Welcome to Contextualized Topic Models’s documentation!¶

Contextualized Topic Models¶

Contextualized Topic Models (CTM) are a family of topic models that use pre-trained representations of language (e.g., BERT) to support topic modeling. See the papers for details:

Bianchi, F., Terragni, S., & Hovy, D. (2021). Pre-training is a Hot Topic: Contextualized Document Embeddings Improve Topic Coherence. ACL. https://aclanthology.org/2021.acl-short.96/
Bianchi, F., Terragni, S., Hovy, D., Nozza, D., & Fersini, E. (2021). Cross-lingual Contextualized Topic Models with Zero-shot Learning. EACL. https://www.aclweb.org/anthology/2021.eacl-main.143/

Topic Modeling with Contextualized Embeddings¶

Our new topic modeling family supports many different languages (i.e., the one supported by HuggingFace models) and comes in two versions: CombinedTM combines contextual embeddings with the good old bag of words to make more coherent topics; ZeroShotTM is the perfect topic model for task in which you might have missing words in the test data and also, if trained with muliglingual embeddings, inherits the property of being a multilingual topic model!

The big advantage is that you can use different embeddings for CTMs. Thus, when a new embedding method comes out you can use it in the code and improve your results. We are not limited by the BoW anymore.

We also have kitty! a new submodule that can be used to quickly create an human in the loop classifier to quickly classify your documents and create named clusters.

Tutorials¶

You can look at our medium blog post or start from one of our Colab Tutorials:

Name	Link
Combined TM on Wikipedia Data (Preproc+Saving+Viz) (stable v2.0.0)
Zero-Shot Cross-lingual Topic Modeling (Preproc+Viz) (stable v2.0.0)
SuperCTM and β-CTM (High-level usage) (stable v2.1.0)

Overview¶

TL;DR¶

In CTMs we have two models. CombinedTM and ZeroShotTM, which have different use cases.
CTMs work better when the size of the bag of words has been restricted to a number of terms that does not go over 2000 elements. This is because we have a neural model that reconstructs the input bag of word, Moreover, in CombinedTM we project the contextualized embedding to the vocab space, the bigger the vocab the more parameters you get, with the training being more difficult and prone to bad fitting. This is NOT a strict limit, however, consider preprocessing your dataset. We have a preprocessing pipeline that can help you in dealing with this.
Check the contextual model you are using, the multilingual model one used on English data might not give results that are as good as the pure English trained one.
Preprocessing is key. If you give a contextual model like BERT preprocessed text, it might be difficult to get out a good representation. What we usually do is use the preprocessed text for the bag of word creating and use the NOT preprocessed text for BERT embeddings. Our preprocessing class can take care of this for you.

Features¶

An important aspect to take into account is which network you want to use: the one that combines contextualized embeddings and the BoW (CombinedTM) or the one that just uses contextualized embeddings (ZeroShotTM)

But remember that you can do zero-shot cross-lingual topic modeling only with the ZeroShotTM model. See cross-lingual-topic-modeling

We also have Kitty: a utility you can use to do a simpler human in the loop classification of your documents. This can be very useful to do document filtering. It also works in cross-lingual setting and thus you might be able to filter documents in a language you don’t know!

References¶

If you find this useful you can cite the following papers :)

ZeroShotTM

@inproceedings{bianchi-etal-2021-cross,
    title = "Cross-lingual Contextualized Topic Models with Zero-shot Learning",
    author = "Bianchi, Federico and Terragni, Silvia and Hovy, Dirk  and
      Nozza, Debora and Fersini, Elisabetta",
    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
    month = apr,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2021.eacl-main.143",
    pages = "1676--1683",
}

CombinedTM

@inproceedings{bianchi-etal-2021-pre,
    title = "Pre-training is a Hot Topic: Contextualized Document Embeddings Improve Topic Coherence",
    author = "Bianchi, Federico  and
      Terragni, Silvia  and
      Hovy, Dirk",
    booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 2: Short Papers)",
    month = aug,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.acl-short.96",
    doi = "10.18653/v1/2021.acl-short.96",
    pages = "759--766",
}

Development Team¶

Federico Bianchi <f.bianchi@unibocconi.it> Bocconi University
Silvia Terragni <s.terragni4@campus.unimib.it> University of Milan-Bicocca
Dirk Hovy <dirk.hovy@unibocconi.it> Bocconi University

Software Details¶

Free software: MIT license
Documentation: https://contextualized-topic-models.readthedocs.io.
Super big shout-out to Stephen Carrow for creating the awesome https://github.com/estebandito22/PyTorchAVITM package from which we constructed the foundations of this package. We are happy to redistribute this software again under the MIT License.

