTube can be seen as

computer mediate communication (CMC)

(Hogenboom et al., 2015). It is very common in

YouTube comments the use of abbreviations,

repetition of characters and signal punctuation and

the use of emoticons.

The use of repeated characters (chars) is often a

way to reinforce an idea. For example, the use of the

word “looooove” has clearly the intention of

emphasizing the positive sentiment. However, that

word does not exist in any dictionary or lexicon. The

removal of repeated characters in this work uses the

same approach used in (Lehnert and Ringle, 2014).

Words with at least three equal consecutive

characters were considered to have repeated

characters. For those words, the repeated chars were

removed by using the following approach: the tool

removed all but two consecutive characters; then the

new word was checked in a dictionary; the word was

considered if the dictionary recognized it; otherwise,

the last repeated char was removed. For example:

 goooooood → good (exists)

 loooooove → loove (does not exist) → love

(exist)

Two percent of the comments for technical

videos had repeated chars that were removed. In

non-technical videos, the number of comments with

repeated chars removed was four percent.

Repeated punctuation was also removed to

achieve a better performance on analysis as

described in (Liu, 2012). The punctuation signals

identified were the full-stop (.), the question-mark

(?) and the exclamation point (!). Two or more

consecutive occurrences of a punctuation signal

were considered as repeated punctuation (except for

reticence - three consecutive full-stops). Repeated

punctuation was removed in 13% of the comments

of technical videos and 18% of the comments of

non-technical videos. The most frequent repeated

punctuation signals are identified in the Table 6.

The most common emoticons presented in

technical and non-technical videos are presented in

Table 7 and Table 8.

Notice that there are 7,746 comments for

technical resources with at least one of the five

emoticons most used for technical videos as listed in

Table 6 and 17,099 for non-technical resources with

the top-5 emoticons listed in Table 8. Proportionally,

Instructional Videos and Others on Youtube - Similarities and Differences in Comments

421

it seems that the first group tends to have more

positive emoticons.

Table 6: Repeated punctuation resume.

Comments

of technical

videos

Comments of

non-technical

videos

Full-stop (.) 53.99% 44.59%

Exclamation mark (!) 30.40% 36.52%

Question mark (?) 13.52% 17.43%

Others (, or ;) 2.09% 1.46%

Table 7: Top 10 of emoticons on technical videos.

Emoticon # comments

1 :) 4,737

2 :D 1,481

3 :P 633

4 ;) 476

5 xD 419

6 :( 414

7 XD 255

8 :/ 166

9 :-) 136

10 :p 127

Table 8: Top 10 of emoticons on non-technical videos.

Emoticon # comments

1 :) 6,822

2 :D 4,760

3 <3 1,969

4 XD 1,814

5 xD 1,734

6 ;) 1,640

7 :( 1,590

8 :P 1,390

9 :/ 818

10 :3 497

By the analysis of these tables, it can be

concluded that the top 2 emoticons are the same.

However, the third most used emoticon on non-

technical video comments (Table 8) is <3, used to

represent a heart, which expresses an intense

sentiment. This emoticon is not often seen in

comments for technical contents.

3.4 Linguistic Analysis

To determine the frequency of the terms used in the

comments, word clouds were generated. For

technical videos, three types of clouds were

produced: i) the most used words (Figure 2), ii) the

most used adjectives (Figure 3), and iii) the most

used verbs (Figure 4).

Figure 2: Words’ frequency in technical video comments.

Figure 3: Adjectives’ frequency in technical video

comments.

Figure 4: Verbs’ frequency in technical video comments.

CSEDU 2017 - 9th International Conference on Computer Supported Education

422

For non-technical videos, the same three types of

clouds were produced. For the most used words

(Figure 5), the word “movie” stands out. Figure 6

and Figure 7 present the most used adjectives and

verbs, respectively. Despite not being highly used,

the words “wait” and “see” appear in Figure 5 and

Figure 7, which normally results from sentences as

“can’t wait to see”.

Figure 5: Words’ frequency in non-technical video

comments.

Figure 6: Adjectives’ frequency in non-technical video

comments.

Figure 7: Verbs’ frequency in non-technical video

comments.

3.5 Comments and

Response-Comments

YouTube allows users to reply other users’

comments. In the collected data, 14% of response-

comments were identified for non-technical

resources and 17% for technical videos (see Table

4). The higher percentage of answers in technical

videos might be an indicator of the behaviour of

users in this type of videos. Videos with learning

content are more prone to have comments with

questions or doubts related to the video content, and

in turn, to have more answers, replies and

thankfulness.

Likewise, for technical videos an average of 1.9

responses was observed for comments with replies,

while for non-technical videos this value is 2.4.

With the method CommentThreads.list of the

API, it was possible to retrieve the number of

positive appreciations (likes) of the comments. The

percentage of comments with positive appreciations

was 22.8% for technical videos and 22.3% for non-

technical videos, which corresponds to 7.3 and 10.7

likes per comment, respectively.

Furthermore, as explained before, our analysis

detected that technical videos have an average of

91.4 characters per comment and non-technical

videos have an average of 116.6

characters/comment. In which regards to correction

for repeated sequential characters, that was

necessary for 4% of non-technical and 2% of

technical video comments. However, the number of

response-comments that needed this type of

correction was close to 0. As the use of repeated

characters tends to reinforce an idea, this value

Instructional Videos and Others on Youtube - Similarities and Differences in Comments

423

reveals that when users reply to comments, they tend

to be more objective in their message. The same

tendency was observed for repeated punctuation

marks: 18% of repeated punctuation marks were

found in non-technical video comments against 13%

in technical video comments.

3.6 What Users Talk about

The comments collected have different

characteristics in which regards to the type of video

they correspond to. For technical videos, the most

common technical terms related to the extracted

videos were searched, and are related to

programming languages. A set of technical terms

related to programming languages was defined and

the most frequent terms on comments were

disclosed. Figure 8 shows the ten technical terms

most frequent in technical video comments and their

frequency.

Figure 8: Frequency of technical terms.

The common use of terms like "class", "int",

"string", "static" and "void" in comments suggest

that users write code in comments to ask or answer

questions.

We also investigated which topics related to the

movies presented in the trailer videos are more

frequent in the comments extracted by identifying

the code of the respective IMDB (Internet Movie

Database) movie. With this ID, and with the Open

Movie Database (OMDB) API (Fritz, 2016), we

extracted information about leading actors, director,

writers and the movie title. Then these comments

were analyzed to know how many comments

mention these topics and which ones are more

mentioned. Only 4.06% of the comments refer to

topics collected from IMDB. Figure 9 shows the

frequency of references to actors, directors,

scriptwriters, and the title.

Figure 9: Frequency of IMDB topics.

The movie title and the name of the actors were

the most referenced topics in these comments (a

total of 13,861 references to the title; 10,667

references to the names of the actors; 2,157

references to scriptwriters; and 1,916 references to

filmmakers).

4 CONCLUSIONS

The findings of this study indicate that there is a

sense of gratitude between viewers of instructional

videos. Not only the word cloud includes the word

“thank” but also users tend to express their

appreciation using some emoticons such as “:-)”.

Considering the top-5 emoticons used for the two

categories of videos, 10,48% of comments for

instructional videos with positive icons and 2,60%

of comments with these positive symbols for the