Well, it's with deep regret that I post my last class related message :-( I must admit that I'm going to miss sharing my thoughts on the net...i've never been much of a "blogger" but there is a certain intriguing freedom about discussing in this manner that I rather enjoy.
Throughout the course of the semester, I managed to listen to an impressive 341 songs. Well, maybe its not so impressive but I was surprised to learn the count was that high. Of the 341, there were 221 that I overwhemingly enjoyed, 2 of them I used to like but Ive heard one too many times and 17 of them I liked somewhat or were at least tolerable. Aggregating a positive pile of counts, 240 of the 341 were positive (loosely defined in some areas but positive nonetheless) providing a 70% rating.
In addition, I only created 6 playlists during the semester. Five of which were artists and 1 was based on a song title. The Mary J Blige playlists, which is an artist, boasted a 82.96% success ratio. I listened to a total of 135 songs on that particular playlist and throughout the testing period I felt this station improve and as a result I listened to it more often because it seemed more reliable. I also had a John Legend, another artist, playlist and it generated a 73.33% success rate after listening to 90 songs. I tried creating a playlist based on a song title rather than an artist but it only generated a 36.36% success rating after 11 songs and I quickly lost trust in the recommendations provided.
During each session, I captured the day, time of day and mood that I was in. I realized that mood is an extremely important factor but rather hard to qualify. For instance, I initially detailed combined mood characteristics such as Awake, Eager to Work; Content, Peaceful; Inspired, Motivated; Upbeat-Energetic, Winding Down, Relaxed, etc... When attempting to perform the analysis its rather hard to qualify exactly how I felt from the combined descriptions and it would require a large matrix to even attempt to quantify where each rating belongs. With that said, I condensed the emotion or mood previously captured to a one word description in hopes of performing some type of analysis. Interestingly enough, when I was extremely tired I listened to 16 songs with a 100% success rate and when I was "pleasant" I listened to 6 songs with a 100% success rate. The two seem to be on opposite ends of the spectrum so perhaps because of the state I was in, I was either less critical or more open to accepting the songs that were played. When I was energetic and eager to work the success rate was between 46% and 58%. When I was motivated and content the success rate was between 67% and 76%. When I was relaxed, I did obtain a 89.77% success rate which is pretty good. These numbers show that the more focused or energetic I was, the lower the success rate....probably because I preconceived notions about the type of music I wanted to hear and not having those songs played probably didn't bode well for my listening experience!! Not hearing certain songs made me more critical of those I heard.
Towards the end of the semester, I listened primarily to the Mary J Blige playlist because it seemed to offer more accurate recommendations. Looking backing on its progress throughout the semester, when I started providing feedback the songs played were hit or miss. However, towards the latter part of the course, the suggestions became better and better. Again, mood played a critical part in my listening experience and if there is a way to account for this facet I think my sessions would improve drastically. On the other hand, I realize how difficult and challenging it is to capture "moods". It is extremely subjective and will require a great deal of knowledge engineering to best assign mood labels to songs as it is user dependent. Songs that may relax me may depress or excite others.....and vice versa.
Overall, removing the mood element, Pandora performs exceptionally!! I have recommended it to several friends and co-workers and absolutely plan on using it in the future! Pandora thanks for all you do!! It's been a trying semester at times and thanks for bearing with my and my "mood swings". I've enjoyed it ;-)
Sincerely yours!
Remi
Monday, May 5, 2008
Thursday, April 10, 2008
April 9-10, 2008 Analysis - My New BFF??
WOW!!!!
WOW!!!
WOW!!!!
So throughout the course of the semester I've been looking for the "PERFECT" listening session. I can finally report to you all that it happened!! Each and every song that played is a song that I enjoy!! Track after track, artist after artist...it was great!! I started at 11:04pm and my mood was exceptionally tired, its been a loooong week :-( Being candid, I listened more out of obligation than desire. Often times, I listen to music while I work but last night I just didnt want to hear anything but I did and to my amazement, I stumbled upon the perfect session!! Though I didn't hear any new material, I did hear tracks that I havent played in a long time and that was enticing! For a split second on way of my old fav's I found my energy swing slightly upward and thats something that I long for on those late working nights!! While I didnt listen for hours on end as I normally do, I did make it through 22 tracks and every single one of them were perfect! The interesting thing though is that I didnt give any song a thumbs up. Though I did hear a few favorites, I opted not to hand out any thumbs out to see what would happen to the station. I dont know if that is indeed the trick or not or maybe my Mary J Blige station, which is the same one I listened to last week, is finally learning me but either way I couldnt be happier!! Unexpecting the expected (yes I've reversed the common phrase-lol), I finally got it...I was too tired to bask in all its glory but this morning I woke up and played just a few more tracks to make sure it wasn't a fluke and I still had it.....a perfect session!!
