Skip to content

Latest commit

 

History

History
56 lines (44 loc) · 6.78 KB

guestbook_michael.md

File metadata and controls

56 lines (44 loc) · 6.78 KB

Ahoy! Welcome to Michael's term-project guestbook. Feel free to have a look at my code and analysis and leave a brief review.

Section 1

visit (Frances):

  • One strength is your code. It is very neat and easy to follow along with, as well as effective considering your data output.
  • One thing to improve is scraping reviews from more books.
  • One interesting thing that you pointed out in your code is how Goodreads labels their reviews not numerically, but with certain phrases in their html.

Hi Frances, thanks for your input! Goodreads in general is a very strange website with some... interesting design choices. I talk a little more about it in my updated JNB. I have also increased the amount of books scraped from 2 to 32.

visit (Jordan):

  • Seems like you have a good base for your project here, I really liked the multi-indexing you did to keep reviews of one book together, and you seem to have done a good job web scraping.
  • My only concern would be that Na-Rae might want more out of this project besides just creating a sentiment analysis model. Maybe you could do some more analysis by genre or by the age, location, etc. of the reviewer
  • Multi-indexing looks very useful for my project, I'm gonna try implementing it.

Thanks for your feedback Jordan! I have expanded my dataframe somewhat to include more data for analysis (such as year of publication, genre, likes, etcetera). As for multiindexing, it is a very useful tool, but I actually ended up removing it from my data cleanup process because it just made it difficult to isolate certain review scores.

visit (Lexy):

  • The basis of your project is very interesting. I was always interested in the reviews on Goodreads because they vary so much. Your code and initial analysis also very easy to follow! I understood what you were doing at every step. Your project is also very well thought out from viewing the project plan.
  • I am concerned that you may not have enough reviews to make a comprehensive sentiment analysis. You may want to include another analysis because this project is just building a model.
  • I learned that Goodreads labeled in html which is strange but I also don't know much about these things.

Hi Lexy, thanks for your feeback! I have indeed expanded my data to include more reviews (around 3000 now) and will increase the scope of my final analysis to incorporate things such as genre and review-popularity.

Section 2

visit (Emma):

  • Your analysis notebook was super easy to follow, especially thanks to the opening commentary. I also really liked how you showed examples of the data frame, it really gave a good perspective into the data.
  • I’m glad you touched on the 2,3, and 4 star reviews in terms of the negative and positive reviews, and it’ll be interesting to see how you take that further! (If 3 is truly a more neutral review, why does the classifier decide they are negative or positive?)
  • Sentiment analysis!!! I’m considering doing something similar for my own project, so this was great to see how someone else is going about that!!!

Hi Emma, thanks for your feedback! I have that same question about why 3-star reviews are classified the way they are and I'll be taking a closer look at them soon. Stay tuned!

Section 3

visit (Emily)

  • I like the organization a lot, and I think the way you built your dataframes is informative and clear; as is the analysis section. I agree that the opening commentary was helpful and set the reader up to understand what you did.
  • I wonder if your method for classifying sentiment the way you did is the best approach. I say that without quite enough knowledge to make it a full critique but I am curious.
  • Wow, your scraping script looks like it was super labor intensive. As someone who also had to scrape a lot of data but used a different method I learned a lot from your approach.

Thanks for the review Emily! Indeed, my approach to classification at the moment is very quick and dirty. I plan to update that section extensively for the next progress report, incorporating some of the techniques we have learned in class over the last couple of weeks.

visit (Abby):

  • Thank you for the extremely reader-friendly Jupyter Notebook and analysis. I'm interested to see what the results are -- I've heard it said before that good reviews generally focus on the actual content of the thing (i.e. "In this book, character 1 did this thing in this place", "This restaurant served this dish and had such-and-such an atmosphere") while negative reviews focus on personal experience (i.e. "I read this book on a Greyhound bus to such-and-such a city during the Fall of 2018", "We stayed in this hotel for a destination wedding for my aunt's brother's coworker's son and we were really looking forward to it"). This project has the potential to prove or disprove whether or not that's true.
  • I honestly don't have any critiques about the overall project. I'm curious to see what will happen when you incorporate non-textual features in the analysis and whether or not they will improve accuracy.
  • I didn't realize you could extract the NB's most useful features with vocabulary_.keys() - that will be helpful. I also learned that Twilight is classified as a fantasy series . . .

Thanks Abby. Glad to hear you liked it! I'm very intrigued by this idea of negative reviews focusing more on personal experience. It makes a ton of sense so I'll be sure to take a look at the data and see what I can find! (As for Twilight: don't forget that old vampire stores like Dracula paved the way for the development of the modern fantasy genre. The two go hand-in-hand!)

visit (Emma pt.2):

  • I really liked seeing how your analysis since my last visit! I also seeing how thorough you were with the grid searching and the multiple ml models. That “yeehaw” comment also resonated quite a bit.
  • Your analysis looks really interesting, and I’m excited to hear more about about the rest of your To-Dos (especially the differences between positive and negative reviews)!
  • from string import punctuation !!!

visit (Sonia)

  • The code is extremely readable, it's clear you formatted things with the reader in mind. When I clean code I tend to go more for what's readable to me then what's readable to some other person, so good job with that.
  • No real critiques, I'm just curious what originally gave you the idea to do this project.
  • It was interesting to see how the repository looks when you remove a part of the commit history, really seems like it's not possible to tell that that happened