BookHive

#

Presentation #

Week 1 report #

This week, our main task was to form a balanced all-around team and choose the general theme and outline of our project. Creating a team was not hard, as we mostly already knew each other and our strengths. Below is the list of our members and the roles they can fill out during project development:

As you can see, we have an assortment of stack variations, and we believe that this will be an advantage during development.

While choosing the project idea, basically everyone had suggestions: we thought about making a site for prediction of currency exchange rate, considered making site+telegram bot for prediction of prices on online shops such as WildBerries. At some point, we even considered making a site for automatic note creation from the user uploaded mp3 file. In the end, we settled on something that in our opinion has a higher chance of being a useful project: a book recommendation site (web/ML project). After finishing a book they really liked, many people have problems finding a similar one, and we aim to aid in that. Here is the brief rundown of the main features of this project: The user enters a book they finished reading and whether they liked it or not. Based on that, our site proposes books that the user has not read and that might be interesting for them using Machine Learning Algorithm. Each book will contain genre, abstract, and links to its page in several online bookshops: both in digital and on paper. Moreover, we plan to have some open source books be available for download directly from our site. Also hope that if this project will work out to get rights to sell books there. More features may be added further down the line, when it will be easier to assess how much time we have after implementing the main ones, among those we consider implementing book crossing, however we first will measure the need for it to decide whether it is worth implementing or not.

To summarize, the problem we are solving is the difficulty to find similar books. The target audience for our project is people who read books. To test our assumptions, we are planning to try this on ourselves and see whether proposed books will be a good fit for us. The metrics for our project will be the amount of site users and their satisfaction with the recommendations. It is currently hard to make plans how to iterate our project, as this will mostly depend on the feedback we will be getting.

The book recommendation algorithm will be AI-based, the GitHub repository will be most likely open, and since we have a Machine Learning algorithm we will have two dedicated ML developers in our team. We are confident that we can complete the project by ourselves without other teams’ help.

Feedback

Hi, I think this is a great project and idea! Since this course aims to finish the semester with an MVP - minimal viable product, consider scaling your project down to creating a web app that does core/main things. Outline main features of your project and focus on execution. I would also think about which particular parts you can strip down - for example, UI can be very simple (I am saying all this because we have only 7 weeks).

One website that I actually use is called goodreads.com - it’s a book recommendation platform which is also a social network of the sort. It helps people to form communities and discuss books. However, this platform mainly oriented towards English-speaking users. Perhaps, your team can take some good ideas/features and replicate them in your website. You can think about parsing their book scores as an initial data for the recommendation service.

Overall, I think this is a very good project idea, that now requires a well-planned execution. Therefore, plan for simplified version that showcases main features.

Week 2 report #

Architecture Design

1. Component Breakdown:

   • Create UI/UX for site
   • Create login/welcome page
   • Create recommendation page
   • Create account information page
   • Design databases for users and books
   • Create starting version of recommendation algorithm
   • Add links to online bookshop for the recommended books (this is not a high priority task, and generally we are not sure whether we can do it without using web scraping)
   • If we have enough time, create some book compilations based on genre

2. Data Management: We are planning to have a database for storing tables about books and users, they most likely will be PostgreSQL, MySQL or SQLite. We are not sure, as we have not decided where the database will be located.

3. User Interface Design: While most of the design is set in stone, it is still subject to some minor changes. Link to Figma Model Screenshots to show progress at the time of report submission:

4. Integration and APIs: We are planning to utilize Google Books API to get important information about books: annotations, genres, etc. This data will be used for recommendation algorithm and will also be displayed together with book’s cover on the site

5. Scalability and Performance: Since we are still in brainstorming stage of our project, we can’t really predict its performance, but we hope that the expected time of our site’s response time will be not higher than a couple of minutes

6. Security and Privacy: Once we choose the database language, we will be able to provide more accurate information about what protection of database we will utilize to avoid malicious actions.

7. Error Handling and Resilience: The lowest bar for us will be that all occurring errors must be handled, or at least not interfere with the product use.

8. Deployment and DevOps:We are yet to decide where we will deploy our product, since there are currently few free deployment options.

Week 2 questionnaire:

1. Tech Stack Resources: Are you utilizing any project-based books that specifically cover your tech stack and help you build your project?

   • Answer: We do not have plans to use any books, as none of us have read books, which would be particularly helpful. And we most likely won’t have enough time to read any book in full, since we have only 6 weeks remaining.