Credits¶

This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template. To ease the use of the library we have also included the rbo package, all the rights reserved to the author of that package.

Note¶

Remember that this is a research tool :)

CombinedTM: Coherent Topic Models¶

Combined TM combines the BoW with SBERT, a process that seems to increase the coherence of the predicted topics (https://arxiv.org/pdf/2004.03974.pdf).

Usage¶

Here is how you can use the CombinedTM. This is a standard topic model that also uses contextualized embeddings. The good thing about CombinedTM is that it makes your topic much more coherent (see the paper https://arxiv.org/abs/2004.03974). n_components=50 specifies the number of topics.

from contextualized_topic_models.models.ctm import CombinedTM
from contextualized_topic_models.utils.data_preparation import TopicModelDataPreparation
from contextualized_topic_models.utils.data_preparation import bert_embeddings_from_file

qt = TopicModelDataPreparation("paraphrase-distilroberta-base-v2")

training_dataset = qt.fit(text_for_contextual=list_of_unpreprocessed_documents, text_for_bow=list_of_preprocessed_documents)

ctm = CombinedTM(bow_size=len(qt.vocab), contextual_size=768, n_components=50) # 50 topics

ctm.fit(training_dataset) # run the model

ctm.get_topics()

Once the model is trained, it is very easy to get the topics!

ctm.get_topics()

Creating the Test Set¶

The transform method will take care of most things for you, for example the generation of a corresponding BoW by considering only the words that the model has seen in training.

If you use CombinedTM you need to include the test text for the BOW:

testing_dataset = qt.transform(text_for_contextual=testing_text_for_contextual, text_for_bow=testing_text_for_bow)

# n_sample how many times to sample the distribution (see the doc)
ctm.get_doc_topic_distribution(testing_dataset, n_samples=20) # returns a (n_documents, n_topics) matrix with the topic distribution of each document

Warning¶

Note that the way we use the transform method here is different from what we do for ZeroShotTM! This is very important!

Tutorial¶

You can find a tutorial here: it will show you how you can use CombinedTM.

ZeroShotTM: Topic Modeling with Missing Words and Cross-Lingual Capabilities¶

Our ZeroShotTM can be used for zero-shot topic modeling. This is because the entire document is encoded into an embedding using a contextualized model. Thus, we are not limited by the usual problems you might encounter with bag of words: at test time, words that are missing from the training set will be encoded using the contextualized model, thus providing a reliable topic model even in sparse context!

More interestingly, this model can be used for cross-lingual topic modeling (See next sections)! You can also read paper (https://www.aclweb.org/anthology/2021.eacl-main.143)

Usage¶

from contextualized_topic_models.models.ctm import ZeroShotTM
from contextualized_topic_models.utils.data_preparation import TopicModelDataPreparation
from contextualized_topic_models.utils.data_preparation import bert_embeddings_from_file

text_for_contextual = [
    "hello, this is unpreprocessed text you can give to the model",
    "have fun with our topic model",
]

text_for_bow = [
    "hello unpreprocessed give model",
    "fun topic model",
]

qt = TopicModelDataPreparation("paraphrase-multilingual-mpnet-base-v2")

training_dataset = qt.fit(text_for_contextual=text_for_contextual, text_for_bow=text_for_bow)

ctm = ZeroShotTM(bow_size=len(qt.vocab), contextual_size=768, n_components=50)

ctm.fit(training_dataset) # run the model

ctm.get_topics(2)

As you can see, the high-level API to handle the text is pretty easy to use; text_for_bert should be used to pass to the model a list of documents that are not preprocessed. Instead, to text_for_bow you should pass the preprocessed text used to build the BoW.

