17  Text Classification and Word Embedding

Author

Phil Chodrow

Major components of this set of lecture notes are based on the Text Classification tutorial from the PyTorch documentation.

Deep Text Classification and Word Embedding

In this set of notes, we’ll discuss the problem of text classification. Text classification is a common problem in which we aim to classify pieces of text into different categories. These categories might be about:

  • Subject matter: is this news article about news, fashion, finance?
  • Emotional valence: is this tweet happy or sad? Excited or calm? This particular class of questions is so important that it has its own name: sentiment analysis.
  • Automated content moderation: is this Facebook comment a possible instance of abuse or harassment? Is this Reddit thread promoting violence? Is this email spam?

We saw text classification previously when we first considered the problem of vectorizing pieces of text. We are now going to look at a somewhat more contemporary approach to text using word embeddings.

import torch
import pandas as pd

import numpy as np

# for embedding visualization later
import plotly.express as px 
import plotly.io as pio

# for VSCode plotly rendering
pio.renderers.default = "notebook"

# for appearance
pio.templates.default = "plotly_white"

# for train-test split
from sklearn.model_selection import train_test_split

# for suppressing bugged warnings from torchinfo
import warnings 
warnings.filterwarnings("ignore", category = UserWarning)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

For this example, we are going to use a data set containing headlines from a large number of different news articles on the website HuffPost. I retrieved this data from Kaggle.

# access the data
url = "https://raw.githubusercontent.com/PhilChodrow/PIC16B/master/datasets/news/News_Category_Dataset_v2.json"
df  = pd.read_json(url, lines=True)
df  = df[["category", "headline"]]

There are over 200,000 headlines listed here, along with the category in which they appeared on the website.

df.head()
category headline
0 CRIME There Were 2 Mass Shootings In Texas Last Week...
1 ENTERTAINMENT Will Smith Joins Diplo And Nicky Jam For The 2...
2 ENTERTAINMENT Hugh Grant Marries For The First Time At Age 57
3 ENTERTAINMENT Jim Carrey Blasts 'Castrato' Adam Schiff And D...
4 ENTERTAINMENT Julianna Margulies Uses Donald Trump Poop Bags...

Our task will be to teach an algorithm to classify headlines by predicting the category based on the text of the headline.

Training a model on this much text data can require a lot of time, so we are going to simplify the problem a little bit, by reducing the number of categories. Let’s take a look at which categories we have:

df.groupby("category").size()
category
ARTS               1509
ARTS & CULTURE     1339
BLACK VOICES       4528
BUSINESS           5937
COLLEGE            1144
COMEDY             5175
CRIME              3405
CULTURE & ARTS     1030
DIVORCE            3426
EDUCATION          1004
ENTERTAINMENT     16058
ENVIRONMENT        1323
FIFTY              1401
FOOD & DRINK       6226
GOOD NEWS          1398
GREEN              2622
HEALTHY LIVING     6694
HOME & LIVING      4195
IMPACT             3459
LATINO VOICES      1129
MEDIA              2815
MONEY              1707
PARENTING          8677
PARENTS            3955
POLITICS          32739
QUEER VOICES       6314
RELIGION           2556
SCIENCE            2178
SPORTS             4884
STYLE              2254
STYLE & BEAUTY     9649
TASTE              2096
TECH               2082
THE WORLDPOST      3664
TRAVEL             9887
WEDDINGS           3651
WEIRD NEWS         2670
WELLNESS          17827
WOMEN              3490
WORLD NEWS         2177
WORLDPOST          2579
dtype: int64

Some of these categories are a little odd:

  • “Women”?
  • “Weird News”?
  • What’s the difference between “Style,” “Style & Beauty,” and “Taste”? ).
  • “Parenting” vs. “Parents”?
  • Etc?…

Well, there are definitely some questions here! Let’s just choose a few categories, and discard the rest. We’re going to give each of the categories an integer that we’ll use to encode the category in the target variable.

