Tag Archives: Natural Language Understanding

Measuring Gendered Correlations in Pre-trained NLP Models

Posted by Kellie Webster, Software Engineer, Google Research

Natural language processing (NLP) has seen significant progress over the past several years, with pre-trained models like BERT, ALBERT, ELECTRA, and XLNet achieving remarkable accuracy across a variety of tasks. In pre-training, representations are learned from a large text corpus, e.g., Wikipedia, by repeatedly masking out words and trying to predict them (this is called masked language modeling). The resulting representations encode rich information about language and correlations between concepts, such as surgeons and scalpels. There is then a second training stage, fine-tuning, in which the model uses task-specific training data to learn how to use the general pre-trained representations to do a concrete task, like classification. Given the broad adoption of these representations in many NLP tasks, it is crucial to understand the information encoded in them and how any learned correlations affect performance downstream, to ensure the application of these models aligns with our AI Principles.

In “Measuring and Reducing Gendered Correlations in Pre-trained Models” we perform a case study on BERT and its low-memory counterpart ALBERT, looking at correlations related to gender, and formulate a series of best practices for using pre-trained language models. We present experimental results over public model checkpoints and an academic task dataset to illustrate how the best practices apply, providing a foundation for exploring settings beyond the scope of this case study. We will soon release a series of checkpoints, Zari1, which reduce gendered correlations while maintaining state-of-the-art accuracy on standard NLP task metrics.

Measuring Correlations
To understand how correlations in pre-trained representations can affect downstream task performance, we apply a diverse set of evaluation metrics for studying the representation of gender. Here, we’ll discuss results from one of these tests, based on coreference resolution, which is the capability that allows models to understand the correct antecedent to a given pronoun in a sentence. For example, in the sentence that follows, the model should recognize his refers to the nurse, and not to the patient.

The standard academic formulation of the task is the OntoNotes test (Hovy et al., 2006), and we measure how accurate a model is at coreference resolution in a general setting using an F1 score over this data (as in Tenney et al. 2019). Since OntoNotes represents only one data distribution, we also consider the WinoGender benchmark that provides additional, balanced data designed to identify when model associations between gender and profession incorrectly influence coreference resolution. High values of the WinoGender metric (close to one) indicate a model is basing decisions on normative associations between gender and profession (e.g., associating nurse with the female gender and not male). When model decisions have no consistent association between gender and profession, the score is zero, which suggests that decisions are based on some other information, such as sentence structure or semantics.

BERT and ALBERT metrics on OntoNotes (accuracy) and WinoGender (gendered correlations). Low values on the WinoGender metric indicate that a model does not preferentially use gendered correlations in reasoning.

In this study, we see that neither the (Large) BERT or ALBERT public model achieves zero score on the WinoGender examples, despite achieving impressive accuracy on OntoNotes (close to 100%). At least some of this is due to models preferentially using gendered correlations in reasoning. This isn’t completely surprising: there are a range of cues available to understand text and it is possible for a general model to pick up on any or all of these. However, there is reason for caution, as it is undesirable for a model to make predictions primarily based on gendered correlations learned as priors rather than the evidence available in the input.

Best Practices
Given that it is possible for unintended correlations in pre-trained model representations to affect downstream task reasoning, we now ask: what can one do to mitigate any risk this poses when developing new NLP models?

  • It is important to measure for unintended correlations: Model quality may be assessed using accuracy metrics, but these only measure one dimension of performance, especially if the test data is drawn from the same distribution as the training data. For example, the BERT and ALBERT checkpoints have accuracy within 1% of each other, but differ by 26% (relative) in the degree to which they use gendered correlations for coreference resolution. This difference might be important for some tasks; selecting a model with low WinoGender score could be desirable in an application featuring texts about people in professions that may not conform to historical social norms, e.g., male nurses.
  • Be careful even when making seemingly innocuous configuration changes: Neural network model training is controlled by many hyperparameters that are usually selected to maximize some training objective. While configuration choices often seem innocuous, we find they can cause significant changes for gendered correlations, both for better and for worse. For example, dropout regularization is used to reduce overfitting by large models. When we increase the dropout rate used for pre-training BERT and ALBERT, we see a significant reduction in gendered correlations even after fine-tuning. This is promising since a simple configuration change allows us to train models with reduced risk of harm, but it also shows that we should be mindful and evaluate carefully when making any change in model configuration.
    Impact of increasing dropout regularization in BERT and ALBERT.
  • There are opportunities for general mitigations: A further corollary from the perhaps unexpected impact of dropout on gendered correlations is that it opens the possibility to use general-purpose methods for reducing unintended correlations: by increasing dropout in our study, we improve how the models reason about WinoGender examples without manually specifying anything about the task or changing the fine-tuning stage at all. Unfortunately, OntoNotes accuracy does start to decline as the dropout rate increases (which we can see in the BERT results), but we are excited about the potential to mitigate this in pre-training, where changes can lead to model improvements without the need for task-specific updates. We explore counterfactual data augmentation as another mitigation strategy with different tradeoffs in our paper.

What’s Next
We believe these best practices provide a starting point for developing robust NLP systems that perform well across the broadest possible range of linguistic settings and applications. Of course these techniques on their own are not sufficient to capture and remove all potential issues. Any model deployed in a real-world setting should undergo rigorous testing that considers the many ways it will be used, and implement safeguards to ensure alignment with ethical norms, such as Google's AI Principles. We look forward to developments in evaluation frameworks and data that are more expansive and inclusive to cover the many uses of language models and the breadth of people they aim to serve.

Acknowledgements
This is joint work with Xuezhi Wang, Ian Tenney, Ellie Pavlick, Alex Beutel, Jilin Chen, Emily Pitler, and Slav Petrov. We benefited greatly throughout the project from discussions with Fernando Pereira, Ed Chi, Dipanjan Das, Vera Axelrod, Jacob Eisenstein, Tulsee Doshi, and James Wexler.


1 Zari is an Afghan Muppet designed to show that ‘a little girl could do as much as everybody else’. ↩

Source: Google AI Blog


Measuring Gendered Correlations in Pre-trained NLP Models

Posted by Kellie Webster, Software Engineer, Google Research

Natural language processing (NLP) has seen significant progress over the past several years, with pre-trained models like BERT, ALBERT, ELECTRA, and XLNet achieving remarkable accuracy across a variety of tasks. In pre-training, representations are learned from a large text corpus, e.g., Wikipedia, by repeatedly masking out words and trying to predict them (this is called masked language modeling). The resulting representations encode rich information about language and correlations between concepts, such as surgeons and scalpels. There is then a second training stage, fine-tuning, in which the model uses task-specific training data to learn how to use the general pre-trained representations to do a concrete task, like classification. Given the broad adoption of these representations in many NLP tasks, it is crucial to understand the information encoded in them and how any learned correlations affect performance downstream, to ensure the application of these models aligns with our AI Principles.