Once the model is trained, it is very easy to get the topics!

ctm.get_topics()

Creating the Test Set¶

The transform method will take care of most things for you, for example the generation of a corresponding BoW by considering only the words that the model has seen in training. However, this comes with some bumps when dealing with the ZeroShotTM, as we will se in the next section.

If you use ZeroShotTM you do not need to use the testing_text_for_bow because if you are using a different set of test documents, this will create a BoW of a different size. Thus, the best way to do this is to pass just the text that is going to be given in input to the contexual model:

testing_dataset = qt.transform(text_for_contextual=testing_text_for_contextual)

# n_sample how many times to sample the distribution (see the doc)
ctm.get_doc_topic_distribution(testing_dataset, n_samples=20)

Warning¶

Note that the way we use the transform method here is different from what we do for CombinedTM! This is very important!

Cross-Lingual Topic Modeling¶

Once you have trained the ZeroShotTM model with multilingual embeddings, you can use this simple pipeline to predict the topics for documents in a different language (as long as this language is covered by paraphrase-multilingual-mpnet-base-v2).

# here we have a Spanish document
testing_text_for_contextual = [
    "hola, bienvenido",
]

# since we are doing multilingual topic modeling, we do not need the BoW in
# ZeroShotTM when doing cross-lingual experiments (it does not make sense, since we trained with an english Bow
# to use the spanish BoW)
testing_dataset = qt.transform(testing_text_for_contextual)

# n_sample how many times to sample the distribution (see the doc)
ctm.get_doc_topic_distribution(testing_dataset, n_samples=20) # returns a (n_documents, n_topics) matrix with the topic distribution of each document

Advanced Notes: We do not need to pass the Spanish bag of word: the bag of words of the two languages will not be comparable! We are passing it to the model for compatibility reasons, but you cannot get the output of the model (i.e., the predicted BoW of the trained language) and compare it with the testing language one.

Kitty: Human-in-the-loop Classification with Contextualized Topic Models¶

Kitty is a utility to generate a simple topic classifier from a topic model. It first runs a CTM instance on the data for you and you can then select and label a set of topics of interest. Once this is done, you can apply this selection to a wider range of documents.

Please cite the following papers if you use Kitty:

Bianchi, F., Terragni, S., & Hovy, D. (2021). Pre-training is a Hot Topic: Contextualized Document Embeddings Improve Topic Coherence. ACL. https://aclanthology.org/2021.acl-short.96/
Bianchi, F., Terragni, S., Hovy, D., Nozza, D., & Fersini, E. (2021). Cross-lingual Contextualized Topic Models with Zero-shot Learning. EACL. https://www.aclweb.org/anthology/2021.eacl-main.143/

Usage¶

from contextualized_topic_models.models.kitty_classifier import Kitty

# read the training data
training_set = list(map(lambda x : x.strip(), open("train_data").readlines()))

kt = Kitty(language="english")
kt.train(training_set, 5) # train a topic model with 5 topics

print(kt.pretty_print_word_classes())

This could probably output topics like these ones:

 family, plant, types, type, moth
 district, mi, area, village, west
 released, series, television, album, film
 school, station, historic, public, states
 born, football, team, played, season

Now, you can then use a simple dictionary to assign the topics to some labels. For example, topic 0 seems to be describing nature related things.

kt.assigned_classes = {0 : "nature", 1 : "location",
                       2 : "entertainment", 3 : "shop/offices", 4: "sport"}

kt.predict(["the village of Puza is a very nice village in Italy"])

>> location

kt.predict(["Pussetto is a soccer player that currently plays for Udiense Calcio"])

>> sport

If you are using a jupyter notebook, you can use the widget to fill in the labels.

kt.widget_annotation()

Cross-Lingual Support¶

A nice feature of Kitty is that it can be used to filter documents in different languages. Assume you have access to a large corpus of Italian documents and a smaller corpus of English documents. You can run Kitty on the English documents, map the labels and apply Kitty on the Italian documents. It is enough to change the embedding model.

