Classifying Amazon Reviews with Scikit-Learn — More Data is Better Turns Out

Last time, I went through some basics of how naive Bayes algorithm works, and the logic behind it, and implemented the classifier myself, as well as using the NLTK. That’s great and all, and hopefully people reading it got a better understanding of what was going on, and possibly how to play along with classification for their own text documents.

But if you’re looking to train and actually deploy a model, say, a website where people can copy paste reviews from Amazon and see how our classifier performs, you’re going to want to use a library like Scikit-Learn. So with this post, I’ll walk through training a Scikit-Learn model, testing various classifiers and parameters, in order to see how we do, and also at the end, will have an initial, version 1, of a Amazon review classifier that we can use in a production setting.

Some notes before we get going:

  • For a lot of the testing, I only use 5 or 10 of the full 26 classes that are in the dataset.
  • Keep in mind, that what works here might not be the same for other data sets. We’re specifically looking at Amazon product reviews. For a different set of texts (you’ll also see the word corpus being thrown around), a different classifier, or parameter sets might be used.
  • The resulting classifier we come up with is, well, really really basic, and probably what we’d guess would perform the best if we guessed what would be the best at the onset. All the time and effort that goes into checking all the combinations
  • I’m going to mention here this good post that popped up when I was looking around for other people who wrote about this. It really nicely outlines going how to classify text with Scikit-learn. To reduce redundancy, something that we all should work towards, I’m going to point you to that article to get up to speed on Scikit-learn and how it can apply to text. In this article, I’m going to start at the end of that article, where we’re working with Scikit-learn pipelines.

As always, you can say hi on twitter, or yell at me there for messing up as well if you want.

How many grams?

First step to think about is how we want to represent the reviews in naive Bayes world, in this case, a bag of words / n-grams. In the other post, I simply used word counts since I wasn’t going into how to make the best model we could have. But besides word counts, we can also bump up the representations to include something called a bigram, which is a two word combos. The idea behind that is that there’s information in two word combos that we aren’t using with just single words. With Scikit-learn, this is very simple to do, and they take care of it for you. Oh, and besides bigrams, we can say we want trigrams, fourgrams, etc. Which we’ll do, to see if that improves performance. Take a look at the wikipedia article for n-grams here.

For example is if a review mentions “coconut oil cream”, as in some sort of face cream (yup, I actually saw this as a mis-classified review), simply using the words and we might get a classification of food since we just see “coconut” “oil” and “cream”. But if we use bigrams as well as the unigrams, we’re also using “coconut oil” and “oil cream” as information. Now this might not get us all the way to a classification of beauty, but it could tip us over the edge.

Continue reading

Practical Naive Bayes — Classification of Amazon Reviews

If you search around the internet looking for applying Naive Bayes classification on text, you’ll find a ton of articles that talk about the intuition behind the algorithm, maybe some slides from a lecture about the math and some notation behind it, and a bunch of articles I’m not going to link here that pretty much just paste some code and call it an explanation.

So I’m going to try to do a little more here, by hopefully writing and explaining enough, is let you yourself write a working Naive Bayes classifier.

There are three sections here. First is setup, and what format I’m expecting your text to be in for the classification. Second, I’ll talk about how to run naive Bayes on your own, using slow Python data structures. Finally, we’ll use Python’s NLTK and it’s classifier so you can see how to use that, since, let’s be honest, it’s gonna be quicker. Note that you wouldn’t want to use either of these in production, so look for a follow up post about how you might go about doing that.

As always, twitter, and check out the full code on github.


Data from this is going to be from this UCSD Amazon review data set. I swear one of the biggest issues with running these algorithms on your own is finding a data set big and varied enough to get interesting results. Otherwise you’ll spend most of your time scraping and cleaning data that by the time you get to the ML part of the project, you’re sufficiently annoyed. So big thanks that this data already exists.

You’ll notice that this set has millions of reviews for products across 24 different classes. In order to keep the complexity down here (this is a tutorial post after all), I’m sticking with two classes, and ones that are somewhat far enough different from each other to show that classification works, we’ll be classifying baby reviews against tools and home improvement reviews.