Two strong weeks in a row!! Perhaps I was intially expecting too much out of the union and now I'm content with allowing things to progress as they may...the foundation of a solid BFF relationship. I'm even considering adding Pandora to my T-Mobile Fave 5 List!! LOL
WOW!!!
WOW!!!!
So throughout the course of the semester I've been looking for the "PERFECT" listening session. I can finally report to you all that it happened!! Each and every song that played is a song that I enjoy!! Track after track, artist after artist...it was great!! I started at 11:04pm and my mood was exceptionally tired, its been a loooong week :-( Being candid, I listened more out of obligation than desire. Often times, I listen to music while I work but last night I just didnt want to hear anything but I did and to my amazement, I stumbled upon the perfect session!! Though I didn't hear any new material, I did hear tracks that I havent played in a long time and that was enticing! For a split second on way of my old fav's I found my energy swing slightly upward and thats something that I long for on those late working nights!! While I didnt listen for hours on end as I normally do, I did make it through 22 tracks and every single one of them were perfect! The interesting thing though is that I didnt give any song a thumbs up. Though I did hear a few favorites, I opted not to hand out any thumbs out to see what would happen to the station. I dont know if that is indeed the trick or not or maybe my Mary J Blige station, which is the same one I listened to last week, is finally learning me but either way I couldnt be happier!! Unexpecting the expected (yes I've reversed the common phrase-lol), I finally got it...I was too tired to bask in all its glory but this morning I woke up and played just a few more tracks to make sure it wasn't a fluke and I still had it.....a perfect session!!
Two strong weeks in a row!! Perhaps I was intially expecting too much out of the union and now I'm content with allowing things to progress as they may...the foundation of a solid BFF relationship. I'm even considering adding Pandora to my T-Mobile Fave 5 List!! LOL
Thursday, April 3, 2008
March 30, 2008 Analysis: HOUSTON - We have lift off!!
So finally, FINALLY, i can report that Pandora and I had the best listening session to date. On Saturday, March 30, 2008 I decided to take a different approach. Normally, Pandora and I secretly meet in the wee hours of the evening and we often have our struggles about what Pandora wants me to hear vs what mood I'm in.....so I decided to start earlier. I opened Pandora at 4:47 pm and my mood at the time was content and peaceful...in other words, i left my critic hat in the closet! lol I decided to start with my Mary J. Blige station, seeing I am going to see her perform this weekend and we had an awesome run!! Out of 34 songs, I was overwhelmingly pleased with 26 of them. Of the remaining 8, 4 of them I liked somewhat so I allowed them to play in their entirety. Like a proud parent, I was ecstatic!! Way to rock Pandora!! While I did put a few songs to sleep, out of 34 there were only 2 songs that I just could not bare to listen to and thats a pretty awesome ratio!! I ended the first session around 6:23 and decided to pick it up again at 9:30pm, noting a relaxed and peaceful mood again. Still listening to my Mary station, I made it through 29 songs and to my amazement 25, yes TWENTY-FIVE songs were PERFECT!!! And the 4 songs that remain, I liked them somewhat so they weren't really bad suggestions!!! Pandora and I are officially BFF's for the moment ;-) I dont have a bad or constructive comment right now....We had an AMAZING run!!!
I dont know if it was my mood or the time of day or just the way the sun aligned with the moon but WE DID WELL!!!
KUDOS to Pandora!!!!
I dont know if it was my mood or the time of day or just the way the sun aligned with the moon but WE DID WELL!!!
KUDOS to Pandora!!!!
Friday, March 28, 2008
What's Missing From Pandora? - Comments from my Nearest Neighbor - No Really :-)
Based on our on-going Pandora discussion, my neighbor in class decided to do a little more investigative working, seeking the advice of a musical expert. He offers an interesting summary of what's missing from pandora and I, as odd as it may seem given my critic-nature, agree wholeheartedly!!
Great read!!
What's Missing From Pandora?
Great read!!
What's Missing From Pandora?
Thursday, March 27, 2008
March 26-27, 2008 Analysis - The Downfalls of Expectancy
As commonly seen in the IT world, we get a new gadget or program or API and play with all it's cool features and tricks. Though we've been through this process countless times before, the first time we run into a problem or are unable to achieve a desired task, aversion sets in! We're amazed and baffled that this new item can't help us take over the world....the nerve! lol Perhaps, I too am guilty of just expecting too much all of the time :-( In an effort to provide constructive self-criticism as well, I'll note this area as an opportunity for improvement in my personal growth!