In “Measuring and Reducing Gendered Correlations in Pre-trained Models” we perform a case study on BERT and its low-memory counterpart ALBERT, looking at correlations related to gender, and formulate a series of best practices for using pre-trained language models. We present experimental results over public model checkpoints and an academic task dataset to illustrate how the best practices apply, providing a foundation for exploring settings beyond the scope of this case study. We will soon release a series of checkpoints, Zari1, which reduce gendered correlations while maintaining state-of-the-art accuracy on standard NLP task metrics.

Measuring Correlations
To understand how correlations in pre-trained representations can affect downstream task performance, we apply a diverse set of evaluation metrics for studying the representation of gender. Here, we’ll discuss results from one of these tests, based on coreference resolution, which is the capability that allows models to understand the correct antecedent to a given pronoun in a sentence. For example, in the sentence that follows, the model should recognize his refers to the nurse, and not to the patient.

The standard academic formulation of the task is the OntoNotes test (Hovy et al., 2006), and we measure how accurate a model is at coreference resolution in a general setting using an F1 score over this data (as in Tenney et al. 2019). Since OntoNotes represents only one data distribution, we also consider the WinoGender benchmark that provides additional, balanced data designed to identify when model associations between gender and profession incorrectly influence coreference resolution. High values of the WinoGender metric (close to one) indicate a model is basing decisions on normative associations between gender and profession (e.g., associating nurse with the female gender and not male). When model decisions have no consistent association between gender and profession, the score is zero, which suggests that decisions are based on some other information, such as sentence structure or semantics.

BERT and ALBERT metrics on OntoNotes (accuracy) and WinoGender (gendered correlations). Low values on the WinoGender metric indicate that a model does not preferentially use gendered correlations in reasoning.

In this study, we see that neither the (Large) BERT or ALBERT public model achieves zero score on the WinoGender examples, despite achieving impressive accuracy on OntoNotes (close to 100%). At least some of this is due to models preferentially using gendered correlations in reasoning. This isn’t completely surprising: there are a range of cues available to understand text and it is possible for a general model to pick up on any or all of these. However, there is reason for caution, as it is undesirable for a model to make predictions primarily based on gendered correlations learned as priors rather than the evidence available in the input.

Best Practices
Given that it is possible for unintended correlations in pre-trained model representations to affect downstream task reasoning, we now ask: what can one do to mitigate any risk this poses when developing new NLP models?

  • It is important to measure for unintended correlations: Model quality may be assessed using accuracy metrics, but these only measure one dimension of performance, especially if the test data is drawn from the same distribution as the training data. For example, the BERT and ALBERT checkpoints have accuracy within 1% of each other, but differ by 26% (relative) in the degree to which they use gendered correlations for coreference resolution. This difference might be important for some tasks; selecting a model with low WinoGender score could be desirable in an application featuring texts about people in professions that may not conform to historical social norms, e.g., male nurses.
  • Be careful even when making seemingly innocuous configuration changes: Neural network model training is controlled by many hyperparameters that are usually selected to maximize some training objective. While configuration choices often seem innocuous, we find they can cause significant changes for gendered correlations, both for better and for worse. For example, dropout regularization is used to reduce overfitting by large models. When we increase the dropout rate used for pre-training BERT and ALBERT, we see a significant reduction in gendered correlations even after fine-tuning. This is promising since a simple configuration change allows us to train models with reduced risk of harm, but it also shows that we should be mindful and evaluate carefully when making any change in model configuration.
    Impact of increasing dropout regularization in BERT and ALBERT.
  • There are opportunities for general mitigations: A further corollary from the perhaps unexpected impact of dropout on gendered correlations is that it opens the possibility to use general-purpose methods for reducing unintended correlations: by increasing dropout in our study, we improve how the models reason about WinoGender examples without manually specifying anything about the task or changing the fine-tuning stage at all. Unfortunately, OntoNotes accuracy does start to decline as the dropout rate increases (which we can see in the BERT results), but we are excited about the potential to mitigate this in pre-training, where changes can lead to model improvements without the need for task-specific updates. We explore counterfactual data augmentation as another mitigation strategy with different tradeoffs in our paper.

What’s Next
We believe these best practices provide a starting point for developing robust NLP systems that perform well across the broadest possible range of linguistic settings and applications. Of course these techniques on their own are not sufficient to capture and remove all potential issues. Any model deployed in a real-world setting should undergo rigorous testing that considers the many ways it will be used, and implement safeguards to ensure alignment with ethical norms, such as Google's AI Principles. We look forward to developments in evaluation frameworks and data that are more expansive and inclusive to cover the many uses of language models and the breadth of people they aim to serve.

Acknowledgements
This is joint work with Xuezhi Wang, Ian Tenney, Ellie Pavlick, Alex Beutel, Jilin Chen, Emily Pitler, and Slav Petrov. We benefited greatly throughout the project from discussions with Fernando Pereira, Ed Chi, Dipanjan Das, Vera Axelrod, Jacob Eisenstein, Tulsee Doshi, and James Wexler.


1 Zari is an Afghan Muppet designed to show that ‘a little girl could do as much as everybody else’. ↩

Source: Google AI Blog


Language-Agnostic BERT Sentence Embedding

Posted by Yinfei Yang and Fangxiaoyu Feng, Software Engineers, Google Research

A multilingual embedding model is a powerful tool that encodes text from different languages into a shared embedding space, enabling it to be applied to a range of downstream tasks, like text classification, clustering, and others, while also leveraging semantic information for language understanding. Existing approaches for generating such embeddings, like LASER or m~USE, rely on parallel data, mapping a sentence from one language directly to another language in order to encourage consistency between the sentence embeddings. While these existing multilingual approaches yield good overall performance across a number of languages, they often underperform on high-resource languages compared to dedicated bilingual models, which can leverage approaches like translation ranking tasks with translation pairs as training data to obtain more closely aligned representations. Further, due to limited model capacity and the often poor quality of training data for low-resource languages, it can be difficult to extend multilingual models to support a larger number of languages while maintaining good performance.

Illustration of a multilingual embedding space.

Recent efforts to improve language models include the development of masked language model (MLM) pre-training, such as that used by BERT, ALBERT and RoBERTa. This approach has led to exceptional gains across a wide range of languages and a variety of natural language processing tasks since it only requires monolingual text. In addition, MLM pre-training has been extended to the multilingual setting by modifying MLM training to include concatenated translation pairs, known as translation language modeling (TLM), or by simply introducing pre-training data from multiple languages. However, while the internal model representations learned during MLM and TLM training are helpful when fine-tuning on downstream tasks, without a sentence level objective, they do not directly produce sentence embeddings, which are critical for translation tasks.

In “Language-agnostic BERT Sentence Embedding”, we present a multilingual BERT embedding model, called LaBSE, that produces language-agnostic cross-lingual sentence embeddings for 109 languages. The model is trained on 17 billion monolingual sentences and 6 billion bilingual sentence pairs using MLM and TLM pre-training, resulting in a model that is effective even on low-resource languages for which there is no data available during training. Further, the model establishes a new state of the art on multiple parallel text (a.k.a. bitext) retrieval tasks. We have released the pre-trained model to the community through tfhub, which includes modules that can be used as-is or can be fine-tuned using domain-specific data.

