Investigating the nature of the conference
SocmedHE21 using sentiment analysis
Zara Sanni
Introduction
SocMedHE started in 2015 with the goal of using social media to open up
more higher education learning spaces online.
SocMedHE, standing for Social Media for Learning in Higher Education, is
an online workshop focused on developing a social media workshop. For this
investigation I am using twitter data from SocMedHE21 to investigate the nature
of the conference, based in part on sentiment analysis. Sentiment analysis is
also known as opinion mining, it’s a natural language processing technique used
to determine whether data is positive, negative, or neutral. It’s an artificial
intelligence which uses certain algorithms to categorise what a person may
think about a certain topic. I will be using it on the SocMedHE21 data to see
whether the conference on a whole was seen as positive or negative.
Method
The data was provided through a twitter archiving google sheet, TAGS.TAGS
is a twitter scraping tool allowing you to automatically collect tweets into a
google sheet from either specific people, hashtags, or search items.
Using TAGS, #SocMedHE21 was put
into the search bar. There was a follower minimum count of 0 so we could
observe as much data as possible and the maximum number of tweets being 3000,
between a time range of December 6th, 2021 and January 24th,
2022. The number of tweets using the SocMedHE hashtag was 419 and the number of
unique tweets was 377, which is very close to the actual number of tweets which
indicates to some sort of conversation and discussion happening between twitter
users.
TAGSExplorer is an interactive version of the twitter conversations
gathered where graphing is used to visually present the data. Using this I was
able to see all the connections between the twitter users. The picture shows
the names of all the users involved in the SocMedHE22 conversation; the bigger
the name, the more connections and conversations the person had with other
people. When you click on a person's username you can see all the tweets,
they’ve made in regards to SocMedHE21 and their replies and mentions. Here you
can see that the top conversationalist isn’t the official SocMedHE account that
I expected, instead it was one of the conference attendees.
Using the
TAGSExplorer arch, I was able to look at the activity of users between the
dates given above and found, as expected, a lot of discourse in December, on December 14th and the few days following but there did
continue to be conversation happening some time after. There were at least a couple tweets fortnightly on twitter most
of which were for a particular person or from a bot (an autonomous program on
the internet or a network that can interact with systems or users) showing
results from an weekly run of another social media analysis tool NodeXL. What
was of interest was that the people named in the tweet were happy to retweet
that tweet. I do think there is a positive correlation between twitter likes
and mentions, specifically with, but not limited to, the official SocMedHE
account. as most of the conference attendees would be following it any
interaction with the account is more likely to be seen especially because the
community involved in this conference are such conversationalists and the
official account itself does also interact and retweet with the community.
Results
LTHEchat is a weekly learning and teaching in higher education chat
where brainstorming in virtual learning spaces allows for conversations and the
sharing of ideas to individuals in the education sector. Many of the LTHEchat
community attended the SocMedHE conference so, using the words collected I
chose the first 150 as they have all been used at least 10 times, making sure I
factored in for spelling mistakes, adding misspelt words of the same nature
together. Using sentiment analysis, I labelled the words either neutral
positive or negative; for example, names of people have been labelled as
neutral but words such as creative, change, and better positive. I didn’t find
any words to label as negative which shows the nature of the conversation to be
beneficial and enthusiastic of the topics discussed in the conference and what
is to come of it. Of the 2992 words sampled, I marked 1152 of them as positive
which counts to 38.5%. after doing this I converted the data into a CSV file
and put it into wordart to create a word cloud with the most used words. I gave
positive words the colour green and the neutral words I made black.
After using
sentiment analysis to look at the data, I can conclude that the nature of the
SocMedHE conference to be a positive one in which people are conversing in a
productive way, carrying on conversations months after the conference which
shows …
Next, to
continue and further this research I think we should look at alternative
software to sentiment analysis to analyse the data and see if the results are
still the same. For example, Amazon Comprehend is a text analysis software that
users can use to ‘find insights and relationships in text”, it identifies key
phrases, people, events etc and can determine whether it is positive or negative.
This is a much more thorough and in-depth way to analyse the data and will also
allow us to see things we may have missed due to limiting factors.
Bibliography
https://blogs.shu.ac.uk/socmedhe/16-with-a-little-help-from-my-followers-facilitating-the-lthechat/
https://www.g2.com/products/sas-sentiment-analysis/competitors/alternatives
https://www.sciencedirect.com/science/article/pii/S2090447914000550
No comments:
Post a Comment