categories = {
    "STYLE"   : 0,
    "SCIENCE" : 1, 
    "TECH" : 2
}

df = df[df["category"].apply(lambda x: x in categories.keys())]
df.head()
category headline
137 TECH Facebook Accused Of Reading Texts And Accessin...
138 TECH Self-Driving Uber In Fatal Accident Had 6 Seco...
155 SCIENCE Scientists Turn To DNA Technology To Search Fo...
272 TECH Instagram Is Adding A 'Mute' Button For The Sa...
285 SCIENCE Unusual Asteroid Could Be An Interstellar Gues... 
df["category"] = df["category"].apply(categories.get)
df
category headline
137 2 Facebook Accused Of Reading Texts And Accessin...
138 2 Self-Driving Uber In Fatal Accident Had 6 Seco...
155 1 Scientists Turn To DNA Technology To Search Fo...
272 2 Instagram Is Adding A 'Mute' Button For The Sa...
285 1 Unusual Asteroid Could Be An Interstellar Gues...
... ... ...
200844 2 Google+ Now Open for Teens With Some Safeguards
200845 2 Web Wars
200846 2 First White House Chief Technology Officer, An...
200847 2 Watch The Top 9 YouTube Videos Of The Week
200848 2 RIM CEO Thorsten Heins' 'Significant' Plans Fo...

6514 rows × 2 columns

The base rate on this problem is the proportion of the data set occupied by the largest label class:

df.groupby("category").size() / len(df)
category
0    0.346024
1    0.334357
2    0.319619
dtype: float64

If we always guessed category 1, then we would expect an accuracy of roughly 35%. So, our task is to see whether we can train a model to beat this.

Next we need to wrap this Pandas dataframe as a Torch data set. While we’ve been using pre-implemented Torch classes for things like directories of images, in this case it’s not so hard to just implement our own Dataset. We just need to implement __getitem__() to return the appropriate row of the dataframe.

from torch.utils.data import Dataset, DataLoader

class TextDataFromDF(Dataset):
    def __init__(self, df):
        self.df = df
    
    def __getitem__(self, index):
        return self.df.iloc[index, 1], self.df.iloc[index, 0]

    def __len__(self):
        return len(self.df)

Now let’s perform a train-validation split and make Datasets from each one.

df_train, df_val = train_test_split(df,shuffle = True, test_size = 0.2)
train_data = TextDataFromDF(df_train)
val_data   = TextDataFromDF(df_val)

Each element of our data sets is a tuple of text and label:

train_data[194]
('LOOK: 8 Top Destinations For Science Nerds This Summer', 1)

Text Vectorization

Now we need to vectorize our text. The most classical way to do this is via one-hot encodings. We’re not going to do this; instead we are going to treat each sentence as a sequence of words, and identify each word via an integer index. First we’ll use a tokenizer to split each sentence into individual words:

from torchtext.data.utils import get_tokenizer
from torchtext.vocab import build_vocab_from_iterator

tokenizer = get_tokenizer('basic_english')

tokenized = tokenizer(train_data[194][0])
tokenized
['look',
 '8',
 'top',
 'destinations',
 'for',
 'science',
 'nerds',
 'this',
 'summer']

You might reasonably disagree about whether this is a good tokenization: should punctuation marks be included? Should “you’re” really have become “you”, “’”, and “re”? These are excellent questions that we won’t discuss too much further right now.

We’re now ready to build a vocabulary. A vocabulary is a mapping from words to integers. The code below loops through the training data and uses it to build such a mapping.

def yield_tokens(data_iter):
    for text, _ in data_iter:
        yield tokenizer(text)

vocab = build_vocab_from_iterator(yield_tokens(train_data), specials=["<unk>"])
vocab.set_default_index(vocab["<unk>"])

Here are the first couple elements of the vocabulary:

vocab.get_itos()[0:10]
['<unk>', "'", 'the', 'to', ',', 's', 'of', 'a', 'in', 'is']

This vocabulary can be applied on a list of tokens like this:

vocab(tokenized)
[63, 264, 74, 5725, 10, 56, 3733, 14, 134]

Batch Collation

Now we’re ready to construct the function that is going to actually pass a batch of data to our training loop. Here are the main steps:

  1. We pull some feature data (i.e. a batch of headlines).
  2. We represent each headline as a sequence of integers using the vocab.
  3. We pad the headlines with an unused integer index if necessary so that all headlines have the same length. This index corresponds to “blank” or “no words in this slot.”
  4. We return the batch of headlines as a consolidated tensor.
max_len = 30
num_tokens = len(vocab.get_itos())
def text_pipeline(x):
    tokens = vocab(tokenizer(x))
    y = torch.zeros(max_len, dtype=torch.int64) + num_tokens
    if len(tokens) > max_len:
        tokens = tokens[0:max_len]
    y[0:len(tokens)] = torch.tensor(tokens,dtype=torch.int64)
    return y

label_pipeline = lambda x: int(x)
text_pipeline("we can't believe")
tensor([  34,   32,    1,   31,  669, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
        9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
        9622, 9622, 9622, 9622, 9622, 9622])
def collate_batch(batch):
    label_list, text_list = [], []
    for (_text, _label) in batch:

