You are on page 1of 5

Semantic Network

Analysis - #Music
TERM 4

PART C

Kanika Mohan
PGP/24/038
Word and Word Pairs – Grouping and Graph Metrics

Post running the graph metrics for words and word pairs on the corpus of the downloaded Twitter
data, the aim was to analyse and understand the way different word pairs are grouped, for which
Grouping by Clustering mechanism was used.

The following Graph Metrics were obtained from the same:

Graph Metric Value


Graph Type Undirected
   
Vertices 2228
   
Unique Edges 2935
Edges With Duplicates 132
Total Edges 3067
   
Self-Loops 8
Connected Components 13
Single-Vertex Connected Components 1
Maximum Vertices in a Connected Component 2204
Maximum Edges in a Connected Component 3053
   
Maximum Geodesic Distance (Diameter) 28
Average Geodesic Distance 5.618577
   
Graph Density 0.001206428
Modularity 0.689402

The vertices were then changed to represent the words they stood for and the edge thickness and
vertex thickness was set to the value of Count to get a clear picture of the importance of different
words in the conversations based on which other words they are connected with and their number
of occurrences in the Twitter data about #Music.

-There are 2935 unique edges and 2228 vertices.

-There is only one single vertex connected component and total connected components are 13.

-The maximum geodesic distance, or the largest possible shortest path between two nodes is 28,
while average distance between them is 5.61.

-The modularity value is towards the higher side, which means that there is a community structure.

-The graph density is low, meaning lesser values of possible permutations of ties have been realised
than what exist.

The graph images for all the groups, the major individual groups, and some of the important
combinations of groups have been provided below and commented on.

1. Overall graph
Here, as can be seen, some of the words with higher count are appearing in different groups but are
paired with each other.

2. Group 1
-The central idea here is the word Music itself, which is paired with lot of other music related
terminology, such as ‘listen’, ‘singer’, ‘artist’, ‘songwriter’, etc. names of different music genres such
as rock, pop, blues, house, trap, etc. and even other words which depict the emotions of these users
associated with music, such as ‘enjoy’.

-Names of various artists and musicians have also been tied to the word music, which is a great way
for Twitter users to connect with not just their favorite artists, but also people across the world who
listen to and love the same type of music.

-Brands associated with the Music industry, such as Spotify, Reverbnation, Beats, YouTube, etc. are
also present in this group, which showcases that these brands use the music hashtag to reach out to
their potential and existing customers who will be interested in music-related content.

3. Groups 1, 5, 9

This is a very interesting combination of groups which are interconnected and provide some great
insights.

-While the word Band in G9 is connected with words like music, rehearsals, and track, which refer to
music terminology, it is also connected with words relating to Dogecoin and cryptocurrency, owing
to the user Rockstardoge, who has started an influential conversation about music crossed with
cryptocurrency.

-Similarly, the word ‘track’ is paired with music, which refers to music tracks recorded or played, and
various other relevant music terminology relating to playing or recording of music.

-This graph provides various interesting insights which are discussed in the PPT.
The following are the group metrics for the top 10 groups:

Maximum
Geodesic Average
Unique Total Self- Distance Geodesic Graph
Group Vertices Edges Edges Loops (Diameter) Distance Density
G1 333 525 595 2 7 3.046 0.010
G2 285 371 389 2 10 4.088 0.009
G3 127 133 137 1 17 4.765 0.017
G4 71 72 74 0 24 8.420 0.029
G5 68 72 74 1 17 5.737 0.032
G6 68 68 68 0 19 6.643 0.030
G7 65 64 64 0 20 7.690 0.031
G8 62 61 61 0 25 8.136 0.032
G9 55 56 56 0 17 6.696 0.038
G10 51 51 51 0 22 7.260 0.040

-The graph density is increasing as the number of edges decreases, as the denominator of possible
combinations decreases, the ratio of combinations realized v/s those possible will increase.

-The highest average geodesic distance is for Group 4 while the least is for Group 1.

You might also like