from contextualized_topic_models.models.kitty_classifier import Kitty

# read the training data
training = list(map(lambda x : x.strip(), open("train_data").readlines()))

# define kitty with a multilingual embedding model
kt = Kitty(embedding_model="paraphrase-multilingual-mpnet-base-v2",  contextual_size=768)

kt.train(training, 5) # train a topic model with 5 topics

print(kt.pretty_print_word_classes())

You can then apply the mapping as we did before and predict in different languages:

kt.predict(["Pussetto è un calciatore che attualmente gioca per l'Udinese Calcio"])

>> sport

You should refer to SBERT Pretrained Models to know if the languages you want to use are supported by SBERT.

What Makes Kitty Different Other Topic Models?¶

Nothing! It just offers a user-friendly utility that makes use of the ZeroShotTM model in the backend.

Extensions: SuperCTM and β-CTM¶

SuperCTM¶

Inspiration for SuperCTM has been taken directly from the work by Card et al., 2018 (you can read this as “we essentially implemented their approach in our architecture”). SuperCTM should give better representations of the documents - this is somewhat expected, since we are using the labels to give more information to the model - and in theory should also make the model able to find topics more coherent with respect to the labels. The model is super easy to use and requires minor modifications to the already implemented pipeline:

from contextualized_topic_models.models.ctm import ZeroShotTM
from contextualized_topic_models.utils.data_preparation import TopicModelDataPreparation

text_for_contextual = [
    "hello, this is unpreprocessed text you can give to the model",
    "have fun with our topic model",
]

text_for_bow = [
    "hello unpreprocessed give model",
    "fun topic model",
]

labels = [0, 1] # we need to have  a label for each document

qt = TopicModelDataPreparation("paraphrase-multilingual-mpnet-base-v2")

# training dataset should contain the labels
training_dataset = qt.fit(text_for_contextual=text_for_contextual, text_for_bow=text_for_bow, labels=labels)

# model should know the label size in advance
ctm = CombinedTM(bow_size=len(qt.vocab), contextual_size=768, n_components=50, label_size=len(set(labels)))

ctm.fit(training_dataset) # run the model

ctm.get_topics(2)

β-CTM¶

We also implemented the intuition found in the work by Higgins et al., 2018, where a weight is applied to the KL loss function. The idea is that giving more weight to the KL part of the loss function helps in creating disentangled representations by forcing independence in the components. Again, the model should be straightforward to use:

ctm = CombinedTM(bow_size=len(qt.vocab), contextual_size=768, n_components=50, loss_weights={"beta" : 3})

Evaluation¶

We have also included some of the metrics normally used in the evaluation of topic models, for example you can compute the coherence of your topics using NPMI using our simple and high-level API.

Metrics Covered¶

The metrics we cover are the one we also describe in our papers.

Coherence (e.g., NPMI, Word Embeddings)
Topic Diversity (e.g., Inversed RBO)
Matches
Centroid Distance

Example¶

from contextualized_topic_models.evaluation.measures import CoherenceNPMI

with open('preprocessed_documents.txt', "r") as fr:
    texts = [doc.split() for doc in fr.read().splitlines()] # load text for NPMI

npmi = CoherenceNPMI(texts=texts, topics=ctm.get_topic_lists(10))
npmi.score()

Mono and Multi-Lingual Embeddings¶

Multilingual¶

Some of the examples below use a multilingual embedding model paraphrase-multilingual-mpnet-base-v2. This means that the representations you are going to use are mutlilinguals. However you might need a broader coverage of languages. In that case, you can check SBERT to find a model you can use.