The collection of the training data for 109 supported languages

The Model
In previous work, we proposed the use of a translation ranking task to learn a multilingual sentence embedding space. This approach tasks the model with ranking the true translation over a collection of sentences in the target language, given a sentence in the source language. The translation ranking task is trained using a dual encoder architecture with a shared transformer encoder. The resulting bilingual models achieved state-of-the-art performance on multiple parallel text retrieval tasks (including United Nations and BUCC). However, the model suffered when the bi-lingual models were extended to support multiple languages (16 languages, in our test case) due to limitations in model capacity, vocabulary coverage, training data quality and more.

Translation ranking task. Given a sentence in a given source language, the task is to find the true translation over a collection of sentences in the target language.

For LaBSE, we leverage recent advances on language model pre-training, including MLM and TLM, on a BERT-like architecture and follow this with fine-tuning on a translation ranking task. A 12-layer transformer with a 500k token vocabulary pre-trained using MLM and TLM on 109 languages is used to increase the model and vocabulary coverage. The resulting LaBSE model offers extended support to 109 languages in a single model.

The dual encoder architecture, in which the source and target text are encoded using a shared transformer embedding network separately. The translation ranking task is applied, forcing the text that paraphrases each other to have similar representations. The transformer embedding network is initialized from a BERT checkpoint trained on MLM and TLM tasks.

Performance on Cross-lingual Text Retrieval
We evaluate the proposed model using the Tatoeba corpus, a dataset consisting of up to 1,000 English-aligned sentence pairs for 112 languages. For more than 30 of the languages in the dataset, the model has no training data. The model is tasked with finding the nearest neighbor translation for a given sentence, which it calculates using the cosine distance.

To understand the performance of the model for languages at the head or tail of the training data distribution, we divide the set of languages into several groups and compute the average accuracy for each set. The first 14-language group is selected from the languages supported by m~USE, which cover the languages from the head of the distribution (head languages). We also evaluate a second language group composed of 36 languages from the XTREME benchmark. The third 82-language group, selected from the languages covered by the LASER training data, includes many languages from the tail of the distribution (tail languages). Finally, we compute the average accuracy for all languages.

The table below presents the average accuracy achieved by LaBSE, compared to the m~USE and LASER models, for each language group. As expected, all models perform strongly on the 14-language group that covers most head languages. With more languages included, the averaged accuracy for both LASER and LaBSE declines. However, the reduction in accuracy from the LaBSE model with increasing numbers of languages is much less significant, outperforming LASER significantly, particularly when the full distribution of 112 languages is included (83.7% accuracy vs. 65.5%).

Model 14 Langs 36 Langs 82 Langs All Langs
m~USE* 93.9 — — —
LASER 95.3 84.4 75.9 65.5
LaBSE 95.3 95.0 87.3 83.7
Average Accuracy (%) on Tatoeba Datasets. The “14 Langs” group consists of languages supported by m~USE; the “36 Langs” group includes languages selected by XTREME; and the “82 Langs” group represents languages covered by the LASER model. The “All Langs” group includes all languages supported by Taoteba.
* The m~USE model comes in two varieties, one built on a convolutional neural network architecture and the other a Transformer-like architecture. Here, we compare only to the Transformer version.

Support to Unsupported Languages
The average performance of all languages included in Tatoeba is very promising. Interestingly, LaBSE even performs relatively well for many of the 30+ Tatoeba languages for which it has no training data (see below). For one third of these languages the LaBSE accuracy is higher than 75% and only 8 have accuracy lower than 25%, indicating very strong transfer performance to languages without training data. Such positive language transfer is only possible due to the massively multilingual nature of LaBSE.

LaBSE accuracy for the subset of Tatoeba languages (represented with ISO 639-1/639-2 codes) for which there was no training data.

Mining Parallel Text from WebLaBSE can be used for mining parallel text (bi-text) from web-scale data. For example, we applied LaBSE to CommonCrawl, a large-scale monolingual corpus, to process 560 million Chinese and 330 million German sentences for the extraction of parallel text. Each Chinese and German sentence pair is encoded using the LaBSE model and then the encoded embedding is used to find a potential translation from a pool of 7.7 billion English sentences pre-processed and encoded by the model. An approximate nearest neighbor search is employed to quickly search through the high-dimensional sentence embeddings. After a simple filtering, the model returns 261M and 104M potential parallel pairs for English-Chinese and English-German, respectively. The trained NMT model using the mined data reaches BLEU scores of 35.7 and 27.2 on the WMT translation tasks (wmt17 for English-to-Chinese and wmt14 for English-to-German). The performance is only a few points away from current state-of-art-models trained on high quality parallel data.

ConclusionWe're excited to share this research, and the model, with the community. The pre-trained model is released at tfhub to support further research on this direction and possible downstream applications. We also believe that what we're showing here is just the beginning, and there are more important research problems to be addressed, such as building better models to support all languages.

AcknowledgementsThe core team includes Wei Wang, Naveen Arivazhagan, Daniel Cer. We would like to thank the Google Research Language team, along with our partners in other Google groups for their feedback and suggestions. Special thanks goes to Sidharth Mudgal, and Jax Law for help with data processing; as well as Jialu Liu, Tianqi Liu, Chen Chen, and Anosh Raj for help on BERT pre-training.

Source: Google AI Blog


Grounding Natural Language Instructions to Mobile UI Actions

Posted by Yang Li, Research Scientist, Google Research

Mobile devices offer a myriad of functionalities that can assist in everyday activities. However, many of these functionalities are not easily discoverable or accessible to users, forcing users to look up how to perform a specific task -- how to turn on the traffic mode in Maps or change notification settings in YouTube, for example. While searching the web for detailed instructions for these questions is an option, it is still up to the user to follow these instructions step-by-step and navigate UI details through a small touchscreen, which can be tedious and time consuming, and results in reduced accessibility. What if one could design a computational agent to turn these language instructions into actions and automatically execute them on the user’s behalf?

In “Mapping Natural Language Instructions to Mobile UI Action Sequences”, published at ACL 2020, we present the first step towards addressing the problem of automatic action sequence mapping, creating three new datasets used to train deep learning models that ground natural language instructions to executable mobile UI actions. This work lays the technical foundation for task automation on mobile devices that would alleviate the need to maneuver through UI details, which may be especially valuable for users who are visually or situationally impaired. We have also open-sourced our model code and data pipelines through our GitHub repository, in order to spur further developments among the research community.

Constructing Language Grounding Models
People often provide one another with instructions in order to coordinate joint efforts and accomplish tasks involving complex sequences of actions, for example, following a recipe to bake a cake, or having a friend walk you through setting up a home network. Building computational agents able to help with similar interactions is an important goal that requires true language grounding in the environments in which the actions take place.

The learning task addressed here is to predict a sequence of actions for a mobile platform given a set of instructions, a sequence of screens produced as the system transitions from one screen to another, as well as the set of interactive elements on those screens. Training such a model end-to-end would require paired language-action data, which is difficult to acquire at a large scale.