        # add label to list
         label_list.append(label_pipeline(_label))

         # add text (as sequence of integers) to list
         processed_text = text_pipeline(_text)
         text_list.append(processed_text)

    label_list = torch.tensor(label_list, dtype=torch.int64)
    text_list = torch.stack(text_list)
    return text_list.to(device), label_list.to(device)
train_loader = DataLoader(train_data, batch_size=8, shuffle=True, collate_fn=collate_batch)
val_loader = DataLoader(val_data, batch_size=8, shuffle=True, collate_fn=collate_batch)

Let’s take a look at a batch of data now:

next(iter(train_loader))
(tensor([[  13,  398,    1,    5,  126,   12,   32, 6813,    2,  500,   44,   12,
           135,   23,   51,  121,    1,   31,  561,   36,    2, 3090,   24, 9622,
          9622, 9622, 9622, 9622, 9622, 9622],
         [  17,  320,  512,   43, 5103,  765,  222,    8,   16, 2835,  152,  254,
          9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622],
         [  20, 1105,    1,    5, 5783,  351, 3024,    7,  404,    8, 5468, 9622,
          9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622],
         [  14, 2302, 2231,    9,    7, 8497,   96, 8450, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622],
         [2161, 2440,    9, 5524,   81, 5026,   19, 1745, 1914, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622],
         [  12,  181,   30,  922,    3, 3793,    7, 8968, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622],
         [   1,   46, 7932,    1,   26, 6549,  265, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622],
         [  51,  481,    3,   63,  424,   21,   14, 1712, 1722,  277,  159, 9622,
          9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622, 9622,
          9622, 9622, 9622, 9622, 9622, 9622]]),
 tensor([2, 2, 2, 0, 0, 1, 2, 0]))

The first element is the sequence of tokens, one for each of the 8 headlines.The second element is the sequence of labels for each piece of text.

Modeling

Word Embedding

A word embedding refers to a representation of a word in a vector space. Each word is assigned an individual vector. The general aim of a word embedding is to create a representation such that words with related meanings are close to each other in a vector space, while words with different meanings are farther apart. One usually hopes for the directions connecting words to be meaningful as well. Here’s a nice diagram illustrating some of the general concepts:

Image credit: Towards Data Science

Word embeddings are often produced as intermediate stages in many machine learning algorithms. In our case, we’re going to add an embedding layer at the very base of our model. We’ll allow the user to flexibly specify the number of dimensions.

We’ll typically expect pretty low-dimensional embeddings for this lecture, but state-of-the-art embeddings will typically have a much higher number of dimensions. For example, the Embedding Projector demo supplied by TensorFlow uses a default dimension of 200.

from torch import nn

class TextClassificationModel(nn.Module):
    
    def __init__(self,vocab_size, embedding_dim, max_len, num_class):
        super().__init__()
        self.embedding = nn.Embedding(vocab_size+1, embedding_dim)
        self.fc   = nn.Linear(max_len*embedding_dim, num_class)
        
    def forward(self, x):
        x = self.embedding(x)
        x = torch.flatten(x, 1)
        x = self.fc(x)
        return(x)

Let’s learn and train a model!

vocab_size = len(vocab)
embedding_dim = 3
model = TextClassificationModel(vocab_size, embedding_dim, max_len, 3).to(device)
import time

optimizer = torch.optim.Adam(model.parameters(), lr=.1)
loss_fn = torch.nn.CrossEntropyLoss()

def train(dataloader):
    epoch_start_time = time.time()
    # keep track of some counts for measuring accuracy
    total_acc, total_count = 0, 0
    log_interval = 300
    start_time = time.time()

    for idx, (text, label) in enumerate(dataloader):
        # zero gradients
        optimizer.zero_grad()
        # form prediction on batch
        predicted_label = model(text)
        # evaluate loss on prediction
        loss = loss_fn(predicted_label, label)
        # compute gradient
        loss.backward()
        # take an optimization step
        optimizer.step()

