- Conclusion -

* * *

Amidst a competitive academic landscape many young adults carry the misconception that big-name, ivy league universities are the best option. However, in reality the most known school is not always the best option on both a personal as well as an academic level. Instead, individuals should seek schools that are particularly excellent in their area of study or interests. Hopefully, by taking a personal-preference-first approach to college search, individuals can narrow their search and set their sites on achievable, personalized applications

Next Steps & Considerations

In the end, the combination of a sentiment analysis, net work, and a summarization model provides users with concise, easy to digest insight into the first-hand experience of students at universities of interest. While the concept of splitting reviews by sentiment prior to summarization, theoretically groups, the overarching pros and cons, the nature of the data added complexity to this task. The evaluation ratings used as labels to train. The model aren’t perfect indications of the reviews sentiment. In many cases, the combined scoring resulting from a students satisfaction with aspects, such as campus food, facilities, Internet, and clubs didn’t provide direct insight into the students overall satisfaction with the University. In fact, it is likely that students took to the open comment section to share opinions not covered in the reading categories. This inconsistency in the numeric rating and the textual review made it hard for the model to learn consistent patterns.

Additionally, very few students feel entirely one way about their school. In a review, students are likely to share both pros and cons of their experience. For this reason, the task of qualifying reviews has either positive or negative can be inherently flawed. Perhaps an improved method could include a third category of “neutral.” Similarly, in a more advanced system, the trained model could be used to classify, smaller segments of individual reviews. In this way, an advanced model could determine the positive comments from the negative comments on a deeper level within the reviews. This would result in a cleaner distinction between the two resulting summaries.