So, Lucia from Pandora, whom is extremely knowledgeable and responsive, provided feedback a few weeks ago to help refine my created playlist. I started last night around 11:48pm (sidebar--based on my previous blogs, it's safe to say that I am somewhat a NIGHT OWL, I don't know when that started or how I've managed to keep it up, maybe it was the nap I took after work)...so anyways, i started around 11:48pm and my mood was both upbeat and energetic. I chose my Kanye West playlist thinking the list previously created would suit my current listening needs. The first track played was Kanye West - Everything I Am. I actually love this track so I provided a thumbs up. Then, attempting to use some of the feedback previously received, I used the "guide us" option to input a song name, still by Kanye West but a track called Good Life which is slightly more upbeat. Of the next nine songs I listened to, 2 were excellent selections but the others.....not so much. Three were songs I used to like but I've heard them so much that they've fallen into the "overplayed" category. One was an intro track and again while I don't mind hearing some of the them during entire album play, I have mixed feelings about here intro/outro/prelude/interlude tracks in this setting. While I understand there are legal constraints that prevent the exact song entered from being played on demand, I was at least hoping to hear something extremely similar to the song I entered. That didn't happen. So I though to myself, maybe I've confused the system. I previously started this Kanye West station and have already tailored/created a certain listening preference. Maybe it was a bad idea to try and use the "guide us" option, after being so far into the station tailoring process. So...I decided to create a new station, using an initial input of the Good Life song, sometimes fresh starts are good for everyone! So at 12:09am, Im still upbeat and energetic and created this new playlist. The first two songs I either like somewhat or used to like but it's been overplayed so recognizing the thumbs down limitation (more on that later), i decided to let them play out. The next two songs offered were not to my liking at all, maybe b/c I didnt rate the first two songs the system thought they were wrongfully acceptable. Of the next 7 songs, only 2 met my needs at the moment :-( I still didn't hear the Good Life song that I initially selected or anything similar to the mood I was trying to create by selecting that song. Thinking that I've officially screwed up everything I worked so hard to achieve by submitting so many changes/requests, I decided to go back to my Kanye West station and try to revive it. I listened to a couple of songs then attempting to "undo" what I've done, I use the "guide us" option to change it back to artist Kanye West opposed to a particular song. I went through 37 songs during this period, only 13 of those were dead-on so the success percentage I've seen in the past is steadily declining. This knowledge engineering process isn't just rough on the developers, it impacts us users too!! (sidebar 2--this is actually a really good exercise as it allows me to see and understand the complexities and hardships on both ends of the spectrum: the developer and the user). I do want to add that of the 13 songs that were played, two of them are songs that I absolutely love and haven't heard in a while so I was like a kid in a candy store when I heard them and there was another track that I've never heard but plan on looking for the album because I liked the song so much...so there still are absolute and obvious benefits to Pandora though my session was sinking. So again while there were some really good recommendations during that period, when it got bad....it got really, really bad! 16 songs were tracks that I've never heard before and they were certainly not what I wanted to hear last night. I was offered a track by AC/DC called Moneytalks Live and I seriously question why?! I decided to investigate why I was hearing that song and the system says it was the hard rock roots, subtle use of vocal harmony, repetitive melodic phrasing, extensive vamping and a vocal-centric aesthetic. The previous track was Rhymefest Build Me Up...and I question how close the previously rated tracks exhibiti the features given for the explanation. Three songs later I was offered a song by The Eagles called New Kid in Town-----speechless :-( Again, I investigated why I was hearing this particular song and it was suggested for its country influences, mellow rock instrumentation, heavy use of vocal harmonies, intricate melodic phrasing and mixed acoustic and electric instrumentation. I don't claim to be the most eclectic individual and while I like to think I'm diverse, there are some genres that just don't do it for me, country is one of them!! The mere indication that the song had country influences, should have been enough to remove this song from recommendation -- this is me falling into the previously mentioned trap/pitfall. I expect this engine to know exactly what I went, when I want it and simultaenously gather those items that should never be offered to me -- perfect case of me trying to "take over the world"! lol About a quarter through the 37 songs I noticed my mood was gradually changing, growing more tired and restless by the song....but I wasn't finished working so I figured I would give a little more input to the system. At 2:45am i switched to OLD FAITHFUL! lol My John Legend playlist, noting my mood as "winding down". I listened to another 14 songs and of those 11 of those were ALL KINDS OF AWESOME!!!!! Two of those, I actually liked somewhat though they didnt get placed in my "awesome bucket"! So with that said, only 1 of the track played did I give a thumbs down!! YAY PANDORA!!! I can conceded that maybe I was attempting to do just a little too much in my other station and attempting to "right the wrong" is an even more daunting task!