Instead, we deconstruct the problem into two sequential steps: an action phrase-extraction step and a grounding step.
The workflow of grounding language instructions to executable actions.
The action phrase-extraction step identifies the operation, object and argument descriptions from multi-step instructions using a Transformer model with area attention for representing each description phrase. Area attention allows the model to attend to a group of adjacent words in the instruction (a span) as a whole for decoding a description.
The action phrase extraction model takes a word sequence of a natural language instruction and outputs a sequence of spans (denoted in red boxes) that indicate the phrases describing the operation, the object and the argument of each action in the task.
Next, the grounding step matches the extracted operation and object descriptions with a UI object on the screen. Again, we use a Transformer model, but in this case, it contextually represents UI objects and grounds object descriptions to them.
The grounding model takes the extracted spans as input and grounds them to executable actions, including the object an action is applied to, given the UI screen at each step during execution.
Results
To investigate the feasibility of this task and the effectiveness of our approach, we construct three new datasets to train and evaluate our model. The first dataset includes 187 multi-step English instructions for operating Pixel phones along their corresponding action-screen sequences and enables assessment of full task performance on naturally occurring instructions, which is used for testing end-to-end grounding quality. For action phrase extraction training and evaluation, we obtain English “how-to” instructions that can be found abundantly from the web and annotate phrases that describe each action. To train the grounding model, we synthetically generate 295K single-step commands to UI actions, covering 178K different UI objects across 25K mobile UI screens from a public android UI corpus.

A Transformer with area attention obtains 85.56% accuracy for predicting span sequences that completely match the ground truth. The phrase extractor and grounding model together obtain 89.21% partial and 70.59% complete accuracy for matching ground-truth action sequences on the more challenging task of mapping language instructions to executable actions end-to-end. We also evaluated alternative methods and representations of UI objects, such as using a graph convolutional network (GCN) or a feedforward network, and found those that can represent an object contextually in the screen lead to better grounding accuracy. The new datasets, models and results provide an important first step on the challenging problem of grounding natural language instructions to mobile UI actions.

Conclusion
This research, and language grounding in general, is an important step for translating multi-stage instructions into actions on a graphical user interface. Successful application of task automation to the UI domain has the potential to significantly improve accessibility, where language interfaces might help individuals who are visually impaired perform tasks with interfaces that are predicated on sight. This also matters for situational impairment when one cannot access a device easily while encumbered by tasks at hand.

By deconstructing the problem into action phrase extraction and language grounding, progress on either can improve full task performance and it alleviates the need to have language-action paired datasets, which are difficult to collect at scale. For example, action span extraction is related to both semantic role labeling and extraction of multiple facts from text and could benefit from innovations in span identification and multitask learning. Reinforcement learning that has been applied in previous grounding work may help improve out-of-sample prediction for grounding in UIs and improve direct grounding from hidden state representations. Although our datasets were based on Android UIs, our approach can be applied generally to instruction grounding on other user interface platforms. Lastly, our work provides a technical foundation for investigating user experiences in language-based human computer interaction.

Acknowledgements
Many thanks to my collaborators on this work at Google Research. Xin Zhou and Jiacong He contributed substantially to the data pipelines and the creation of the datasets. Yuan Zhang and Jason Baldridge provided much valuable advice for the project and contributed to the presentation of the work. Gang Li provided generous help for creating open-source datasets. Many thanks to Ashwin Kakarla, Muqthar Mohammad and Mohd Majeed for their help with the annotations.

Source: Google AI Blog


Presenting a Challenge and Workshop in Efficient Open-Domain Question Answering

Posted by Eunsol Choi, Visiting Faculty Researcher and Tom Kwiatkowski, Research Scientist, Google Research

One of the primary goals of natural language processing is to build systems that can answer a user's questions. To do this, computers need to be able to understand questions, represent world knowledge, and reason their way to answers. Traditionally, answers have been retrieved from a collection of documents or a knowledge graph. For example, to answer the question, “When was the declaration of independence officially signed?” a system might first find the most relevant article from Wikipedia, and then locate a sentence containing the answer, “August 2, 1776”. However, more recent approaches, like T5, have also shown that neural models, trained on large amounts of web-text, can also answer questions directly, without retrieving documents or facts from a knowledge graph. This has led to significant debate about how knowledge should be stored for use by our question answering systems — in human readable text and structured formats, or in the learned parameters of a neural network.

Today, we are proud to announce the EfficientQA competition and workshop at NeurIPS 2020, organized in cooperation with Princeton University and the University of Washington. The goal is to develop an end-to-end question answering system that contains all of the knowledge required to answer open-domain questions. There are no constraints on how the knowledge is stored — it could be in documents, databases, the parameters of a neural network, or any other form — but entries will be evaluated based on the number of bytes used to access this knowledge, including code, corpora, and model parameters. There will also be an unconstrained track, in which the goal is to achieve the best possible question answering performance regardless of system size. To build small, yet robust systems, participants will have to explore new methods of knowledge representation and reasoning.
An illustration of how the memory budget changes as a neural network and retrieval corpus grow and shrink. It is possible that successful systems will also use other resources such as a knowledge graph.
Competition Overview
The competition will be evaluated using the open-domain variant of the Natural Questions dataset. We will also provide further human evaluation of all the top performing entries to account for the fact that there are many correct ways to answer a question, not all of which will be covered by any set of reference answers. For example, for the question “What type of car is a Jeep considered?” both “off-road vehicles” and “crossover SUVs” are valid answers.

The competition is divided between four separate tracks: best performing system under 500 Mb; best performing system under 6 Gb; smallest system to get at least 25% accuracy; and the best performing system with no constraints. The winners of each of these tracks will be invited to present their work during the competition track at NeurIPS 2020, which will be hosted virtually. We will also put each of the winning systems up against human trivia experts (the 2017 NeurIPS Human-Computer competition featured Jeopardy! and Who Wants to Be a Millionaire champions) in a real-time contest at the virtual conference.

Participation
To participate, go to the competition site where you will find the data and evaluation code available for download, as well as dates and instructions on how to participate, and a sign-up form for updates. Along with our academic collaborators, we have provided some example systems to help you get started.

We believe that the field of natural language processing will benefit from a greater exploration and comparison of small system question answering options. We hope that by encouraging the development of very small systems, this competition will pave the way for on-device question answering.

Acknowledgements
Creating this challenge and workshop has been a large team effort including Adam Roberts, Colin Raffel, Chris Alberti, Jordan Boyd-Graber, Jennimaria Palomaki, Kenton Lee, Kelvin Guu, and Michael Collins from Google; as well as Sewon Min and Hannaneh Hajishirzi from the University of Washington; and Danqi Chen from Princeton University.