2. Mentorship Support: Do you currently have a mentor actively involved in your project?

   • Answer: Our team does not have a mentor, and we are not looking for one, as we are confident we will be able to deliver the MVP at the end by ourselves.

3. Exploring Alternative Resources: In addition to project-based books, what other resources have you explored to expand your understanding of your tech stack?

   • Answer: our ML developers use Kaggle to learn more about it. One of the particularly useful videos on our topic. Frontend developers read React documentation.

4. Identifying Knowledge Gaps: Are there any specific areas within your tech stack where you or your team feel there are knowledge gaps or expertise is lacking?

   • Answer: We might have some problems connecting ML to backend, as it is not something our ML developers have a lot of experience in. But our team should be able to find out how to do it using open source tutorials and documentations.

5. Engaging with the Tech Community: Have you actively engaged with the broader tech community to seek guidance or learn from experienced professionals in your tech stack?

   • Answer: One of our members attended Innopolis meetup regarding Data Science and Machine Learning recently and shared information he learned there with the rest of the team.

6. Learning Objectives: What specific learning objectives have you set for yourself and your team in relation to your tech stack this week?

   • Answer: All of us had an Introduction to Machine Learning course, where we performed most of the tasks by the steps given by instructors. However now the development process is basically unsupervised, and we want to test ourselves in fully organizing project workflow, while also picking up knowledge in related fields.

7. Sharing Knowledge with Peers: How have you been sharing your knowledge and expertise with your teammates?

   • Answer: We had a number of meetups/brainstorms to make sure that everyone had time to speak their mind, and we were all on the same page regarding the vision of our project.

8. How have you leveraged AI to compensate for any lacking expertise in your tech stack?

   • Answer: We picked some inspiration from AI generated content, while picking logo and name for our product, so the AI involvement was rather limited so far.

Tech Stack and Team Allocation:

• We will code Backend and Machine Learning parts on Python and Frontend on React

Anastasia Martynova (Team Lead) – Designer/Frontend

Maxim Piniagin – Product manager/Backend

Yegor Sokolov – Backend

Marat Medvedev – ML/Backend

Lev Gorbunkov – Frontend

Bulat Kutlugallyamov – Backend/Frontend

Roman Voronov - ML/Backend

Week 2 Feedback by Rustam
This report is well written and structured, and it’s a pleasure to see how your project takes shape. I liked the figma UI and overall description of the project. Seems to me, you have fully aligned team that can effectively deliver results.
Keep up the pace and make sure to put an emphasis on the execution. Overall - good project and report.
5/5

Week 3 report #

1. Prototype Features:

   • Link to GitHub repository.
   • Since this is our first week of working directly on the code of our site, not the entirety of functionality has been implemented yet.
   • Current functionality. Below are also the images of work in progress pages, you can view planned full design in week 2 report:
     • User is asked to either to log in or register:
     • User is then asked to enter 5 books and rate them
     • In personal cabinet, one can view their username and status which is based on amount of rated books, change password and log out.
   • Technical details:
     • All information about users, books, and ratings is stored in a locally deployed PostgreSQL database, which was created based on this dataset and was extended with book genres and annotations using Google Books API.

2. Challenges and Solutions: During the development, we faced a number of problems:

   1. Our ML developers had a lot of issues with preprocessing, which severely slowed them down:
      • The dataset was pretty raw: some books had no ratings and there were ratings of books, which had no information on them in other tables. Solution was to just remove such books.
      • For Matrix Factorization, books’ IDs had to be normalized from ISBNs, and this was quite difficult on Pandas. The solution was to use PySpark, which uses SQL, however further data augmentation was also hard due to little experience with PySpark. Because of that, after PySpark, data was converted back to Pandas.
      • All these problems took a lot of time for our ML developers, so ML part is not quite ready for the first prototype.
   2. Due to some books in dataset being extremely obscure, Google Books API did not have its annotations and/or genres, we are still working on solution of this issue, for now they just can’t be recommended.
   3. Because of imprecise workload estimation on Frontend side, we did not have enough time to create recommendation page.

3. Next Steps:

   • ML team still has a lot to work on for the next weeks, as the recommendation model is planned to be implemented using users’ ratings from Dataset.
   • Frontend team will have to finish recommendation page structure and change all pages to look like their planned Figma design.
   • Also, right now, our system lacks the function to save recommended to books into want to read tab for user.

