Introduction
On this submit we are going to use Keras to categorise duplicated questions from Quora.
The dataset first appeared within the Kaggle competitors Quora Query Pairs and consists of roughly 400,000 pairs of questions together with a column indicating if the query pair is taken into account a replica.
Our implementation is impressed by the Siamese Recurrent Structure, with modifications to the similarity
measure and the embedding layers (the unique paper makes use of pre-trained phrase vectors). Utilizing this sort
of structure dates again to 2005 with Le Cun et al and is helpful for
verification duties. The concept is to study a operate that maps enter patterns right into a
goal area such {that a} similarity measure within the goal area approximates
the “semantic” distance within the enter area.
After the competitors, Quora additionally described their method to this downside on this weblog submit.
Dowloading knowledge
Information might be downloaded from the Kaggle dataset webpage
or from Quora’s launch of the dataset:
We’re utilizing the Keras get_file() operate in order that the file obtain is cached.
Studying and preprocessing
We’ll first load knowledge into R and do some preprocessing to make it simpler to
embody within the mannequin. After downloading the info, you may learn it
utilizing the readr read_tsv() operate.
We’ll create a Keras tokenizer to rework every phrase into an integer
token. We will even specify a hyperparameter of our mannequin: the vocabulary measurement.
For now let’s use the 50,000 commonest phrases (we’ll tune this parameter later).
The tokenizer can be match utilizing all distinctive questions from the dataset.
tokenizer <- text_tokenizer(num_words = 50000)
tokenizer %>% fit_text_tokenizer(distinctive(c(df$question1, df$question2)))Let’s save the tokenizer to disk to be able to use it for inference later.
save_text_tokenizer(tokenizer, "tokenizer-question-pairs")We’ll now use the textual content tokenizer to rework every query into a listing
of integers.
question1 <- texts_to_sequences(tokenizer, df$question1)
question2 <- texts_to_sequences(tokenizer, df$question2)Let’s check out the variety of phrases in every query. It will helps us to
determine the padding size, one other hyperparameter of our mannequin. Padding the sequences normalizes them to the identical measurement in order that we are able to feed them to the Keras mannequin.
80% 90% 95% 99%
14 18 23 31 We will see that 99% of questions have at most size 31 so we’ll select a padding
size between 15 and 30. Let’s begin with 20 (we’ll additionally tune this parameter later).
The default padding worth is 0, however we’re already utilizing this worth for phrases that
don’t seem throughout the 50,000 most frequent, so we’ll use 50,001 as an alternative.
question1_padded <- pad_sequences(question1, maxlen = 20, worth = 50000 + 1)
question2_padded <- pad_sequences(question2, maxlen = 20, worth = 50000 + 1)Now we have now completed the preprocessing steps. We’ll now run a easy benchmark
mannequin earlier than shifting on to the Keras mannequin.
Easy benchmark
Earlier than creating a sophisticated mannequin let’s take a easy method.
Let’s create two predictors: share of phrases from question1 that
seem within the question2 and vice-versa. Then we are going to use a logistic
regression to foretell if the questions are duplicate.
perc_words_question1 <- map2_dbl(question1, question2, ~imply(.x %in% .y))
perc_words_question2 <- map2_dbl(question2, question1, ~imply(.x %in% .y))
df_model <- knowledge.body(
perc_words_question1 = perc_words_question1,
perc_words_question2 = perc_words_question2,
is_duplicate = df$is_duplicate
) %>%
na.omit()Now that we’ve our predictors, let’s create the logistic mannequin.
We’ll take a small pattern for validation.
val_sample <- pattern.int(nrow(df_model), 0.1*nrow(df_model))
logistic_regression <- glm(
is_duplicate ~ perc_words_question1 + perc_words_question2,
household = "binomial",
knowledge = df_model[-val_sample,]
)
abstract(logistic_regression)Name:
glm(method = is_duplicate ~ perc_words_question1 + perc_words_question2,
household = "binomial", knowledge = df_model[-val_sample, ])
Deviance Residuals:
Min 1Q Median 3Q Max
-1.5938 -0.9097 -0.6106 1.1452 2.0292
Coefficients:
Estimate Std. Error z worth Pr(>|z|)
(Intercept) -2.259007 0.009668 -233.66 <2e-16 ***
perc_words_question1 1.517990 0.023038 65.89 <2e-16 ***
perc_words_question2 1.681410 0.022795 73.76 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial household taken to be 1)
Null deviance: 479158 on 363843 levels of freedom
Residual deviance: 431627 on 363841 levels of freedom
(17 observations deleted on account of missingness)
AIC: 431633
Variety of Fisher Scoring iterations: 3Let’s calculate the accuracy on our validation set.
[1] 0.6573577We acquired an accuracy of 65.7%. Not all that a lot better than random guessing.
Now let’s create our mannequin in Keras.
Mannequin definition
We’ll use a Siamese community to foretell whether or not the pairs are duplicated or not.