        # for printing accuracy
        total_acc   += (predicted_label.argmax(1) == label).sum().item()
        total_count += label.size(0)
        
    print(f'| epoch {epoch:3d} | train accuracy {total_acc/total_count:8.3f} | time: {time.time() - epoch_start_time:5.2f}s')
    
def evaluate(dataloader):

    total_acc, total_count = 0, 0

    with torch.no_grad():
        for idx, (text, label) in enumerate(dataloader):
            predicted_label = model(text)
            total_acc += (predicted_label.argmax(1) == label).sum().item()
            total_count += label.size(0)
    return total_acc/total_count
EPOCHS = 20
for epoch in range(1, EPOCHS + 1):
    train(train_loader)
| epoch   1 | train accuracy    0.600 | time:  0.90s
| epoch   2 | train accuracy    0.860 | time:  0.82s
| epoch   3 | train accuracy    0.922 | time:  0.75s
| epoch   4 | train accuracy    0.956 | time:  0.70s
| epoch   5 | train accuracy    0.965 | time:  0.71s
| epoch   6 | train accuracy    0.964 | time:  0.68s
| epoch   7 | train accuracy    0.976 | time:  0.68s
| epoch   8 | train accuracy    0.976 | time:  0.68s
| epoch   9 | train accuracy    0.983 | time:  0.75s
| epoch  10 | train accuracy    0.984 | time:  0.73s
| epoch  11 | train accuracy    0.987 | time:  0.83s
| epoch  12 | train accuracy    0.986 | time:  0.73s
| epoch  13 | train accuracy    0.988 | time:  0.82s
| epoch  14 | train accuracy    0.987 | time:  0.78s
| epoch  15 | train accuracy    0.990 | time:  0.74s
| epoch  16 | train accuracy    0.991 | time:  0.74s
| epoch  17 | train accuracy    0.993 | time:  0.72s
| epoch  18 | train accuracy    0.993 | time:  0.73s
| epoch  19 | train accuracy    0.991 | time:  0.72s
| epoch  20 | train accuracy    0.996 | time:  0.70s
evaluate(val_loader)
0.841903300076746

Our accuracy on validation data is much lower than what we achieved on the training data. This is a possible sign of overfitting. Regardless, this predictive performance is much better than the roughly 34% that we would have achieved by guesswork:

Inspecting Word Embeddings

Recall from our discussion of image classification that the intermediate layers learned by the model can help us understand the representations that the model uses to construct its final outputs. In the case of word embeddings, we can simply extract this matrix from the corresponding layer of the model:

embedding_matrix = model.embedding.cpu().weight.data.numpy()

Let’s also extract the words from our vocabulary:

tokens = vocab.get_itos()

The embedding matrix itself has 3 columns, which is too many for us to conveniently visualize. So, instead we are going to use our friend PCA to extract a 2-dimensional representation that we can plot.

from sklearn.decomposition import PCA
pca = PCA(n_components=2)
weights = pca.fit_transform(embedding_matrix)

We’ll use the Plotly package to do the plotting. Plotly works best with dataframes:

tokens = vocab.get_itos()
tokens.append(" ")
embedding_df = pd.DataFrame({
    'word' : tokens, 
    'x0'   : weights[:,0],
    'x1'   : weights[:,1]
})

embedding_df
word x0 x1
0 <unk> 0.397705 0.675089
1 ' 2.056857 -1.362140
2 the 1.535659 -0.833715
3 to 1.825663 1.002944
4 , 0.807101 -1.193210
... ... ... ...
9618 ‘what 1.168435 5.259336
9619 ’ 0.787976 3.724449
9620 ’90s 1.304341 0.266395
9621 💯 5.319955 -2.834618
9622 0.344932 0.078845

9623 rows × 3 columns

And, let’s plot! We’ve used Plotly for the interactivity: hover over a dot to see the word it corresponds to.

fig = px.scatter(embedding_df, 
                 x = "x0", 
                 y = "x1", 
                 size = list(np.ones(len(embedding_df))),
                 size_max = 10,
                 hover_name = "word")

fig.show()

We’ve made an embedding! We might notice that this embedding appears to be a little bit “stretched out” in three main directions. Each one corresponds to one of the three classes in our training data.

Although modern methods for training word embeddings are much more complex, this example illustrates a key point: word embeddings are trained as byproducts of the process of training a model that learns to do something else, like text classification or predictive text generation.