Lastly, I would like to add the thumbs down limitation is a bit of a pest at times. Lucia, if you're listening, can you offer any insight as to why the limitation exist? During that rough spell I found myself doing other things around the house or lowering the volume since I had reached my thumbs down limitation for the time being. I even hit mute one time and apparently Pandora is smart enough to pick up on that...ooops! lol One of the biggest downfalls that Ive noticed by this limitation, other than me wanting to change the song immediately and not having the ability to do so, it seems as if it steers the recommendation engine in the wrong direction. If I could tell the system that I truly don't like that song, without worrying about the constraint, perhaps the system would be able to perform better analysis and queue songs that are more to my liking...just my thoughts though!
I think that's enough for now!!
Signing off!
Remi A
Hoping for better sessions in the future....
So, Lucia from Pandora, whom is extremely knowledgeable and responsive, provided feedback a few weeks ago to help refine my created playlist. I started last night around 11:48pm (sidebar--based on my previous blogs, it's safe to say that I am somewhat a NIGHT OWL, I don't know when that started or how I've managed to keep it up, maybe it was the nap I took after work)...so anyways, i started around 11:48pm and my mood was both upbeat and energetic. I chose my Kanye West playlist thinking the list previously created would suit my current listening needs. The first track played was Kanye West - Everything I Am. I actually love this track so I provided a thumbs up. Then, attempting to use some of the feedback previously received, I used the "guide us" option to input a song name, still by Kanye West but a track called Good Life which is slightly more upbeat. Of the next nine songs I listened to, 2 were excellent selections but the others.....not so much. Three were songs I used to like but I've heard them so much that they've fallen into the "overplayed" category. One was an intro track and again while I don't mind hearing some of the them during entire album play, I have mixed feelings about here intro/outro/prelude/interlude tracks in this setting. While I understand there are legal constraints that prevent the exact song entered from being played on demand, I was at least hoping to hear something extremely similar to the song I entered. That didn't happen. So I though to myself, maybe I've confused the system. I previously started this Kanye West station and have already tailored/created a certain listening preference. Maybe it was a bad idea to try and use the "guide us" option, after being so far into the station tailoring process. So...I decided to create a new station, using an initial input of the Good Life song, sometimes fresh starts are good for everyone! So at 12:09am, Im still upbeat and energetic and created this new playlist. The first two songs I either like somewhat or used to like but it's been overplayed so recognizing the thumbs down limitation (more on that later), i decided to let them play out. The next two songs offered were not to my liking at all, maybe b/c I didnt rate the first two songs the system thought they were wrongfully acceptable. Of the next 7 songs, only 2 met my needs at the moment :-( I still didn't hear the Good Life song that I initially selected or anything similar to the mood I was trying to create by selecting that song. Thinking that I've officially screwed up everything I worked so hard to achieve by submitting so many changes/requests, I decided to go back to my Kanye West station and try to revive it. I listened to a couple of songs then attempting to "undo" what I've done, I use the "guide us" option to change it back to artist Kanye West opposed to a particular song. I went through 37 songs during this period, only 13 of those were dead-on so the success percentage I've seen in the past is steadily declining. This knowledge engineering process isn't just rough on the developers, it impacts us users too!! (sidebar 2--this is actually a really good exercise as it allows me to see and understand the complexities and hardships on both ends of the spectrum: the developer and the user). I do want to add that of the 13 songs that were played, two of them are songs that I absolutely love and haven't heard in a while so I was like a kid in a candy store when I heard them and there was another track that I've never heard but plan on looking for the album because I liked the song so much...so there still are absolute and obvious benefits to Pandora though my session was sinking. So again while there were some really good recommendations during that period, when it got bad....it got really, really bad! 16 songs were tracks that I've never heard before and they were certainly not what I wanted to hear last night. I was offered a track by AC/DC called Moneytalks Live and I seriously question why?! I decided to investigate why I was hearing that song and the system says it was the hard rock roots, subtle use of vocal harmony, repetitive melodic phrasing, extensive vamping and a vocal-centric aesthetic. The previous track was Rhymefest Build Me Up...and I question how close the previously rated tracks exhibiti the features given for the explanation. Three songs later I was offered a song by The Eagles called New Kid in Town-----speechless :-( Again, I investigated why I was hearing this particular song and it was suggested for its country influences, mellow rock instrumentation, heavy use of vocal harmonies, intricate melodic phrasing and mixed acoustic and electric instrumentation. I don't claim to be the most eclectic individual and while I like to think I'm diverse, there are some genres that just don't do it for me, country is one of them!! The mere indication that the song had country influences, should have been enough to remove this song from recommendation -- this is me falling into the previously mentioned trap/pitfall. I expect this engine to know exactly what I went, when I want it and simultaenously gather those items that should never be offered to me -- perfect case of me trying to "take over the world"! lol About a quarter through the 37 songs I noticed my mood was gradually changing, growing more tired and restless by the song....but I wasn't finished working so I figured I would give a little more input to the system. At 2:45am i switched to OLD FAITHFUL! lol My John Legend playlist, noting my mood as "winding down". I listened to another 14 songs and of those 11 of those were ALL KINDS OF AWESOME!!!!! Two of those, I actually liked somewhat though they didnt get placed in my "awesome bucket"! So with that said, only 1 of the track played did I give a thumbs down!! YAY PANDORA!!! I can conceded that maybe I was attempting to do just a little too much in my other station and attempting to "right the wrong" is an even more daunting task!