Source: Google AI Blog


PEGASUS: A State-of-the-Art Model for Abstractive Text Summarization

Posted by Peter J. Liu and Yao Zhao, Software Engineers, Google Research

Students are often tasked with reading a document and producing a summary (for example, a book report) to demonstrate both reading comprehension and writing ability. This abstractive text summarization is one of the most challenging tasks in natural language processing, involving understanding of long passages, information compression, and language generation. The dominant paradigm for training machine learning models to do this is sequence-to-sequence (seq2seq) learning, where a neural network learns to map input sequences to output sequences. While these seq2seq models were initially developed using recurrent neural networks, Transformer encoder-decoder models have recently become favored as they are more effective at modeling the dependencies present in the long sequences encountered in summarization.

Transformer models combined with self-supervised pre-training (e.g., BERT, GPT-2, RoBERTa, XLNet, ALBERT, T5, ELECTRA) have shown to be a powerful framework for producing general language learning, achieving state-of-the-art performance when fine-tuned on a wide array of language tasks. In prior work, the self-supervised objectives used in pre-training have been somewhat agnostic to the down-stream application in favor of generality; we wondered whether better performance could be achieved if the self-supervised objective more closely mirrored the final task.

In “PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization” (to appear at the 2020 International Conference on Machine Learning), we designed a pre-training self-supervised objective (called gap-sentence generation) for Transformer encoder-decoder models to improve fine-tuning performance on abstractive summarization, achieving state-of-the-art results on 12 diverse summarization datasets. Supplementary to the paper, we are also releasing the training code and model checkpoints on GitHub.

A Self-Supervised Objective for Summarization
Our hypothesis is that the closer the pre-training self-supervised objective is to the final down-stream task, the better the fine-tuning performance. In PEGASUS pre-training, several whole sentences are removed from documents and the model is tasked with recovering them. An example input for pre-training is a document with missing sentences, while the output consists of the missing sentences concatenated together. This is an incredibly difficult task that may seem impossible, even for people, and we don’t expect the model to solve it perfectly. However, such a challenging task encourages the model to learn about language and general facts about the world, as well as how to distill information taken from throughout a document in order to generate output that closely resembles the fine-tuning summarization task. The advantage of this self-supervision is that you can create as many examples as there are documents, without any human annotation, which is often the bottleneck in purely supervised systems.
A self-supervised example for PEGASUS during pre-training. The model is trained to output all the masked sentences.
We found that choosing “important” sentences to mask worked best, making the output of self-supervised examples even more similar to a summary. We automatically identified these sentences by finding those that were most similar to the rest of the document according to a metric called ROUGE. ROUGE computes the similarity of two texts by computing n-gram overlaps using a score from 0 to 100 (ROUGE-1, ROUGE-2, and ROUGE-L are three common variants).

Similar to other recent methods, such as T5, we pre-trained our model on a very large corpus of web-crawled documents, then we fine-tuned the model on 12 public down-stream abstractive summarization datasets, resulting in new state-of-the-art results as measured by automatic metrics, while using only 5% of the number of parameters of T5. The datasets were chosen to be diverse, including news articles, scientific papers, patents, short stories, e-mails, legal documents, and how-to directions, showing that the model framework is adaptive to a wide-variety of topics.

Fine-Tuning with Small Numbers of Examples
While PEGASUS showed remarkable performance with large datasets, we were surprised to learn that the model didn’t require a large number of examples for fine-tuning to get near state-of-the-art performance:
ROUGE scores (three variants, higher is better) vs. the number of supervised examples across four selected summarization datasets. The dotted-line shows the Transformer encoder-decoder performance with full-supervision, but without pre-training.
With only 1000 fine-tuning examples, we were able to perform better in most tasks than a strong baseline (Transformer encoder-decoder) that used the full supervised data, which in some cases had many orders of magnitude more examples. This “sample efficiency” greatly increases the usefulness of text summarization models as it significantly lowers the scale and cost of supervised data collection, which in the case of summarization is very expensive.

Human-Quality summaries
While we find automatic metrics such as ROUGE are useful proxies for measuring progress during model development, they only provide limited information and don’t tell us the whole story, such as fluency or a comparison to human performance. To this end, we conducted a human evaluation, where raters were asked to compare summaries from our model with human ones (without knowing which is which). This has some similarities to the Turing test.
Human raters were asked to rate model and human-written summaries without knowing which was which. The document is truncated here for illustration, but raters see the full text.
We performed the experiment with 3 different datasets and found that human raters do not consistently prefer the human summaries to those from our model. Furthermore, our models trained with only 1000 examples performed nearly as well. In particular, with the much studied XSum and CNN/Dailymail datasets, the model achieves human-like performance using only 1000 examples. This suggests large datasets of supervised examples are no longer necessary for summarization, opening up many low-cost use-cases.

A Test of Comprehension: Counting Ships
Following this post is an example article from the XSum dataset and the model-generated abstractive summary. As we can see, the model correctly abstracts and paraphrases four named frigates (HMS Cumberland, HMS Campbeltown, HMS Chatham and HMS Cornwall) as “four Royal Navy frigates”, something an extractive approach could not do since “four” is not mentioned anywhere. Was this a fluke or did the model actually count? One way to find out is to add and remove ships to see if the count changes.

As can be seen below, the model successfully “counts” ships from 2 to 5. However, when we add a sixth ship, the “HMS Alphabet”, it miscounts it as “seven”. So it appears the model has learned to count small numbers of items in a list, but does not yet generalize as elegantly as we would hope. Still, we think this rudimentary counting ability is impressive as it was not explicitly programmed into the model, and it demonstrates a limited amount of “symbolic reasoning” by the model.

PEGASUS code and model release
To support on-going research in this field and ensure reproducibility, we are releasing the PEGASUS code and model checkpoints on GitHub. This includes fine-tuning code which can be used to adapt PEGASUS to other summarization datasets.

Acknowledgements
This work has been a collaborative effort involving Jingqing Zhang, Yao Zhao, Mohammad Saleh, and Peter J. Liu. We thank the T5 and Google News teams for providing datasets for pre-training PEGASUS.

Source: Google AI Blog


PEGASUS: A State-of-the-Art Model for Abstractive Text Summarization

Posted by Peter J. Liu and Yao Zhao, Software Engineers, Google Research

Students are often tasked with reading a document and producing a summary (for example, a book report) to demonstrate both reading comprehension and writing ability. This abstractive text summarization is one of the most challenging tasks in natural language processing, involving understanding of long passages, information compression, and language generation. The dominant paradigm for training machine learning models to do this is sequence-to-sequence (seq2seq) learning, where a neural network learns to map input sequences to output sequences. While these seq2seq models were initially developed using recurrent neural networks, Transformer encoder-decoder models have recently become favored as they are more effective at modeling the dependencies present in the long sequences encountered in summarization.

Transformer models combined with self-supervised pre-training (e.g., BERT, GPT-2, RoBERTa, XLNet, ALBERT, T5, ELECTRA) have shown to be a powerful framework for producing general language learning, achieving state-of-the-art performance when fine-tuned on a wide array of language tasks. In prior work, the self-supervised objectives used in pre-training have been somewhat agnostic to the down-stream application in favor of generality; we wondered whether better performance could be achieved if the self-supervised objective more closely mirrored the final task.

