A Q&A with a UCIPT Postdoctoral Researcher
Using sentiment analysis, are you able to accurately predict the fashion preferences of UCLA students?
We have not tried to predict fashion preferences for UCLA students. To do so, we would need data from multiple sources (e.g., Pinterest, Instagram) in order to build an accurate prediction model. A person’s fashion preferences can be influenced by many things, such as their social circle, job, or hobbies. Therefore, in order to build a fashion preference prediction model, we would need to get data that are able to gain insights into these areas. Twitter data and sentiment analysis is a good start, but we probably need more than just sentiment analysis to make an accurate prediction model. We would need to look at the person’s social network to see what things that person is interested in (e.g., movies, hobbies, pop culture). Moreover, we would have to conduct topic modeling combined with sentiment analysis to understand how the person feels about certain celebrities and music genres. Overall, I think it is possible to predict someone’s fashion preferences, but it would require multiple data sources and analysis techniques.
Based on Twitter analysis, what are the most positive and negative sentiment brands at UCLA?
We have not examined positive and negative brands at UCLA. We could do this if we had a list of brand hashtags popular with UCLA students and faculty, and then conduct a sentiment analysis to figure out what people think about the brands. Here is a list of the top 10 brands on Twitter this year. You can also find out about the top brands in different countries at this website.
What are successful brands doing to attract positive sentiment in the UCLA Twitterverse?
I think there are several methods that successful brands are using to attract positive customer experience:
- Connecting with users: On Twitter, brands can connect with consumers in new and more direct ways than ever before. For example, brands can tweet at customers and answer their questions and concerns directly. The company can engage with users in their tone of voice or preferred styles and personalize the message based on the user’s interests and needs.
- Influencer marketing: A lot of brands are collaborating with new digital creators (e.g., Youtube stars, Twitter influencers) to market their brand.
- Real-time marketing: a great example of this is when the power went out during Super Bowl XLVII and Oreo’s famous tweet sparked the age of real-time marketing. Real-time marketing enables a brand to connect their brand to the culture at large as well as to a particular moment at scale.
These are just some of the new creative possibilities that successful brands are using to ensure a positive customer experience. You can read about other methods here.
Can you predict if a specific movie will be popular based on Twitter analysis? For example, have you compared weekend box office income with Twitter sentiment for a specific film opening?
There is a recent study by Krushikanth R. Apala that was published at the ACM International Conference on Advances in Social Networks Analysis and Mining that attempted to answer this question. Researchers collected data from multiple social media and web sources including Twitter, YouTube, and the IMDb movie database. The prediction model was built based on decision factors derived from a historical movie database, follower counts from Twitter, and a sentiment analysis of YouTube viewer comments. You can read about how predictive data modeling works in earlier Q&As with my colleagues Jin Yu and Wenchao Yu. Based on the models presented by Dr. Apala, he identified the following patterns: 1) the popularity of a leading actress is crucial to the success of a movie; 2) the combination of past successful genre and a sequel movie is another pattern for success; 3) a new movie in an unpopular genre and an actor with low popularity could be a pattern for a flop. The authors have plans to validate their model in a follow-up study in the near future.
What does research show it takes to convince a person to go from expressing a positive comment about a movie or product on Twitter to acting on it in the real world, such as purchasing a ticket?
There are several psychological theories related to persuasion, such as the amplification hypothesis, information manipulation theory, priming, the scarcity principle, and the sleeper effect. We see these techniques used in advertising everyday, including on Twitter. Based on these theories, I think in order to change someone’s mind, the following strategies can be useful:
- The number: The more people that try to persuade the original poster, the greater the chance of someone being influenced.
- Timing: individuals who write back first to a post are more likely to persuade the original poster than those who write back later.
- Language use: There are some research studies showing that longer replies tend to be more convincing, as do arguments that use calmer language. Using specific examples is also useful. Sometimes using plural pronouns (“we”) may be more convincing than first- person pronouns (“I”). (First-person pronouns can indicate an opinion is malleable.)
You can find out more about the psychological strategies to change someone’s mind here.