You are on page 1of 4

The Measurement of the Search Charts of Music

Yue Yang, Changjia Chen, Yishuai Chen,


School of Electric and Information Engineering, Beijing Jiaotong University
caddielily@yahoo.com.cn, changjiachen@sina.com, chenyishuai@gmial.com

Abstract charts and so on. We will discuss the charts of


searching in this paper.
Nowadays, more and more singers published A discussion of real-time charts is provided in the
various songs. There are about 50,000 songs in the next section. Section 3 introduces a special chart that
internet. The songs are refreshed everyday, so we need we study in this paper. We show the data and results
a real-time chart of music. In this paper, our study is we got in section 4. Section 5 is a conclusion of our
based on the real-time search charts of music. We study. The last two sections are about the future work
studied the algorithm of the charts and analyzed the and acknowledgment.
ranking of the charts. According to the ranking of the
charts, we got some special characters. Those 2. The real-time charts
characters can not be seen from other music charts,
such as selling charts and other week’s charts of music. Most music charts can only show the musical rank
While, there are also some problems about the real- of the last day, or the last week. Those charts are only
time charts, like special users of charts. Those able to tell us what or who is the most popular song or
problems need us to solve. We will get more singer last day or week. So the message of the charts is
impersonality charts after solving those problems. not the real-time.
Sometimes we need to know the latest rank of the
Keywords: search charts, special users, real-time music.
charts For example, when a new single is published in
today's morning, then many people want to listen to the
latest song at the first time. If you don't know the latest
1. Introduction song, you will not be the in-man. Especially, when we
search the music charts, we want to know the latest
Music, likes movies and pictures now are and the hottest song. So we will expend the charts can
important parts of people's lives. Music charts play an tell us the latest results. In 2008, when the Sichuan
important role in the music industry. Giles [1] told us earthquake happened in china, the songs for the
that the charts of music can return the information to earthquake are popular. But many charts of music
the music market. Many music charts are based on the don’t show that until one or two weeks later.
selling of CDs, such as UK charts and Billboard’s [2]
charts. But nowadays, we can’t estimate the popularity 3. The naver-charts
of music only by the selling. Today there are many
ways for people to listen the music. Many people In this paper, we get the data from the naver-charts.
choose to listen to the music on internet instead of The naver-charts are real-time common charts. As we
buying the CDs. So the most popular music must not to know, the www.naver.com has over than 70% of the
be the best selling one. Today, if we want to know the Korean market. So we can say that more than 35
popular music, we can search on the internet. Through million Korean people use it. The messages of the
the hit charts of the music we will get the popularity of www.naver.com are translated to many languages and
the music. quoted in any web station of Korean entertainment.
There are many kinds of music charts, such as the The naver-charts we studied have many differences
hit charts, the download charts, the mobile ringtone from other charts. Firstly, in fact, the naver-charts is
not the hit charts, it is a searching charts. In this charts,
the rank is based on the searching times. Secondly, the there are eight days. The forth day is Saturday. The
naver-charts show the top-10 hit keywords of music in blue line in figure 2 is about the “singer and song”; the
real-time. So it is a real-time chart. Thirdly, the chart is yellow line is only about the keyword of the “song”.
keywords charts of music. The keywords of music We select another famous singer to compare with
include not only the names of the songs, but also the the A, we call it B (B=MC Mong). The song of B is
names of the singers and the lyrics. So we can see the b(b=circus). A and B are both famous Korean singers.
trend of the singer, the song and the lyrics from the Figure 3 illustrates the ranking of “B & b” and b in
figures below. eight days. And the figure 4 illustrates those in one
week. There three lines in figure 3. The blue and green
4. Data and results lines are same as the ones in figure 1. The blue line
shows the dynamic of the hits of “B & b”, the green
The data used in this paper is collected in one week. line is about the keyword of b and the red line is only
From June 9th to June 17th, we collect the data of the based on the “lyrics of b”.
music charts. In this paper, we use part of the data to Then we will know how to distinguish the
study. Korean people work for six days in one week. popularity of each singer. If a song is popular, not only
They only rest in every Sunday. So we just concerned the song, but also the singer will be hot. There is a
about the data of working days. The duration of data is correlation between the popularity of one singer and
from June 9th 12am to 13th 00am and 14th 00am to 17th the one of his song. From the figures we know that the
12am. The time showed in the figures is Seoul Time. b is more popular than a.
For the charts we studied is a searching charts, the
numbers of the keywords in the data is about 800. 4. 2. Correlation between the A&a and a
Those keywords conclude the name of the singer, the
name of the song, the singer with his song, especially
the lyrics of the song. We sorted the keywords into
four kinds. Also we can sort the singers into two kinds
according to the searching keywords. Parts of them are
popular for their names are searched in different forms.
Compared the rest singers are common. Table 1 and 2
show us the sorts.
Table 1. The sorts of keywords
total singers songs Singer& lyrics others Figure 1. The ranking of A&a and a in 9th
song
758 55 578 81 48 6
Table 2. The sorts of singers
Singers Popular singers Common singers
55 32 23

4.1. The basic characteristic of the ranking

According to the ranking of the music, we can get Figure 2. The ranking in one week
the popularity of each singer and song. Here are four
figures to explain it.
In the figure 1 below, we can see two lines. These
two lines show the rank of two keywords. The
keywords user searched for are the singer’s name A
(A=wonder girls) and their song a (a=so hot). The blue
line is about the “singer and song”, the other line is
about the keyword of the “song”. The data of the
figure 1 is only from 12am to 12pm in one day.
Figure 2 is also about the rank of “A & a” and a. Figure 3. The ranking of B&b and b in 9th
Figure 2 shows the ranking of keywords in one week.
The x-axis of figure 2 is from 0am to 24pm. The y-axis
of that show the dates of the data. From top to bottom,
and weekend. Among the users of the naver, there are
many users only concerning about one special singer's
rank. We call those special users fans. The age of the
special users are almost 10 to 20.