First thing I want to do now, after unpacking the .gz file, is to get a train and test set that’s smaller than the 160,792 and 134,476 of baby and tool reviews respectively. For purposes here, I’m going to use 1000 of each, with 800 used for training, and 200 used for testing. The algorithms are able to support any number of training and test reviews, but for demonstration purposes, we’re making that number lower.

Check the github repo if you want to see the code, but I wrote a script that just takes the full file, picks 1000 random numbers, segments 800 into the training set, and 200 into the test set, and saves them to files with the names “train_CLASSNAME.json” and “test_CLASSNAME.json” where classname is either “baby” or “tool”.

Also, the files from that dataset are really nice, in that they’re already python objects. So to get them into a script, all you have to do is run “eval” on each line of the file if you want the dict object.


There really wasn’t a good place to talk about this, so I’ll mention it here before getting into either of the self, and nltk running of the algorithm. The features we’re going to use are simply the lowercased version of all the words in the review. This means, in order to get a list of these words from the block of text, we remove punctuation, lowercase every word, split on spaces, and then remove words that are in the NLTK corpus of stopwords (basically boring words that don’t have any information about class).

from nltk.corpus import stopwords
STOP_WORDS = set(stopwords.words('english'))
def clean_review(review):
  exclude = set(string.punctuation)
  review = ''.join(ch for ch in review if ch not in exclude)
  split_sentence = review.lower().split(" ")
  clean = [word for word in split_sentence if word not in STOP_WORDS]
  return clean

Realize here that there are tons of different ways to do this, and ways to get more sophisticated that hopefully can get you better results! Things like stemming, which takes words down to their root word (wikipedia gives the example of “stems”, “stemmer”, “stemming”, “stemmed” as based on “stem”). You might want to include n-grams, for an n larger than 1 in our case as well.

Basically, there’s tons of processing on the text that you could do here. But since this I’m just talking about how Naive Bayes works, I’m sticking with simplicity. Maybe in the future I can get fancy and see how well I can do in classifying these reviews.

Ok, on to the actual algorithm.

Continue reading

Classifying Country Music Songs is an Art — Getting Training Data

If you’ve been following along recently, I’ve been writing about my theory of country music, and how unlike most other genres out there, country music song topics are, let’s just say, much more centralized. And so in my continuing effort to automatically classify the country songs topic, I need to take all the songs lyrics I downloaded, and manually classify them so I have some training data.

This is actually the third post on this topic I’ve written. In the first post where I showed how to get song lyrics using Genius’s API and scraping, and then the second post, where I gathered up all the lyrics from country artists, removed the duplicates, and realized that Lee Brice talks about beer and trucks much more than he does about love. The stats I ran at the end of the second entry are fine and all, but really what I have at the moment is some 5 thousand songs that are uncategorized, which isn’t going to allow me to do any more sophisticated classification than simple word analysis.

What this means is I’m going to need some help classifying those 5000 songs. To do this, I wrote a rails app deployed on Heroku free mode that will allow anyone to sign up and help with this task. Obviously I’m not expecting people to get through all 5000 themselves (other than me of course), but hopefully if I can get enough people to do more than a few songs, I can get a good representation from which I can get interesting results.

Rest of the article is as follows. First, I’ll have a section where I talk about my theory of country music song topics, which I’ve been annoying my friends by talking about whenever we talk about country music. Then in the next / last section, I’ll talk about how I’m looking to get these songs classified, and what the interface is like and what’s going on behind the scenes.

As an aside, I am somewhat of a fan of country music. I usually just say it’s pop music with a slide, and by definition, pop music is catchy. But still, those country song lyrics can get quite ridiculous () which is definitely fun to laugh at.

Also, follow me along on twitter for more updates on this, and other topics.