The concept is to create a mannequin that may embed the questions (sequence of phrases)
right into a vector. Then we are able to examine the vectors for every query utilizing a similarity
measure and inform if the questions are duplicated or not.
First let’s outline the inputs for the mannequin.
Then let’s the outline the a part of the mannequin that can embed the questions in a
vector.
word_embedder <- layer_embedding(
input_dim = 50000 + 2, # vocab measurement + UNK token + padding worth
output_dim = 128, # hyperparameter - embedding measurement
input_length = 20, # padding measurement,
embeddings_regularizer = regularizer_l2(0.0001) # hyperparameter - regularization
)
seq_embedder <- layer_lstm(
models = 128, # hyperparameter -- sequence embedding measurement
kernel_regularizer = regularizer_l2(0.0001) # hyperparameter - regularization
)Now we are going to outline the connection between the enter vectors and the embeddings
layers. Notice that we use the identical layers and weights on each inputs. That’s why
that is known as a Siamese community. It is sensible, as a result of we don’t need to have totally different
outputs if question1 is switched with question2.
We then outline the similarity measure we need to optimize. We wish duplicated questions
to have greater values of similarity. On this instance we’ll use the cosine similarity,
however any similarity measure could possibly be used. Do not forget that the cosine similarity is the
normalized dot product of the vectors, however for coaching it’s not essential to
normalize the outcomes.
cosine_similarity <- layer_dot(record(vector1, vector2), axes = 1)Subsequent, we outline a closing sigmoid layer to output the chance of each questions
being duplicated.
output <- cosine_similarity %>%
layer_dense(models = 1, activation = "sigmoid")Now that allow’s outline the Keras mannequin when it comes to it’s inputs and outputs and
compile it. Within the compilation part we outline our loss operate and optimizer.
Like within the Kaggle problem, we are going to reduce the logloss (equal
to minimizing the binary crossentropy). We’ll use the Adam optimizer.
We will then check out out mannequin with the abstract operate.
_______________________________________________________________________________________
Layer (sort) Output Form Param # Related to
=======================================================================================
input_question1 (InputLayer (None, 20) 0
_______________________________________________________________________________________
input_question2 (InputLayer (None, 20) 0
_______________________________________________________________________________________
embedding_1 (Embedding) (None, 20, 128) 6400256 input_question1[0][0]
input_question2[0][0]
_______________________________________________________________________________________
lstm_1 (LSTM) (None, 128) 131584 embedding_1[0][0]
embedding_1[1][0]
_______________________________________________________________________________________
dot_1 (Dot) (None, 1) 0 lstm_1[0][0]
lstm_1[1][0]
_______________________________________________________________________________________
dense_1 (Dense) (None, 1) 2 dot_1[0][0]
=======================================================================================
Whole params: 6,531,842
Trainable params: 6,531,842
Non-trainable params: 0
_______________________________________________________________________________________Mannequin becoming
Now we are going to match and tune our mannequin. Nonetheless earlier than continuing let’s take a pattern for validation.
set.seed(1817328)
val_sample <- pattern.int(nrow(question1_padded), measurement = 0.1*nrow(question1_padded))
train_question1_padded <- question1_padded[-val_sample,]
train_question2_padded <- question2_padded[-val_sample,]
train_is_duplicate <- df$is_duplicate[-val_sample]
val_question1_padded <- question1_padded[val_sample,]
val_question2_padded <- question2_padded[val_sample,]
val_is_duplicate <- df$is_duplicate[val_sample]Now we use the match() operate to coach the mannequin:
Practice on 363861 samples, validate on 40429 samples
Epoch 1/10
363861/363861 [==============================] - 89s 245us/step - loss: 0.5860 - acc: 0.7248 - val_loss: 0.5590 - val_acc: 0.7449
Epoch 2/10
363861/363861 [==============================] - 88s 243us/step - loss: 0.5528 - acc: 0.7461 - val_loss: 0.5472 - val_acc: 0.7510
Epoch 3/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5428 - acc: 0.7536 - val_loss: 0.5439 - val_acc: 0.7515
Epoch 4/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5353 - acc: 0.7595 - val_loss: 0.5358 - val_acc: 0.7590
Epoch 5/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5299 - acc: 0.7633 - val_loss: 0.5358 - val_acc: 0.7592
Epoch 6/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5256 - acc: 0.7662 - val_loss: 0.5309 - val_acc: 0.7631
Epoch 7/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5211 - acc: 0.7701 - val_loss: 0.5349 - val_acc: 0.7586
Epoch 8/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5173 - acc: 0.7733 - val_loss: 0.5278 - val_acc: 0.7667
Epoch 9/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5138 - acc: 0.7762 - val_loss: 0.5292 - val_acc: 0.7667
Epoch 10/10
363861/363861 [==============================] - 88s 242us/step - loss: 0.5092 - acc: 0.7794 - val_loss: 0.5313 - val_acc: 0.7654After coaching completes, we are able to save our mannequin for inference with the save_model_hdf5()
operate.