English¶

If you are doing topic modeling in English, you SHOULD use an English sentence-bert model, for example paraphrase-distilroberta-base-v2. In that case, it’s really easy to update the code to support monolingual English topic modeling. If you need other models you can check SBERT for other models.

qt = TopicModelDataPreparation("paraphrase-distilroberta-base-v2")

Language-Specific¶

In general, our package should be able to support all the models described in the sentence transformer package and in HuggingFace. You need to take a look at HuggingFace models and find which is the one for your language. For example, for Italian, you can use UmBERTo. How to use this in the model, you ask? well, just use the name of the model you want instead of the english/multilingual one:

qt = TopicModelDataPreparation("Musixmatch/umberto-commoncrawl-cased-v1")

Published Papers on CTM¶

Contextualized Topic Models (CTM) are a family of topic models that use pre-trained representations of language (e.g., BERT) to support topic modeling. See the papers for details:

Bianchi, F., Terragni, S., & Hovy, D. (2021). Pre-training is a Hot Topic: Contextualized Document Embeddings Improve Topic Coherence. ACL. https://aclanthology.org/2021.acl-short.96/
Bianchi, F., Terragni, S., Hovy, D., Nozza, D., & Fersini, E. (2021). Cross-lingual Contextualized Topic Models with Zero-shot Learning. EACL. https://www.aclweb.org/anthology/2021.eacl-main.143/

If you want to replicate our results, you can use our code. You will find the W1 dataset in the colab and here: https://github.com/vinid/data, if you need the W2 dataset, send us an email (it is a bit bigger than W1 and we could not upload it on github).

Note: Thanks to Camille DeJarnett (Stanford/CMU), Xinyi Wang (CMU), and Graham Neubig (CMU) we found out that with the current version of the package and all the dependencies (e.g., the sentence transformers embedding model, CUDA version, PyTorch version), results with the model distiluse-base-multilingual-cased are lower than what appears in the paper. We suggest to use paraphrase-multilingual-mpnet-base-v2 which is a newer multilingual model that has results that are higher than those in the paper.

See for example the results on the matches metric for Italian in the following table.

Model Name	Matches
paraphrase-multilingual-mpnet-base-v2	0.67
distiluse-base-multilingual-cased	0.57
paper	0.62

Thus, if you use ZeroShotTM for a multilingual task, we suggest the use of paraphrase-multilingual-mpnet-base-v2.

Visualization¶

We offer two main plots to show your topics. The first one is based on PyLDAvis, a great package to analyze how your topic model is behaving. You can take a look to the PyLDAvis Github Page to get a better idea on how to interpret the different components.

The second visualization is a very simple topic visualization with a wordcloud.

PyLdaVis Visualization¶

If you already know how to use that, you’ll see that using PyLDAvis is very simple. First of all you need to install pyldavis.

We support pyLDA visualizations with few lines of code!

import pyLDAvis as vis

lda_vis_data = ctm.get_ldavis_data_format(tp.vocab, training_dataset, n_samples=10)

ctm_pd = vis.prepare(**lda_vis_data)
vis.display(ctm_pd)

https://raw.githubusercontent.com/MilaNLProc/contextualized-topic-models/master/img/pyldavis.png

Showing The Topic Word Cloud¶

You can also create a word cloud of the topic!

ctm.get_wordcloud(topic_id=47, n_words=15)

https://raw.githubusercontent.com/MilaNLProc/contextualized-topic-models/master/img/displaying_topic.png

Installation¶

Stable release¶

To install Contextualized Topic Models, run this command in your terminal:

$ pip install contextualized_topic_models

This is the preferred method to install Contextualized Topic Models, as it will always install the most recent stable release.

If you don’t have pip installed, this Python installation guide can guide you through the process.

From sources¶

The sources for Contextualized Topic Models can be downloaded from the Github repo.

You can either clone the public repository:

$ git clone git://github.com/vinid/contextualized_topic_models

Or download the tarball:

$ curl -OJL https://github.com/MilaNLProc/contextualized_topic_models/tarball/master

Once you have a copy of the source, you can install it with:

$ python setup.py install

Frequently Asked Questions¶

I am getting very poor results. What can I do?¶

There are many elements that can influence the final results in a topic model. A good preprocessing is fundamental for obtaining meaningful topics. On this link you can find some suggestions on how to preprocess your data, but be careful because each dataset may have their pecularities. If you still get poor results, don’t hesitate to contact us! We would be happy to help you :)

Am I forced to use the SBERT Embeddings?¶

Not at all! you can actually the embedding you like most. Keep in mind to check that the matrix you create as the same number of documents (first dimension) as the Bag of Word representation and you are good to go! Check the code here to see how we create the representations.