In “PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization” (to appear at the 2020 International Conference on Machine Learning), we designed a pre-training self-supervised objective (called gap-sentence generation) for Transformer encoder-decoder models to improve fine-tuning performance on abstractive summarization, achieving state-of-the-art results on 12 diverse summarization datasets. Supplementary to the paper, we are also releasing the training code and model checkpoints on GitHub.

A Self-Supervised Objective for Summarization
Our hypothesis is that the closer the pre-training self-supervised objective is to the final down-stream task, the better the fine-tuning performance. In PEGASUS pre-training, several whole sentences are removed from documents and the model is tasked with recovering them. An example input for pre-training is a document with missing sentences, while the output consists of the missing sentences concatenated together. This is an incredibly difficult task that may seem impossible, even for people, and we don’t expect the model to solve it perfectly. However, such a challenging task encourages the model to learn about language and general facts about the world, as well as how to distill information taken from throughout a document in order to generate output that closely resembles the fine-tuning summarization task. The advantage of this self-supervision is that you can create as many examples as there are documents, without any human annotation, which is often the bottleneck in purely supervised systems.
A self-supervised example for PEGASUS during pre-training. The model is trained to output all the masked sentences.
We found that choosing “important” sentences to mask worked best, making the output of self-supervised examples even more similar to a summary. We automatically identified these sentences by finding those that were most similar to the rest of the document according to a metric called ROUGE. ROUGE computes the similarity of two texts by computing n-gram overlaps using a score from 0 to 100 (ROUGE-1, ROUGE-2, and ROUGE-L are three common variants).

Similar to other recent methods, such as T5, we pre-trained our model on a very large corpus of web-crawled documents, then we fine-tuned the model on 12 public down-stream abstractive summarization datasets, resulting in new state-of-the-art results as measured by automatic metrics, while using only 5% of the number of parameters of T5. The datasets were chosen to be diverse, including news articles, scientific papers, patents, short stories, e-mails, legal documents, and how-to directions, showing that the model framework is adaptive to a wide-variety of topics.

Fine-Tuning with Small Numbers of Examples
While PEGASUS showed remarkable performance with large datasets, we were surprised to learn that the model didn’t require a large number of examples for fine-tuning to get near state-of-the-art performance:
ROUGE scores (three variants, higher is better) vs. the number of supervised examples across four selected summarization datasets. The dotted-line shows the Transformer encoder-decoder performance with full-supervision, but without pre-training.
With only 1000 fine-tuning examples, we were able to perform better in most tasks than a strong baseline (Transformer encoder-decoder) that used the full supervised data, which in some cases had many orders of magnitude more examples. This “sample efficiency” greatly increases the usefulness of text summarization models as it significantly lowers the scale and cost of supervised data collection, which in the case of summarization is very expensive.

Human-Quality summaries
While we find automatic metrics such as ROUGE are useful proxies for measuring progress during model development, they only provide limited information and don’t tell us the whole story, such as fluency or a comparison to human performance. To this end, we conducted a human evaluation, where raters were asked to compare summaries from our model with human ones (without knowing which is which). This has some similarities to the Turing test.
Human raters were asked to rate model and human-written summaries without knowing which was which. The document is truncated here for illustration, but raters see the full text.
We performed the experiment with 3 different datasets and found that human raters do not consistently prefer the human summaries to those from our model. Furthermore, our models trained with only 1000 examples performed nearly as well. In particular, with the much studied XSum and CNN/Dailymail datasets, the model achieves human-like performance using only 1000 examples. This suggests large datasets of supervised examples are no longer necessary for summarization, opening up many low-cost use-cases.

A Test of Comprehension: Counting Ships
Following this post is an example article from the XSum dataset and the model-generated abstractive summary. As we can see, the model correctly abstracts and paraphrases four named frigates (HMS Cumberland, HMS Campbeltown, HMS Chatham and HMS Cornwall) as “four Royal Navy frigates”, something an extractive approach could not do since “four” is not mentioned anywhere. Was this a fluke or did the model actually count? One way to find out is to add and remove ships to see if the count changes.

As can be seen below, the model successfully “counts” ships from 2 to 5. However, when we add a sixth ship, the “HMS Alphabet”, it miscounts it as “seven”. So it appears the model has learned to count small numbers of items in a list, but does not yet generalize as elegantly as we would hope. Still, we think this rudimentary counting ability is impressive as it was not explicitly programmed into the model, and it demonstrates a limited amount of “symbolic reasoning” by the model.

PEGASUS code and model release
To support on-going research in this field and ensure reproducibility, we are releasing the PEGASUS code and model checkpoints on GitHub. This includes fine-tuning code which can be used to adapt PEGASUS to other summarization datasets.

Acknowledgements
This work has been a collaborative effort involving Jingqing Zhang, Yao Zhao, Mohammad Saleh, and Peter J. Liu. We thank the T5 and Google News teams for providing datasets for pre-training PEGASUS.

Source: Google AI Blog


An NLU-Powered Tool to Explore COVID-19 Scientific Literature

Posted by Keith Hall, Research Scientist, Natural Language Understanding, Google Research

Due to the COVID-19 pandemic, scientists and researchers around the world are publishing an immense amount of new research in order to understand and combat the disease. While the volume of research is very encouraging, it can be difficult for scientists and researchers to keep up with the rapid pace of new publications. Traditional search engines can be excellent resources for finding real-time information on general COVID-19 questions like "How many COVID-19 cases are there in the United States?", but can struggle with understanding the meaning behind research-driven queries. Furthermore, searching through the existing corpus of COVID-19 scientific literature with traditional keyword-based approaches can make it difficult to pinpoint relevant evidence for complex queries.

To help address this problem, we are launching the COVID-19 Research Explorer, a semantic search interface on top of the COVID-19 Open Research Dataset (CORD-19), which includes more than 50,000 journal articles and preprints. We have designed the tool with the goal of helping scientists and researchers efficiently pore through articles for answers or evidence to COVID-19-related questions.

When the user asks an initial question, the tool not only returns a set of papers (like in a traditional search) but also highlights snippets from the paper that are possible answers to the question. The user can review the snippets and quickly make a decision on whether or not that paper is worth further reading. If the user is satisfied with the initial set of papers and snippets, we have added functionality to pose follow-up questions, which act as new queries for the original set of retrieved articles. Take a look at the animation below to see an example of a query and a corresponding follow-up question. We hope these features will foster knowledge exploration and efficient gathering of evidence for scientific hypotheses.

Semantic Search
A key technology powering the tool is semantic search. Semantic search aims to not just capture term overlap between a query and a document, but to really understand whether the meaning of a phrase is relevant to the user’s true intent behind their query.

Consider the query, “What regulates ACE2 expression?” Even though this seems like a simple question, certain phrases can still confuse a search engine that relies solely on text matching. For example, “regulates” can refer to a number of biological processes. While traditional information retrieval (IR) systems use techniques like query expansion to mitigate this confusion, semantic search models aim to learn these relationships implicitly.