Country Music Song Topics

Topic 1: Love

Love. The classic song topic — universally relatable, unbounded in subtopics, and somewhat of a default topic for any story, song or otherwise.Now that I think about it, I’m not sure there’s any song genre out there that doesn’t have love as a main song topic. So it makes sense that love is quite prevalent in many of the country songs.

Whether happy songs, about how Brett Young can’t go to sleep unless his girl is next to him at night,

or sad songs, about how the guy in Billy Currington’s song, “It Don’t Hurt Like It Used To”, was broken up with, but got over it eventually. Or somewhat, cause it still hurts.

Now that I think about it, I’m not sure there’s any song genre out there that doesn’t have love as a main song topic.

Topic 2: Small Town Life

Nothing says small town life like boots, dirt roads, dumpy bars with a cover band, railroad tracks, barns, white churches, and crop fields. No, I’m not making this up, those are just some of the things the band LoCash sings about in their recent song titled “I Love this Life”

If that wasn’t enough, how about picket fences, blue sky and green grass, old Ford trucks, back porches, homemade wine, tire swings, fireworks, and dead deer heads waiting to be hung on the wall. Yup, that’s just what Drake White is singing about in his song “Livin’ The Dream”. I admit, if I hear this when scrolling through stations on the radio in my car that doesn’t have an aux port, I’ll turn it up.

Continue reading

Talkin’ ‘Bout Trucks, Beer, and Love in Country Songs — Analyzing Genius Lyrics

Trucks, beer, and love, all things that make country music go round. I’ve said before that country music is just pop music with a slide, and then lyrics about slightly different topics than what you’ll hear in hip hop or “normal” pop music on the radio.

In my continuing quest to validate my theory that all country songs can fit into one of four different topics, in this post, I go through lyrics to see which artists talk about trucks, beer, and love the most. In my first post on this topic, I talked about how to get song lyrics from genius and print them out on the command line.

The goal here, and what I’m going to walk you through, is how I stored stored info and lyrics for all the songs for the country artists, how I made sure that all the lyrics were unique, and then ran some stats on the songs. Another note before we go is that a lot of data work is just janitorial. The actual code for getting “interesting” results is fairly simple. The key it to enjoy doing the janitor-style coding and then you’ll be good.

If you’re interested in which country music people talk most about trucks, beer, alcohol, or small towns, skip to the end where I list out some stats. For the rest, here’s some code.

I wonder how they feel about beer trucks. I’m guessing they’d all be fans of them.

Step 1 — Save the Lyrics!

When doing anything with web scraping, the one thing to always, always keep in mind here, is that you want to avoid hitting the server for as little as possible. With that in mind, we’re going to do here is assume the inputs are names of artists. For each of those artists, find all of their songs, and then for each of those songs, grab the lyrics in the way that I did in the first post, and then save them locally along with some meta information the API provides.

Now when I post the following code, don’t imagine that I knew what I wanted . Everything in here was created iteratively. Here’s a list of all the features of this piece of code does that were created iteratively.

Directory structure — Within the folder that contains the main .py file, there’s a folder named artists. And within that folder, when the code runs, a folder with the artist’s name is created (if not already). And within that folder, there are two more folders, info and lyrics. When we run the code, I put the lyrics in /artists/artist_name/lyrics/Song Title.txt and the info from the API, containing information about the song, like annotations, title, and song API id so we can grab it again if need be, in the file /artists/artist_name/info/Song Title.txt. The key, again, being saving all the info given to avoid unnecessary requests.

Redundancy Checking — Along with making sure to save all the info given, if we run an artist for the second time, we don’t want to get lyrics that we already have. So once we have all the songs for that artist, I run a check to see if we have a file with the name of the song already, and that the file isn’t empty. If the file is there, we continue to the next song.