Is the BoW needed even for the ZeroShotTM?¶

Yes, it is. The BoW is necessary in the reconstruction phase, without that we would lose the symbolic information that allows us to get the topics.

ZeroShotTM or CombinedTM? Which one should I use?¶

ZeroShotTm and CombinedTM can be basically used for the same tasks. ZeroShotTM has two main pros:

1) it can handle unseen words in the test phase. This makes it very useful to be used in our Kitty module, for example.

2) If your objective is to do cross-lingual topic modeling (i.e. train a topic model on a dataset in one language and predict the topics for data in other languages), then ZeroShotTM is the model for you.

If you just aim at extracting topics from a corpus, you can use either the CombinedTM or the ZeroShotTM. We have designed the CombinedTM for the purpose of obtaining more coherent topics, so we suggest you use this for more general topic extraction. Yet, as you can read in this paper, the ZeroShotTM model still gets results that are very similar to the ones of the CombinedTM.

How do I choose the correct number of topics?¶

The “n_component” parameter represents the number of topics for CTM. There is not a “right” answer about the choice of the number of topics. Usually, researchers try a different number of topics (10, 30, 50, etc, depending on the prior knowledge on the dataset) and select the number of topics that guarantees the highest average topic coherence. We also suggest you take into consideration the topic diversity.

Credits¶

Development Lead¶

Federico Bianchi <f.bianchi@unibocconi.it>
Silvia Terragni <s.terragni4@campus.unimib.it>
Dirk Hovy <dirk.hovy@unibocconi.it>

Contributors¶

None yet. Why not be the first?

History¶

2.2.0 (2021-09-??)¶

introducing kitty
improving the documentation a lot

2.1.2 (2021-09-03)¶

patching Issue 38
improvements PR 80

2.1.0 (2021-07-16)¶

new model introduced SuperCTM
new model introduced β-CTM

2.0.0 (2021-xx-xx)¶

warning, breaking changes were introduced:
- the order of the parameters in CTMDataset was changed (now first is contextual embeddings)
- CTM takes in input bow_size, contextual_size instead of input_size and bert_size
- changed the name of the parameters in the dataset
introduced early stopping
introduced visualization with pyldavis

1.8.2 (2021-02-08)¶

removed constraint over pytorch version. This should solve problems for Windows users

1.8.0 (2021-01-11)¶

novel way to handle text, we now allow for an easy usage of training and testing data
better visualization of the training progress and of the sampling process
removed old stuff from the documentation

1.7.1 (2020-12-17)¶

some minor updates to the documentation
adding a new method to visualize the topic using a wordcloud
save and load will now generate a warning since the feature has not been tested

1.7.0 (2020-12-10)¶

adding a new and much simpler way to handle text for topic modeling

1.6.0 (2020-11-03)¶

introducing the two different classes for ZeroShotTM and CombinedTM
depracating CTM class in favor of ZeroShotTM and CombinedTM

1.5.3 (2020-11-03)¶

adding support for Windows encoding by defaulting file load to UTF-8

1.5.2 (2020-11-03)¶

updated sentence-transformers version to 0.3.6
beta support for model saving and loading
new evaluation metrics based on coherence

1.5.0 (2020-09-14)¶

Introduced a method to predict the topics for a set of documents (supports multiple sampling to reduce variation)
Adding some features to bert embeddings creation like increased batch size and progress bar
Supporting training directly from lists without the need to deal with files
Adding a simple quick preprocessing pipeline

1.4.3 (2020-09-03)¶

Updating sentence-transformers package to avoid errors

1.4.2 (2020-08-04)¶

Changed the encoding on file load for the SBERT embedding function

1.4.1 (2020-08-04)¶

Fixed bug over sparse matrices

1.4.0 (2020-08-01)¶

New feature handling sparse bow for optimized processing
New method to return topic distributions for words

1.0.0 (2020-04-05)¶

Released models with the main features implemented

0.1.0 (2020-04-04)¶

First release on PyPI.