Word order also matters. ACE2 (angiotensin converting enzyme-2) itself regulates certain biological processes, but the question is actually asking what regulates ACE2. Matching on terms alone will not distinguish between “what regulates ACE2 ” and “what ACE2 regulates.” Traditional IR systems use tricks like n-gram term matching, but semantic search methods strive to model word order and semantics at their core.

The semantic search technology we use is powered by BERT, which has recently been deployed to improve retrieval quality of Google Search. For the COVID-19 Research Explorer we faced the challenge that biomedical literature uses a language that is very different from the kinds of queries submitted to Google.com. In order to train BERT models, we required supervision — examples of queries and their relevant documents and snippets. While we relied on excellent resources produced by BioASQ for fine-tuning, such human-curated datasets tend to be small. Neural semantic search models require large amounts of training data. To augment small human-constructed datasets, we used advances in query generation to build a large synthetic corpus of questions and relevant documents in the biomedical domain.

Specifically, we used large amounts of general domain question-answer pairs to train an encoder-decoder model (part a in the figure below). This kind of neural architecture is used in tasks like machine translation that encodes one piece of text (e.g., an English sentence) and produces another piece of text (e.g., a French sentence). Here we trained the model to translate from answer passages to questions (or queries) about that passage. Next we took passages from every document in the collection, in this case CORD-19, and generated corresponding queries (part b). We then used these synthetic query-passage pairs as supervision to train our neural retrieval model (part c).
Synthetic query construction.
However, we found that there were examples where the neural model performed worse than a keyword-based model. This is because of the memorization-generalization continuum, which is well known in most fields of artificial intelligence and psycholinguistics. Keyword-based models, like tf-idf, are essentially memorizers. They memorize terms from the query and look for documents that have them. Neural retrieval models, on the other hand, learn generalizations about concepts and meaning and try to match based on those. Sometimes they can over-generalize when precision is important. For example, if I query, “What regulates ACE2 expression?”, one may want the model to generalize the concept of “regulation,” but not ACE2 beyond acronym expansion.

Hybrid Term-Neural Retrieval Model
To improve our system we built a hybrid term-neural retrieval model. A crucial observation is that both term-based and neural models can be cast as a vector space model. In other words, we can encode both the query and documents and then treat retrieval as looking for the document vectors that are most similar to the query vector, also known as k-nearest neighbor retrieval. There is a lot of research and engineering that is needed to make this work at scale, but it allows us a simple mechanism to combine methods. The simplest approach is to combine the vectors with a trade-off parameter.
Hybrid Term and Neural Retrieval.
In the figure above, the blue boxes are the term-based vectors, and the red, the neural vectors. We represent documents by concatenating these vectors. We concatenate the two vectors for queries as well, but we control the relative importance of exact term matches versus neural semantic matching. This is done via a weight parameter k. While more complex hybrid schemes are possible, we found that this simple hybrid model significantly increased quality on our biomedical literature retrieval benchmarks.

Availability and Community Feedback
The COVID-19 Research Explorer is freely available to the research community as an open alpha. Over the coming months we will be making a number of usability enhancements, so please check back often. Try out the COVID-19 Research Explorer, and please share any comments you have with us via the feedback channels on the site.

Acknowledgements
This effort has been successful thanks to the hard work of many people, including, but not limited to the following (in alphabetical order of last name): John Alex, Waleed Ammar, Greg Billock, Yale Cong, Ali Elkahky, Daniel Francisco, Stephen Greco, Stefan Hosein, Johanna Katz, Gyorgy Kiss, Margarita Kopniczky, Ivan Korotkov, Dominic Leung, Daphne Luong, Ji Ma, Ryan Mcdonald, Matt Pearson-Beck, Biao She, Jonathan Sheffi, Kester Tong, Ben Wedin

Source: Google AI Blog


An NLU-Powered Tool to Explore COVID-19 Scientific Literature

Posted by Keith Hall, Research Scientist, Natural Language Understanding, Google Research

Due to the COVID-19 pandemic, scientists and researchers around the world are publishing an immense amount of new research in order to understand and combat the disease. While the volume of research is very encouraging, it can be difficult for scientists and researchers to keep up with the rapid pace of new publications. Traditional search engines can be excellent resources for finding real-time information on general COVID-19 questions like "How many COVID-19 cases are there in the United States?", but can struggle with understanding the meaning behind research-driven queries. Furthermore, searching through the existing corpus of COVID-19 scientific literature with traditional keyword-based approaches can make it difficult to pinpoint relevant evidence for complex queries.

To help address this problem, we are launching the COVID-19 Research Explorer, a semantic search interface on top of the COVID-19 Open Research Dataset (CORD-19), which includes more than 50,000 journal articles and preprints. We have designed the tool with the goal of helping scientists and researchers efficiently pore through articles for answers or evidence to COVID-19-related questions.

When the user asks an initial question, the tool not only returns a set of papers (like in a traditional search) but also highlights snippets from the paper that are possible answers to the question. The user can review the snippets and quickly make a decision on whether or not that paper is worth further reading. If the user is satisfied with the initial set of papers and snippets, we have added functionality to pose follow-up questions, which act as new queries for the original set of retrieved articles. Take a look at the animation below to see an example of a query and a corresponding follow-up question. We hope these features will foster knowledge exploration and efficient gathering of evidence for scientific hypotheses.

Semantic Search
A key technology powering the tool is semantic search. Semantic search aims to not just capture term overlap between a query and a document, but to really understand whether the meaning of a phrase is relevant to the user’s true intent behind their query.

Consider the query, “What regulates ACE2 expression?” Even though this seems like a simple question, certain phrases can still confuse a search engine that relies solely on text matching. For example, “regulates” can refer to a number of biological processes. While traditional information retrieval (IR) systems use techniques like query expansion to mitigate this confusion, semantic search models aim to learn these relationships implicitly.

Word order also matters. ACE2 (angiotensin converting enzyme-2) itself regulates certain biological processes, but the question is actually asking what regulates ACE2. Matching on terms alone will not distinguish between “what regulates ACE2 ” and “what ACE2 regulates.” Traditional IR systems use tricks like n-gram term matching, but semantic search methods strive to model word order and semantics at their core.

The semantic search technology we use is powered by BERT, which has recently been deployed to improve retrieval quality of Google Search. For the COVID-19 Research Explorer we faced the challenge that biomedical literature uses a language that is very different from the kinds of queries submitted to Google.com. In order to train BERT models, we required supervision — examples of queries and their relevant documents and snippets. While we relied on excellent resources produced by BioASQ for fine-tuning, such human-curated datasets tend to be small. Neural semantic search models require large amounts of training data. To augment small human-constructed datasets, we used advances in query generation to build a large synthetic corpus of questions and relevant documents in the biomedical domain.