save_model_hdf5(mannequin, "model-question-pairs.hdf5")Mannequin tuning
Now that we’ve an inexpensive mannequin, let’s tune the hyperparameters utilizing the
tfruns package deal. We’ll start by including FLAGS declarations to our script for all hyperparameters we need to tune (FLAGS permit us to fluctuate hyperparmaeters with out altering our supply code):
FLAGS <- flags(
flag_integer("vocab_size", 50000),
flag_integer("max_len_padding", 20),
flag_integer("embedding_size", 256),
flag_numeric("regularization", 0.0001),
flag_integer("seq_embedding_size", 512)
)With this FLAGS definition we are able to now write our code when it comes to the flags. For instance:
input1 <- layer_input(form = c(FLAGS$max_len_padding))
input2 <- layer_input(form = c(FLAGS$max_len_padding))
embedding <- layer_embedding(
input_dim = FLAGS$vocab_size + 2,
output_dim = FLAGS$embedding_size,
input_length = FLAGS$max_len_padding,
embeddings_regularizer = regularizer_l2(l = FLAGS$regularization)
)The total supply code of the script with FLAGS might be discovered right here.
We moreover added an early stopping callback within the coaching step to be able to cease coaching
if validation loss doesn’t lower for five epochs in a row. It will hopefully cut back coaching time for dangerous fashions. We additionally added a studying fee reducer to scale back the training fee by an element of 10 when the loss doesn’t lower for 3 epochs (this method usually will increase mannequin accuracy).
We will now execute a tuning run to probe for the optimum mixture of hyperparameters. We name the tuning_run() operate, passing a listing with
the doable values for every flag. The tuning_run() operate can be accountable for executing the script for all combos of hyperparameters. We additionally specify
the pattern parameter to coach the mannequin for less than a random pattern from all combos (lowering coaching time considerably).
library(tfruns)
runs <- tuning_run(
"question-pairs.R",
flags = record(
vocab_size = c(30000, 40000, 50000, 60000),
max_len_padding = c(15, 20, 25),
embedding_size = c(64, 128, 256),
regularization = c(0.00001, 0.0001, 0.001),
seq_embedding_size = c(128, 256, 512)
),
runs_dir = "tuning",
pattern = 0.2
)The tuning run will return a knowledge.body with outcomes for all runs.
We discovered that the very best run attained 84.9% accuracy utilizing the mix of hyperparameters proven beneath, so we modify our coaching script to make use of these values because the defaults:
FLAGS <- flags(
flag_integer("vocab_size", 50000),
flag_integer("max_len_padding", 20),
flag_integer("embedding_size", 256),
flag_numeric("regularization", 1e-4),
flag_integer("seq_embedding_size", 512)
)Making predictions
Now that we’ve skilled and tuned our mannequin we are able to begin making predictions.
At prediction time we are going to load each the textual content tokenizer and the mannequin we saved
to disk earlier.
Since we received’t proceed coaching the mannequin, we specified the compile = FALSE argument.
Now let`s outline a operate to create predictions. On this operate we we preprocess the enter knowledge in the identical method we preprocessed the coaching knowledge:
predict_question_pairs <- operate(mannequin, tokenizer, q1, q2) {
q1 <- texts_to_sequences(tokenizer, record(q1))
q2 <- texts_to_sequences(tokenizer, record(q2))
q1 <- pad_sequences(q1, 20)
q2 <- pad_sequences(q2, 20)
as.numeric(predict(mannequin, record(q1, q2)))
}We will now name it with new pairs of questions, for instance:
predict_question_pairs(
mannequin,
tokenizer,
"What's R programming?",
"What's R in programming?"
)[1] 0.9784008Prediction is sort of quick (~40 milliseconds).
Deploying the mannequin
To exhibit deployment of the skilled mannequin, we created a easy Shiny software, the place
you may paste 2 questions from Quora and discover the chance of them being duplicated. Strive altering the questions beneath or coming into two fully totally different questions.
The shiny software might be discovered at https://jjallaire.shinyapps.io/shiny-quora/ and it’s supply code at https://github.com/dfalbel/shiny-quora-question-pairs.
Notice that when deploying a Keras mannequin you solely must load the beforehand saved mannequin file and tokenizer (no coaching knowledge or mannequin coaching steps are required).
Wrapping up
- We skilled a Siamese LSTM that provides us affordable accuracy (84%). Quora’s state-of-the-art is 87%.
- We will enhance our mannequin through the use of pre-trained phrase embeddings on bigger datasets. For instance, attempt utilizing what’s described in this instance. Quora makes use of their very own full corpus to coach the phrase embeddings.
- After coaching we deployed our mannequin as a Shiny software which given two Quora questions calculates the chance of their being duplicates.