Workflow: Applying artificial neural networks to build models to analyze customer emotions based on comments and evaluation serves for determination business-related trends
- Train Data: 7,786 samples
- Dev: 1,112 samples
- Test Data: 2,224 samples
- Link: Dataset Repository
- id : the id of comment
- comment: Commentary content
- n_star: User rating for the smartphone (stars)
- data_time: Date and time the comment was posted
- label: Sentiment label of the comment
Based on the chart above, in the years from early 2017 to early 2019, customer reviews for products were quite high, averaging about 4.5. But between mid-2019 and the end of 2020, average reviews dropped alarmingly, demonstrating that customers are very dissatisfied with the quality of our products or services. It is necessary to urgently re-check product or service quality management steps to improve the situation
<
We can see that the number of Positive Reviews is always more than other categories. Another notable point is that from early 2019 to mid-2020, the number of classified ad reviews increased rapidly, meaning the number of customers skyrocketed during that time.
It is observed that users tend to use less than 40 words to rate. The number of Positive reviews is always higher than Negative, this is a good sign for the product business.
4.The heatmap chart represents the correlation matrix for the columns positive count, negative count, neutral_count, n_star
The chart shows that the positive_count value is positively correlated with n_star (correlation index 0.65), meaning that in user reviews, the higher the number of positive_count words, the higher the likelihood that that user will give a high rating. On the contrary, the negative_count value is negatively correlated with n_star (correlation index -0.69), meaning that in user reviews, the more negative_count words there are, the lower the rating will be.
