Dear Srilekha, Thank you for the update and for sharing your approach. Using a local model for mood detection makes sense, especially given the API constraints. The *"facebook/bart-large-mnli"* model seems like a solid choice for emotion classification. For song recommendation, *RecBole, RecCobeats, and LightFM* each have their strengths. If scalability and flexibility are key, *RecBole* might be a strong candidate due to its broad support for different recommendation algorithms. However, if we prioritize music-specific embeddings, *RecCobeats* could be worth further exploration. We would advice to make a decision now and maybe refine it when you make more progress in the actual implementation. Thank you! Best, Giannis Prokopiou & Thanos Aidinis Στις Σάβ 29 Μαρ 2025 στις 5:56 π.μ., ο/η Srilekha <kslekha02 [ at ] gmail [ dot ] com> έγραψε: > Dear *Giannis Prokopiou *and *Thanos Aidinis*, > > I hope you both are doing well. I apologize for mailing you multiple > times, but I wanted to share my thoughts on our current approach and seek > your guidance. I’ve been working on the mood detection model and opted for > a local approach instead of using an API. Considering the rate limits and > potential costs associated with APIs—especially since we’re already using > one and might need another for song recommendation—I felt that managing > everything locally would be more efficient. > > Currently, I'm using a model based on the *"**facebook/bart-large-mnli* > *"* pipeline for mood detection. By mood, I mean identifying emotions > like *sad*, *happy*, *angry*, etc from the user’s input. As for intent > detection, we aim to classify whether the input is a *mood expression*, *direct > song request*, or *mixed*. For instance, if a user says, *"I’m feeling > low, play something upbeat,"* it would be classified as *mixed*, while *"Play > something by The Weeknd"* would be a *direct request*. > > For the song recommendation model, I’m currently exploring options like *RecBole, > Reccobeats* and *LightFM*. Since scalability and data availability are > key considerations, I would greatly value your input on which of these > models—or any other alternative—you think would be more suitable for our > use case. I’m confident that I can do more research on this, but I’ve > already spent significant time on this, and I don’t want to waste any more > if it’s not the right path altogether. > > Your thoughts and suggestions would really help us make informed > decisions. Thank you once again for your guidance and support! > > Best regards, > Sai Sri Lekha > kslekha02 [ at ] gmail [ dot ] com > ---- > Λαμβάνετε αυτό το μήνυμα απο την λίστα: Λίστα αλληλογραφίας και συζητήσεων > που απευθύνεται σε φοιτητές developers \& mentors έργων του Google Summer > of Code - A discussion list for student developers and mentors of Google > Summer of Code projects., > https://lists.ellak.gr/gsoc-developers/listinfo.html > Μπορείτε να απεγγραφείτε από τη λίστα στέλνοντας κενό μήνυμα ηλ. > ταχυδρομείου στη διεύθυνση <gsoc-developers+unsubscribe [ at ] ellak [ dot ] gr>. >
---- Λαμβάνετε αυτό το μήνυμα απο την λίστα: Λίστα αλληλογραφίας και συζητήσεων που απευθύνεται σε φοιτητές developers \& mentors έργων του Google Summer of Code - A discussion list for student developers and mentors of Google Summer of Code projects., https://lists.ellak.gr/gsoc-developers/listinfo.html Μπορείτε να απεγγραφείτε από τη λίστα στέλνοντας κενό μήνυμα ηλ. ταχυδρομείου στη διεύθυνση <gsoc-developers+unsubscribe [ at ] ellak [ dot ] gr>.