Lyric Error Checking — Ahh unicode. While great for allowing multitudes of different characters rather than the standard English alphabet along with a few specialty characters, they’re not ideal when I’m trying to deal with simple song lyrics. And when saving the lyrics, I encountered more than a few random, unnecessary characters that Python threw errors for encoding problems. In a semi-janky rule-based solution (which isn’t great to use, see below), when I saw these errors being thrown, I would specifically replace them with the correct “normal” character. I assume there’s some library out there that would take care of all the encoding issues, but this worked for me. Also, on Genius’s end, it would be sweet if they, you know, checked for abnormal characters when lyrics were uploaded and didn’t have them in the first place. Also would be cool if they included the lyrics in the API.

def clean_lyrics(lyrics):
  lyrics = lyrics.replace(u"\u2019", "'") #right quotation mark
  lyrics = lyrics.replace(u"\u2018", "'") #left quotation mark
  lyrics = lyrics.replace(u"\u02bc", "'") #a with dots on top
  lyrics = lyrics.replace(u"\xe9", "e") #e with an accent
  lyrics = lyrics.replace(u"\xe8", "e") #e with an backwards accent
  lyrics = lyrics.replace(u"\xe0", "a") #a with an accent
  lyrics = lyrics.replace(u"\u2026", "...") #ellipsis apparently
  lyrics = lyrics.replace(u"\u2012", "-") #hyphen or dash
  lyrics = lyrics.replace(u"\u2013", "-") #other type of hyphen or dash
  lyrics = lyrics.replace(u"\u2014", "-") #other type of hyphen or dash
  lyrics = lyrics.replace(u"\u201c", '"') #left double quote
  lyrics = lyrics.replace(u"\u201d", '"') #right double quote
  lyrics = lyrics.replace(u"\u200b", ' ') #zero width space ?
  lyrics = lyrics.replace(u"\x92", "'") #different quote
  lyrics = lyrics.replace(u"\x91", "'") #still different quote
  lyrics = lyrics.replace(u"\xf1", "n") #n with tilde!
  lyrics = lyrics.replace(u"\xed", "i") #i with accent
  lyrics = lyrics.replace(u"\xe1", "a") #a with accent
  lyrics = lyrics.replace(u"\xea", "e") #e with circumflex
  lyrics = lyrics.replace(u"\xf3", "o") #o with accent
  lyrics = lyrics.replace(u"\xb4", "") #just an accent, so remove
  lyrics = lyrics.replace(u"\xeb", "e") #e with dots on top
  lyrics = lyrics.replace(u"\xe4", "a") #a with dots on top
  lyrics = lyrics.replace(u"\xe7", "c") #c with squigly bottom
  return lyrics

Check out the most of the main function below. If you’re looking for the actual full file, check out this gist. It’s easier to post that on Github than formatting the entire thing here.

Continue reading

Getting Song Lyrics from Genius’s API + Scraping

Genius is a great resource. At a high level, Genius has song lyrics and allows users to comment on what the artist meant. Starting as Rap Genius, where users annotated rap lyrics, the site rebranded as “Genius”, allowing all songs to be talked about. According to their website, “Genius is the world’s biggest collection of song lyrics and crowdsourced musical knowledge.” Recently even, they’ve moved to allowing annotations of pretty much anything posted online.

I’ve have used it a bunch recently while trying to figure out what the hell Frank Ocean was trying to say in his new album Blond. Users of the site explained tons of Frank’s references that went whoosh right over my head when I listened the first time and all the times after.

And recently, when I had some ideas for mini projects using song lyrics, I was pretty happy to find that Genius had a API for getting the data on their site. Whenever I’m trying to get data elsewhere, I’m much happier with an API, or at least being able to get it from JSON responses rather than parsing HTML. It’s just cleaner to look at, and with an API, I can expect good documentation that isn’t going to change with css updates.

Their API docs looked pretty good at first glance, with endpoints for artists, songs, albums, and annotations. One things I did notice was that they don’t have an artist entry point. A lot of what I want to do is artist based, meaning I need to know the artist id for everyone. And in order for me to get that, I have to search the artist, grab a song from the results, hit the song endpoint for that song’s information, and then grab the artist id from there. It’d be nice if you could specify what I’m searching for when I hit the search endpoint so I don’t have to go through that whole charade just to get the artist. But that’s a blog post for another time. Overall, they give out tons of information pretty easily.