Lastly, I would like to add the thumbs down limitation is a bit of a pest at times. Lucia, if you're listening, can you offer any insight as to why the limitation exist? During that rough spell I found myself doing other things around the house or lowering the volume since I had reached my thumbs down limitation for the time being. I even hit mute one time and apparently Pandora is smart enough to pick up on that...ooops! lol One of the biggest downfalls that Ive noticed by this limitation, other than me wanting to change the song immediately and not having the ability to do so, it seems as if it steers the recommendation engine in the wrong direction. If I could tell the system that I truly don't like that song, without worrying about the constraint, perhaps the system would be able to perform better analysis and queue songs that are more to my liking...just my thoughts though!
I think that's enough for now!!
Signing off!
Remi A
Hoping for better sessions in the future....
Thursday, March 20, 2008
March 19, 2008 Analysis
Last night, Pandora and I reunited and went through 28 songs in just about 2 hours. 18 of the songs I heard were great recommendations. Many of which I had previously given a "thumbs up" so it's safe to assume that these songs are my "favs"!! I started at 11:30pm and my mood initially was extremely tired but I wanted to work just a little longer and thought Pandora's assistance could be valuable in helping me make it through the next few hours. So, I played a station that I previously created for one of my favorite artists..John Legend and the first five songs recommended were perfect!! Actually, the next song wasn't a bad selection either per say. I had previously given the songs a thumbs up but I just wasn't in the mood to hear it last night. 8 of the next 10 recommended were again, great selections!!! And then, it happened, the next four were just not what I wanted to hear...at all -( On the fourth song, I was going to call it an evening, figuring Pandora had given me the best I could ask for and maybe I was pushing my luck a little thinking we could continue like that for another hour....but something said try one more!! So i obliged and made it through 7 more songs. Five of which were dead-on!! Wanting to end my session on a plesant note and given the fact that I had to get up for work in less than 6 hours, I decided to call it a night! 18 out of 24 isnt all that bad but we definitely still have opportunities for improvement.
I want to give Lucia, a Pandora representative, a big THANK YOU. She provided several tips via blog comments that I will implement during future sessions in hopes of taking advantage of all Pandora has to offer!!
I want to give Lucia, a Pandora representative, a big THANK YOU. She provided several tips via blog comments that I will implement during future sessions in hopes of taking advantage of all Pandora has to offer!!
Thursday, February 21, 2008
February 17, 2008 Analysis
On Sunday, February 17, I woke up inspired and motivated to work. I'm working on research for another class and woke up anxious to do work so I wanted to listen to music that would match, if not elevate, that mood! So after doing a few things around the house, I opened up Pandora around 1:12pm and started a new search for Kirk Franklin. Over the next 2.5 hours I listened to 33 songs. Of the first 4 songs recommended, only 1 was to my liking. After that though, Pandora, hit it's stride. 14 of the next 15 songs were awesome! In fact 11 of those 15 songs were brand new to my ears and the one that I didn't want to listen to, was one that I had heard before. The next 14 songs were pretty much hit or miss....7 I enjoyed and 7, not so much! I'm not sure how Pandora hit it's stride or what features caused it to degrade somewhat but for a period of time, Pandora and I were in it to win it! lol I wish there was a way to identify what about that period of time worked so well or what attributes aligned with my mood/taste and tags from the recommendations but something obviously went awry!! Stay tuned, we'll keep the analysis going and maybe I can find the answer to my own question of the remainder of the semester!
Subscribe to:
Posts (Atom)