Mar 25, 2016 · post

H.P. Luhn and the Heuristic Value of Simplicity


The Fast Forward Labs team is putting final touches on our Summarization research, which explains approaches to making text quantifiable and computable. Stay tuned for a series of resources on the topic, including an online talk May 24 where we’ll cover technical details and survey use cases for financial services, media, and professional services with Agolo. Sign up here!

In writing our reports, we try not only to inform readers about the libraries, math, and techniques they can use to put a system into production today, but also the lessons they can learn from historical approaches to a given topic. Turning a retrospective eye towards past work can be particularly helpful when using an algorithm like a recurrent neural network. That’s because neural networks are notoriously hard to interpret: feature engineering is left to the algorithm, and involves a complex interplay between the weight of individual computing nodes and the connections that link them together. In the face of this complexity, it can be helpful to keep former, more simple techniques in mind as a heuristic guide - or model - to shape our intuition about how contemporary algorithms work.

For example, we found that H.P. Luhn’s 1958 paper The Automatic Creation of Literary Abstracts provided a simple heuristic to help wrap our heads around the basic steps that go into probabilistic models for summarizing text. (For those interested in history, Luhn also wrote a paper about business intelligence in 1958 that feels like it could have been written today, as it highlights the growing need to automate information retrieval to manage an unwieldy amount of information.) Our design lead, Grant Custer, designed a prototype you can play with to walk through Luhn’s method. 

Here’s the link to access the live demo. Feel free to use the suggested text, or to play around your own (and share results on Twitter!). 

Luhn’s algorithm begins by transforming the content of sentences into a mathematical expression, or vector. He uses a “bag of words” model, which ignores filler words like “the” or “and”, and counts the number of times remaining words appear in each sentence.


Luhn’s intuition was that a word that appears many times in a document is important to the document’s meaning. Under this assumption, a sentence that contains many of the words that appear many times in the overall document is itself highly representative of that document. In our demo, if a document’s most significant words are protein (appears 8 times) and DNA (appears 7 times), then this implies that the sentence “Proteins are made by tiny machines in the cell called ribosomes” is a useful one to extract in the summary. Once this sentence scoring is complete, the last step is simply to select those sentences with the highest overall rankings. 


Luhn himself notes that his method for determining the relative significance of sentences “gives no consideration to the meaning of words or the arguments expressed by word combinations.” It is, rather, a “probabilistic one” based on counting. The algorithm does not select the sentence about ribosomes given its understanding of the importance ribosomes have to the topic in question; it selects that sentence because it densely packs together words that appear frequently across the longer document. 

The contemporary approaches we study in our upcoming report build upon this basic theme. One approach, topic modeling using Latent Dirichlet Allocation (LDA), groups together words that co-occur into mathematical expressions called topics and then represents documents as a short vector of different topics weights (e.g., 50% words from the the protein topic, 30% words from the DNA topic, and 20% from the gene topic). While the mathematical models that determine these topics are much more complex than simply counting a bag of words, LDA borrows Luhn’s basic insight: that we can quantify semantic meaning as the relative distribution of like items in a data set.

Stay tuned for more exciting language processing and deep learning resources throughout the spring! 

- Kathryn 

Read more

Mar 28, 2016 · announcement
Feb 24, 2016 · post

Latest posts

Nov 15, 2020 · post

Representation Learning 101 for Software Engineers

by Victor Dibia · Figure 1: Overview of representation learning methods. TLDR; Good representations of data (e.g., text, images) are critical for solving many tasks (e.g., search or recommendations). Deep representation learning yields state of the art results when used to create these representations. In this article, we review methods for representation learning and walk through an example using pretrained models. Introduction Deep Neural Networks (DNNs) have become a particularly useful tool in building intelligent systems that simplify cognitive tasks for users. more
Jun 22, 2020 · post

How to Explain HuggingFace BERT for Question Answering NLP Models with TF 2.0

by Victor · Given a question and a passage, the task of Question Answering (QA) focuses on identifying the exact span within the passage that answers the question. Figure 1: In this sample, a BERTbase model gets the answer correct (Achaemenid Persia). Model gradients show that the token “subordinate ..” is impactful in the selection of an answer to the question “Macedonia was under the rule of which country?". This makes sense .. good for BERTbase. more
Jun 16, 2020 · notebook

Evaluating QA: Metrics, Predictions, and the Null Response →

by Melanie · A deep dive into computing QA predictions and when to tell BERT to zip it! In our last post, Building a QA System with BERT on Wikipedia, we used the HuggingFace framework to train BERT on the SQuAD2.0 dataset and built a simple QA system on top of the Wikipedia search engine. This time, we’ll look at how to assess the quality of a BERT-like model for Question Answering.
May 19, 2020 · notebook

Building a QA System with BERT on Wikipedia →

by Melanie · So you’ve decided to build a QA system. You want to start with something simple and general so you plan to make it open domain using Wikipedia as a corpus for answering questions. You want to use the best NLP that your compute resources allow (you’re lucky enough to have access to a GPU) so you’re going to focus on the big, flashy Transformer models that are all the rage these days.
Apr 28, 2020 · notebook

Intro to Automated Question Answering →

by Melanie · Welcome to the first edition of the Cloudera Fast Forward blog on Natural Language Processing for Question Answering! Throughout this series, we’ll build a Question Answering (QA) system with off-the-shelf algorithms and libraries and blog about our process and what we find along the way. We hope to wind up with a beginning-to-end documentary that provides:
Apr 1, 2020 · newsletter

Enterprise Grade ML

by Shioulin · At Cloudera Fast Forward, one of the mechanisms we use to tightly couple machine learning research with application is through application development projects for both internal and external clients. The problems we tackle in these projects are wide ranging and cut across various industries; the end goal is a production system that translates data into business impact. What is Enterprise Grade Machine Learning? Enterprise grade ML, a term mentioned in a paper put forth by Microsoft, refers to ML applications where there is a high level of scrutiny for data handling, model fairness, user privacy, and debuggability. more

Popular posts

Oct 30, 2019 · newsletter
Exciting Applications of Graph Neural Networks
Nov 14, 2018 · post
Federated learning: distributed machine learning with data locality and privacy
Apr 10, 2018 · post
PyTorch for Recommenders 101
Oct 4, 2017 · post
First Look: Using Three.js for 2D Data Visualization
Aug 22, 2016 · whitepaper
Under the Hood of the Variational Autoencoder (in Prose and Code)
Feb 24, 2016 · post
"Hello world" in Keras (or, Scikit-learn versus Keras)


In-depth guides to specific machine learning capabilities


Machine learning prototypes and interactive notebooks


A usable library for question answering on large datasets.

Explain BERT for Question Answering Models

Tensorflow 2.0 notebook to explain and visualize a HuggingFace BERT for Question Answering model.

NLP for Question Answering

Ongoing posts and code documenting the process of building a question answering model.

Interpretability Revisited: SHAP and LIME

Explore how to use LIME and SHAP for interpretability.


Cloudera Fast Forward is an applied machine learning reseach group.
Cloudera   Blog   Twitter