Week 4 report #

1. External Feedback

   • This week we conducted 5 interviews from which we gained a lot of useful feedback which we will surely take into account when iterating our product.
   • All of the interviewees were students of Innopolis University of approximately the same age as us.
   • Here are the bullet points from all interviews combined:
     1. When searching for books to rate, sometimes it would show several copies of the same book as different books. We suspect that the reason for it is that database has the same book published by different publishers, resulting in being viewed as different books.
     2. Currently, implemented design is a little empty and some colors are off. However, we have not fully implemented the planned design and when the full design in Figma was shown to interviewees, every one of them agreed that it will look better.
     3. In the current iteration of design, the search bar on the starting screen is too low and because of that some search results are outside the screen. This will absolutely be fixed, as the final planned design will have the search bar located higher.
     4. When searching for books to rate, there is no animation for search progress. So the user does not know whether the system is still looking for the book in the base, or it is finished and there is just no such books in it.
     5. There should be some indications whether book ratings are saved or not.
     6. Notifications regarding registration and login do not look good according to some interviewees.
     7. A lot of books are missing from the dataset, making it hard for some interviewees to find their favorite books.
     8. Unexpectedly for us, there is demand for search by author name, not just by book name. We will try to implement it before the final presentaion.
     9. Some books have broken links to their cover images, resulting in being replaced by question marks. This does not look good in combination with the fact that we haven’t added the book names and annotations to the recommendations page.
     10. Change password padding overlays the status, so by clicking status you can change password.
     11. Some interviewees wanted options to delete some of their ratings to not affect recommendations and to delete account altogether.
     12. 20 books are too many for users to rate. Right now we set the threshold to 3, but we will continue observations and tests to find the optimal number of ratings.

2. Testing

   • The best way of testing is of course using the product extensively: we had every member of our team and 5 interviewees test the product multiple times, as another can person can find issues that others might never think about.
   • Testing procedure:
     1. Register as a new user
     2. Login
     3. Rate 3–6 books
     4. View recommendations,
     5. Check how close they are to rated books
     6. Check user information
     7. Sometimes change password
     8. Repeat
   • Most of the technical issues mentioned above were found this way.

3. Iteration

This week we improved our UI, finished the recommendation model, collected feedback and patched some bugs. As of now we are looking into ways of solving important issues such as lack of annotations and names of books on recommendations page, unpredictability of book cover existences, finishing the planned design.

Below are screenshots of the current site iteration to show the progress we have made:

Feedback by Rustam
I’ve read through week 3 and week 4 progress reports and so far looks good. It is nice to see that you and your team have accomplished a lot during this weeks. Make sure to keep the pace and move fast.

Overall, good reports and progress. 5/5 for both weeks

Week 5 report #

To begin with, the target audience of our project is reading students and adults. In terms of feedback collection, we showed an updated version of our product to some of the people who gave us feedback last week, and they were satisfied with the changes. We also conducted 3 more one on one interviews with other Innopolis students, and converted their feedback into a bullet point list of issues, that were added to issues on GitHub. Most of the interviewees pointed out the same issues as the previous ones, that we just didn’t have time to fix yet, but there were some new ones:

1. On Want to Read page, there is no way to remove books from the list.
2. On Want to Read page, there is information about books other than their covers, we should add at least book names and annotations.
3. Some system alerts have typos in them.
4. Some interviewees didn’t like the bright honey-themed color palette, however, we will not change it, as it is a part of our vision.

Overall, the customers liked the design direction and primary functionality, but generally wanted more little features. We did not make any online forms for feedback collection, as we believe that one on one interviews are more effective, generate more feedback and are more personal. Moreover, we have not deployed our product yet, as we are still working on it, so the online form would at best provide screenshots and descriptions of the app flow, which is not as good as a live demo. We also tried getting feedback from ChatGPT, but it was hard to explain our app in such a way, that would result in AI giving any useful feedback.

This week, we implemented a primary functionality that wasn’t done yet - want to read list. We also made some tweaks to UI and fixed broken links to some books’ covers. We are not planning to expand our product after the final presentation, so we will not be making a product roadmap. Until the final presentation, we will be working on fixing the problems of the product, such as lack of information on Want to Read page, and finalizing the UI. After that is done, if we have time, we will be working on adding extra features proposed during interviews, such as a separate page for more information about books, and a functionality for deleting ratings and the account itself.

Alright, good to see the convergence of the feedback - so you know exactly what to improve. I personally liked the bee related design - it’s pretty. Noted the final presentation on the top of the file. 5 the week! And good luck with the project and the final presentation!