But why, Genius, why don’t you have an endpoint for getting the raw lyrics of a song?! You have a songs endpoint on the API, and you give me a ton of information from there — the song title, album name, featured artists on the song, number of annotations, images associated with the song, album information, page views for that song, and a whole host of more data. But the one thing you don’t give me, and the one thing that people using the API probably want the most, is plain text lyrics!

Pre-Genius, I was stuck with these jankily laid out sites with super old looking css that would have the lyrics, but not necessarily correct, and definitely no annotations. Those sites are probably easily scrapeable considering their simplicity, but searching for the right song would be more difficult, and the lyrics might not be correct. Genius solved this all now for a web user, but dammit, I want the lyrics in the API!

Now you might be able to get the entire set of lyrics by using the annotations endpoint, which had information about all the annotations for a certain song or article, but that would require a song to have annotations for every word in the song. For someone like Chance the Rapper who like Frank Ocean (and most other hip hop artists uses tons of references in his lyrics, having complete annotations might not be an issue. But of Jake Owen, who’s new single “American Country Love Song” has probably the most self explanatory lyrics ever (sorry for throwing you under the bus here, Jake. Still a fan), there’s no need to annotate anything, and getting the lyrics in this manner wouldn’t work.

The lyrics are there on the internet however, and I can get at them by hitting the song endpoint, and using the web url that it returns. The rest of this article will show you how to do that using Python and it’s requests and BeautifulSoup libraries. But I don’t have to have to resort to HTML parsing, and I don’t think Genius wants users doing that either.

I’m left here wondering why they don’t want to give up the lyrics so easily, and I really don’t have much to go on. Genius’s goal seems to be wanting to annotate the internet. It has already moved on from their initial site of Rap Genius, into all music, and now into speech transcripts, as well as pretty much any other content on the web. Their value comes from those annotations themselves, not the information they’re annotating. They give away the annotations freely, but not the information (lyrics) in this case.

Enough speculation on why Genius doesn’t spit out the lyrics to a song when you get the other information. And as I’m writing this, I realize I easily could have overlooked something in their API and Genius might return the full lyrics, but I overlooked it. In that case, half of this article will be pointless and I’ll hold my head in shame from yelling at them like I did.

For purposes here, I’m going to show you how to get the song lyrics from Genius if you have the song title, and also talk through my process of getting there.

Note of clarification, just to make sure I’m not violating their terms of service, this post is for informational purposes only. Hopefully this can help programmers out there learn. Don’t do something bad with this knowledge. Code time!

First thing you’re going to need is an account set up with Genius. You can sign up from the upper right hand corner of the homepage. After that, navigate to the api docs where you’ll then see your Bearer token that you’ll need for all API requests.

I’m using the requests library here, and once you have the bearer token, here’s what all the API requests to Genius should look like if, for example, you’re searching for a song title.

import requests

#TOKEN below should be the string that the API docs tells you
#Clearly I'm not giving mine out here on the internet. That'd be dumb
base_url = ""
#Key line below here when, this is how to authorize your request when
#using the API
headers = {'Authorization': 'Bearer TOKEN'}
search_url = base_url + "/search"
song_title = "In the Midst of It All"
params = {'q': song_title}
response = requests.get(search_url, params=params, headers=headers)

The response, according to the Genius API, would be a list of songs that match that string passed in, with the first result being the Tom Misch song that I was going for. By changing around the url that is passed into the request method, you can access all the information that Genius supplies from the API (pretty much everything but the lyrics).

Continue reading

Predicting PGA Tour Scoring Average from Statistics Using Linear Regression

First off, I admit, that’s probably the most boring title for a blog post ever. It gets a negative value on the clickbait scale that is generally unseen in the modern, “every click equals dollars” era that we live in. On the other hand, it tells you exactly what this article is about — predicting scoring average using stats.

In this article, I’ll go through getting the data from the database, cleaning that data for use, and then running a linear regression in order to generate coefficients for each of the stats to generate scoring average predictions. Oh, and some analysis and commentary at the end!