Bias in Text Embeddings

Whenever we create a machine learning model that might conceivably have impact on the thoughts or actions of human beings, we have a responsibility to understand the limitations and biases of that model. Biases can enter into machine learning models through several routes, including the data used as well as choices made by the modeler along the way. For example, in our case:

  1. Data: we used data from a popular news source.
  2. Modeler choice: we only used data corresponding to a certain subset of labels.

With these considerations in mind, let’s see what kinds of words our model associates with female and male genders.

feminine = ["she", "her", "woman"]
masculine = ["he", "him", "man"]

highlight_1 = ["strong", "powerful", "smart",     "thinking", "brave", "muscle"]
highlight_2 = ["hot",    "sexy",     "beautiful", "shopping", "children", "thin"]

def gender_mapper(x):
    if x in feminine:
        return 1
    elif x in masculine:
        return 4
    elif x in highlight_1:
        return 3
    elif x in highlight_2:
        return 2
    else:
        return 0

embedding_df["highlight"] = embedding_df["word"].apply(gender_mapper)
embedding_df["size"]      = np.array(1.0 + 50*(embedding_df["highlight"] > 0))

# 
sub_df = embedding_df[embedding_df["highlight"] > 0]
import plotly.express as px 

fig = px.scatter(sub_df, 
                 x = "x0", 
                 y = "x1", 
                 color = "highlight",
                 size = list(sub_df["size"]),
                 size_max = 10,
                 hover_name = "word", 
                 text = "word")

fig.update_traces(textposition='top center')

fig.show()

What do you notice about some of the similarities represented in these embeddings? What do you wonder?

Representational Harm and Representational Bias

Earlier in this course, we discussed allocative bias. Allocative bias occurs when different groups have inequitable opportunities to access important resources or opportunities on the basis of their identity. We discussed examples that raised questions about equitable access to personal liberty, employment, and insurance.

Representational bias refers to the systematic cultural representation of marginalized groups in harmful ways, or of denying them cultural representation at all. The perpetuation of harmful stereotypes is perhaps the most well-known form of representational harm. Erasure is another form of representational harm in which representations or topics of interest to marginalized groups are suppressed.

Here’s a very recent example (from Margaret Mitchell) illustrating how representational gender bias shows up in ChatGPT:

Another form of representational harm in ML systems is the famous historical tendency of Google Search to surface demeaning and offensive search results related to people of color. This tendency was studied by Dr. Safiya Noble in her book Algorithms of Oppression. In one of Dr. Nobel’s most famous examples, top results for the phrase “black girls” in 2011 consisted of links to porn sites, which did not hold true of searches for “white girls” or “black men.” As late as 2016, an image search for “gorillas” would surface pictures of Black individuals. You can find a brief synopsis of some of Dr. Noble’s findings here (content warning: highly sexually explicit language). Google has since taken steps to improve these specific examples.

Bias in Google Translate

It is well-documented that machine learning algorithms trained on natural text can inherit biases present in those texts. One of the most direct ways in which we can observe such bias is in Google Translate. Some languages, such as Hungarian, do not possess gendered pronouns. When Google Translate attempts to render these pronouns into a gendered language like English, assumptions are made, as pointed out in this Tweet by Dora Vargha. Let’s demonstrate with the following English sentences.

he cooks. she is a political leader. she is an engineer. he is a cleaner. he is beautiful. she is strong.

Translate these into Hungarian and back via Google Translate, and here’s what you’ll get:

she cooks. he is a political leader. he is an engineer. she is cleaning. she is beautiful. he is strong.

Considering that English has a gender neutral pronoun (they), this would be an easy item to fix, which Google has thus far declined to do.

Intersections of Representational and Allocative Harms

In some cases, representational and allocative harms can intersect and reinforce each other. For example, modern translation systems perform impressively in certain languages but much less well in others. Some of these languages, such as Pashto and Dari, are spoken by many refugees and asylum-seekers to the US. The use of automated translation software when processing these asylum cases has drawn considerable scrutiny and appears to have resulted in the denial of at least one case due to a machine translation error.

More on Bias in Language Models

For more on the topic of bias in language models, you may wish to read the now-infamous paper by Emily Bender, Angelina McMillan-Major, Timnit Gebru, and “Shmargret Shmitchell” (Margaret Mitchell), “On the Dangers of Stochastic Parrots.” This is the paper that ultimately led to the firing of the final two authors by Google in late 2020 and early 2021.



© Phil Chodrow, 2024