Among so many text analysis techniques that are available out there, sentiment analysis is one of the most precise and accurate one—hence widely used by companies that have social media monitoring as one of their services. However, there’s an even better alternative to the traditional sentiment analysis we all know, and it is called aspect-based sentiment analysis—which Sonar also uses in our social media monitoring tools.
Also known as aspect-level sentiment analysis, feature-based sentiment analysis, or aspect sentiment analysis, this technique gives an upper hand for businesses in learning more about what customers really need by analysing their customer feedback data by categorizing them according to each aspect and identifying the sentiment attributed to each one. The fact that customers usually talk about more than one aspect of service in social media conversations is what makes it hard for machines to categorize the real sentiment behind each sentence.
For example, when a customer posts their review for a restaurant they recently visited, they would usually not only write:
“The food here tastes good.”
Instead, they will write a full-on review about their whole experience there, such as:
“I hate the fact that I have to wait in line for SO LONG just to get in. However, the food here tastes heavenly, and it makes all the waiting worth it!”
In these kinds of sentences, there can never be only one sentiment for the entire review. That is why aspect-based sentiment analysis is used to understand the context in a deeper and help find out what kind of actionable insights we can take from this customer’s review—in this case, maintaining the quality of the food while enlarging the space so more customers can get in without waiting for too long.
So, basically, aspect-based sentiment analysis is able extract two things: sentiments and aspects. Sentiment as we know it is the positive, neutral, or negative tone behind customers’ opinions regarding a particular aspect. On the other hand, aspect is the topic, category, or feature that is being talked about by customers.
To put it simply, after deciding on which aspects to analyse from customers’ reviews, we can categorize and label these reviews according to the aspect that is mentioned in the review as seen on the table below.
Case Study: Yoiko Ramen 415 Restaurant (Blok M)
Customer’s Review | Taste | Price | Ambience | Service |
Ambience resto ini biasa aja, tetapi konsep Japanese-nya masih masuk. Overall, harga dan rasa dari menu yang saya pesan di sini kurang worth it. | Negative | Negative | Neutral | |
Chicken Karage yang saya pesen enak banget, begitu pun dengan Goma Tonkotsu Ramen-nya. Suasananya berasa lagi makan di Jepang karena interiornya sangat mendukung. Sayangnya, pelayanannya kurang ramah. | Positive | Positive | Negative | |
Enak sih Tonkotsu Ramen-nya, tapi yaudah biasa aja gitu. Kalo disuruh balik lagi kayaknya aku nggak mau. Menurutku ramen dan kuahnya kurang rich. | Negative | |||
Soal rasa, nggak usah dibahas lagi. Tadi pesen Gyoza, Tonkotsu, dan Gila Ramen. All good as usual. Tapi, waiting time-nya parah. Makanan pertama baru keluar setelah nunggu 45 menit! | Positive | Negative |
Some cells that are left blank on the table above means that the customer did not mention or talk about that aspect on their reviews. This kind of categorization is the key to extracting the most important insights from customers’ reviews and decide the next strategy that can really help improve your business.
On our next blog, we will talk more about why aspect-based sentiment analysis is better than the conventional ones. So, stay tuned!
If you have any questions about how Sonar’s aspect-based sentiment analysis works and how we can help your company grow, please feel free to contact us.