4. 4. The special singers

If a singer published a song recently, he or she


should be in the top-k ranking. If a singer has not
published new songs for a long time, he will not be in
Figure 4. The ranking in one week the top-10 of the charts. In the music charts, there is a
We can see that through the correlation between the special singer who doesn’t have new songs. But for the
ranking of “singer and song” and the one of the “song”. music charts, we only want to see the latest singer and
We only use the data of July 9th to study in this paper. song in the top-k. So this is not right for the music
The expression of the binomial correlation is: charts. Figure 5 shows this thing. The keyword in
E ( XY ) − E ( X ) E (Y ) figure 5 is the name of a very famous singer, TVXQ.
ρ ( X ,Y ) = Though he published last new song in September of
D( X ) D(Y ) 2006, his name is still in the top-k of music charts
In our study, X means the ranking of “singer and everyday. Compared with it, figure 6 to 10 show the
song”. Y means the ranking of the “song”. E(X) ranking of other five singers. We named the five
means the average rank of the “singer and song” in one singers with A, B, C, D and E. The singers A and B are
day. E(Y) means the average rank of the “song” in one the same as the ones before. C published his album in
day. D(X) shows the variance of the “singer and song”. May 22nd. E published her album just in July 14th. The
D(Y) shows the one of the “song”. Ρwill show us the figures show the ranking of the singer from July 9th to
correlation between the ranking of the “singer and 17th. The 9th’s data is on the top of the picture and the
song” and the one of the “song”. 17th’s data is on the bottom. For the 13th’s data is
Finally, we got the correlation of A and B. similar with 12th’s, we just concerned about the other
ρ(A&a,a)=0.0573. 7days.
ρ(B&b,b)=0.2975.
Obviously, ρ(B&b,b) is much bigger than ρ(A&a,a).
So the influence of A’s fans is smaller than the one of
B’s. If a singer’s song is popular, the number of that
singer’s fans will be bigger; therefore the song will be
in the top-k ranking for long time.

4. 3. The special users

In the figure 1, we can see two things. The first


thing they tell us is the popularity of the singer; the
second thing is the time that the users search for their Figure 5. Singer=T
singer. We tested from 12am to 12pm. So we can
know that the fans of “A” searched their singer from
3pm to 8pm. They rest from 5pm to 7pm.
In the figure 3, the singer’s name is B (B=MC
Mong), and his song is b (b=circus). We can also see
that the users who search for the song b are common
users. The searching time is continuously. But those
who search the “singer and song” are not common
users. Those uncommon users appeared from 2pm to
3pm and 5pm to 7pm.
We also noticed the same thing in the figure 2 and 4. Figure 6. Singer=A
The keywords of “singer and song” only disappeared
in the afternoon and Saturday. So the users who search
the “singer and song” only have time in the afternoon
5. Conclusion
The real-time charts have many advantages. They
will show the most popular songs in time. They can
transfer the newest messages to users as soon as
possible. Otherwise, the real-time charts also have
some problems. The special users will be a big
problem of the real-time charts, especially for the
common charts. They will have an effect on the results
of ranking. The affection of special users should be
eliminated or weakened.

Figure 7. Singer=B 6. Future works


Why do the users search for musical material in the
internet, and how? What kind of application do the
users like? [3] Whether the charts we used are
impersonal or not? Those things are still concerned.
We need to address a new algorithm of ranking to
avoid the infection of special users and let the music
charts to be more reliable.
On the other hand, the RSS and the news are also
popular. So the real-time charts of RSS and the news
will be popular in future. The fresh time of that and the
users' study are our future work.
Figure 8. Singer=C
7. Acknowledgements

This work is supported by:


(1) Chinese Ministry of Science and Technology “973”
(2007CB307101)
(2) The National Natural Science Foundation of China
(60772043, 60672069)
(3) Chinese Ministry of Education (20050004033)
(4) The Foundation of BJTU (2003SM017)

8. References
Figure 9. Singer=D [1]David E. Giles, “Increasing returns to information in the
U.S. popular music industry”, Econometrics Working Paper ,
EWP0510.

[2] David E. Giles, “Survival of the Hippest: Life at the Top


of the Hot 100”, Econometrics Working Paper, EWP0507.

[3] Joe Futrelle, J. Stephen Downie, “Interdisciplinary


Communities and Research Issues in Music Information
Retrieval”, 2002 IRCAM – Centre Pompidou, 2002

Figure 10. Singer=E

You might also like