Shameless shoutout to my other blog, Golf on the Mind. Check it out and subscribe to the newsletter / twitter / instagram if you’re into golf at all. Or ignore, and keep reading for some code!

Here's a pic of a golf course to get you in the mood.

Here’s a pic of a golf course to get you in the mood.

Getting the data

Last time if you remember, I spent all this effort taking the csv stat files, and putting the information into a database. Start there if you haven’t read that post yet. It’ll show how I grabbed the stats and formatted them.

Now that you’re back in the present we need to create a query that gets the stats for the players for a specific year. An example row in a CSV file of the data would be something like:

player_id, player_name, stat_1_value, stat_2_value, … , stat_n_value

for stats 1 to n where n (the number of stats), and the which stats themselves (driving distance, greens in regulation, etc.) vary depending on inputs.

Now let me say, I am not an expert in writing sql queries. And since people on the internet loooove to dole out hate in comments sections, I’m just going to say that there’s probably a better way of writing this query. Feel free to let me know and I can throw an edit in here, but this query works just fine.

  max(case when stat_lines.stat_id=330 then stat_lines.raw else null end) as putting_average,
  max(case when stat_lines.stat_id=157 then stat_lines.raw else null end) as driving_distance,
  max(case when stat_lines.stat_id=250 then stat_lines.raw else null end) as gir,
  max(case when stat_lines.stat_id=156 then stat_lines.raw else null end) as driving_accuracy,
  max(case when stat_lines.stat_id=382 then stat_lines.raw else null end) as scoring_average
from players
  join stat_lines on stat_lines.player_id =
  join stats on
where stat_lines.year=2012 and ( or or or or and stat_lines.raw is not null
group by,;

High level overview time! We’re selecting player id, and player name, along with their stats for putting average, driving distance, greens in regulation, driving accuracy and scoring average for the year 2012. In order to get the right stats, we need to know the stat id for the stats.

One more thing. This query is funky, and I probably could have designed the schema differently to make this prettier. For example, I could have just gone with one table, stat_lines, with fields for player_name and stat_name (along with all the current fields) and then the sql would be very simple. But there are other applications to keep in mind. What if you wanted to display all stats by a player? Or all of a players stats for a certain year? With the way I have the schema set up, those queries are simple and logical. For this specific case, I’ll deal with the complexity.

Loading the Data

That query above is great, but it’s not going to cut it if I have to specify what the year, and the stat ids in that string every time I run the script. Gotta be dynamic here.

Continue reading

Python, Postgres, SQLAlchemy, and PGA Tour Stats

A little ago, I wrote an article about scraping a bunch of PGA Tour stats. The end result of that was writing those stats out into CSV files. While this was suitable for that task of gathering the stats, let’s face it, you’re probably going to want to put those into some database to allow for easier querying, or possibly integrate it into to web app in the future. There are a bunch of different reasons for wanting this, so I’m going to go through the process I took to put all the data in the CSV files into the database.

Adding players to the database

First step is to fire up postgres! I’m not going to cover starting postgres since there’s so much good content about it, for example, this super good tutorial here by Digital Ocean. I created at database called ‘pgatour’, created a user named ‘pgatour_user’ with password ‘pgatour_user_password’, logged in, and created the first table, Player.

pgatour=# create table players (
  id serial PRIMARY KEY,
  name varchar (255) NOT NULL

Ok, now, as a test, I’m going to add myself into the database from the psql command line.

pgatour=# insert into players (name) values ('Jack Schultz')

Note that since id is serial, we don’t need to insert that value, just the name. Alas, I am not on the PGA Tour, so I’m going to need to delete myself.

pgatour=# select * from players;
id | name
1 | Jack Schultz
(1 row)

pgatour=# delete from players where name = 'Jack Schultz';
pgatour=# select * from players;
id | name
(0 rows)

Looking good. Now onto the Python side.

Python SQL Alchemy Interface

Now that the Players table in the database is set up, we’re going to want to be able to modify the contents of it in Python.

Continue reading