Specifically, we used large amounts of general domain question-answer pairs to train an encoder-decoder model (part a in the figure below). This kind of neural architecture is used in tasks like machine translation that encodes one piece of text (e.g., an English sentence) and produces another piece of text (e.g., a French sentence). Here we trained the model to translate from answer passages to questions (or queries) about that passage. Next we took passages from every document in the collection, in this case CORD-19, and generated corresponding queries (part b). We then used these synthetic query-passage pairs as supervision to train our neural retrieval model (part c).
Synthetic query construction.
However, we found that there were examples where the neural model performed worse than a keyword-based model. This is because of the memorization-generalization continuum, which is well known in most fields of artificial intelligence and psycholinguistics. Keyword-based models, like tf-idf, are essentially memorizers. They memorize terms from the query and look for documents that have them. Neural retrieval models, on the other hand, learn generalizations about concepts and meaning and try to match based on those. Sometimes they can over-generalize when precision is important. For example, if I query, “What regulates ACE2 expression?”, one may want the model to generalize the concept of “regulation,” but not ACE2 beyond acronym expansion.

Hybrid Term-Neural Retrieval Model
To improve our system we built a hybrid term-neural retrieval model. A crucial observation is that both term-based and neural models can be cast as a vector space model. In other words, we can encode both the query and documents and then treat retrieval as looking for the document vectors that are most similar to the query vector, also known as k-nearest neighbor retrieval. There is a lot of research and engineering that is needed to make this work at scale, but it allows us a simple mechanism to combine methods. The simplest approach is to combine the vectors with a trade-off parameter.
Hybrid Term and Neural Retrieval.
In the figure above, the blue boxes are the term-based vectors, and the red, the neural vectors. We represent documents by concatenating these vectors. We concatenate the two vectors for queries as well, but we control the relative importance of exact term matches versus neural semantic matching. This is done via a weight parameter k. While more complex hybrid schemes are possible, we found that this simple hybrid model significantly increased quality on our biomedical literature retrieval benchmarks.

Availability and Community Feedback
The COVID-19 Research Explorer is freely available to the research community as an open alpha. Over the coming months we will be making a number of usability enhancements, so please check back often. Try out the COVID-19 Research Explorer, and please share any comments you have with us via the feedback channels on the site.

Acknowledgements
This effort has been successful thanks to the hard work of many people, including, but not limited to the following (in alphabetical order of last name): John Alex, Waleed Ammar, Greg Billock, Yale Cong, Ali Elkahky, Daniel Francisco, Stephen Greco, Stefan Hosein, Johanna Katz, Gyorgy Kiss, Margarita Kopniczky, Ivan Korotkov, Dominic Leung, Daphne Luong, Ji Ma, Ryan Mcdonald, Matt Pearson-Beck, Biao She, Jonathan Sheffi, Kester Tong, Ben Wedin

Source: Google AI Blog


Using Neural Networks to Find Answers in Tables

Posted by Thomas Müller, Software Engineer, Google Research

Much of the world’s information is stored in the form of tables, which can be found on the web or in databases and documents. These might include anything from technical specifications of consumer products to financial and country development statistics, sports results and much more. Currently, one needs to manually look at these tables to find the answer to a question or rely on a service that gives answers to specific questions (e.g., about sports results). This information would be much more accessible and useful if it could be queried through natural language.

For example, the following figure shows a table with a number of questions that people might want to ask. The answer to these questions might be found in one, or multiple, cells in a table (“Which wrestler had the most number of reigns?”), or might require aggregating multiple table cells (“How many world champions are there with only one reign?”).
A table and questions with the expected answers. Answers can be selected (#1, #4) or computed (#2, #3).
Many recent approaches apply traditional semantic parsing to this problem, where a natural language question is translated to an SQL-like database query that executes against a database to provide the answers. For example, the question “How many world champions are there with only one reign?” would be mapped to a query such as “select count(*) where column("No. of reigns") == 1;” and then executed to produce the answer. This approach often requires substantial engineering in order to generate syntactically and semantically valid queries and is difficult to scale to arbitrary questions rather than questions about very specific tables (such as sports results).

In, “TAPAS: Weakly Supervised Table Parsing via Pre-training”, accepted at ACL 2020, we take a different approach that extends the BERT architecture to encode the question jointly along with tabular data structure, resulting in a model that can then point directly to the answer. Instead of creating a model that works only for a single style of table, this approach results in a model that can be applied to tables from a wide range of domains. After pre-training on millions of Wikipedia tables, we show that our approach exhibits competitive accuracy on three academic table question-answering (QA) datasets. Additionally, in order to facilitate more exciting research in this area, we have open-sourced the code for training and testing the models as well as the models pre-trained on Wikipedia tables, available at our GitHub repo.

How to Process a Question
To process a question such as “Average time as champion for top 2 wrestlers?”, our model jointly encodes the question as well as the table content row by row using a BERT model that is extended with special embeddings to encode the table structure.

The key addition to the transformer-based BERT model are the extra embeddings that are used to encode the structured input. We rely on learned embeddings for the column index, the row index and one special rank index, which indicates the order of elements in numerical columns. The following image shows how all of these are added together at the input and fed into the transformer layers. The figure below illustrates how the question is encoded, together with the small table shown on the left. Each cell token has a special embedding that indicates its row, column and numeric rank within the column.
BERT layer input: Every input token is represented as the sum of the embeddings of its word, absolute position, segment (whether it belongs to the question or table), column and row and numeric rank (the position the cell would have if the column was sorted by its numeric values).
The model has two outputs: 1) for each table cell, a score indicates the probability that this cell will be part of the answer and 2) an aggregation operation that indicates which operation (if any) is applied to produce the final answer. The following figure shows how, for the question “Average time as champion for top 2 wrestlers?”, the model should select the first two cells of the “Combined days” column and the “AVERAGE” operation with high probability.
Model schematic: The BERT layer encodes both the question and table. The model outputs a probability for every aggregation operation and a selection probability for every table cell. For the question “Average time as champion for top 2 wrestlers?” the AVERAGE operation and the cells with the numbers 3,749 and 3,103 should have a high probability.
Pre-training
Using a method similar to how BERT is trained on text, we pre-trained our model on 6.2 million table-text pairs extracted from the English Wikipedia. During pre-training, the model learns to restore words in both table and text that have been replaced with a mask. We find that the model can do this with relatively high accuracy (71.4% of the masked tokens are restored correctly for tables unseen during training).

Learning from Answers Only
During fine-tuning the model learns how to answer questions from a table. This is done through training with either strong or weak supervision. In the case of strong supervision, for a given table and questions, one must provide the cells and aggregation operation to select (e.g., sum or count), a time-consuming and laborious process. More commonly, one trains using weak supervision, where only the correct answer (e.g., 3426 for the question in the example above) is provided. In this case, the model attempts to find an aggregation operation and cells that produce an answer close to the correct answer. This is done by computing the expectation over all the possible aggregation decisions and comparing it with the true result. The weak supervision scenario is beneficial because it allows for non-experts to provide the data needed to train the model and takes less time than strong supervision.

Results
We applied our model to three datasets — SQA, WikiTableQuestions (WTQ) and WikiSQL — and compared it to the performance of the top three state-of-the-art (SOTA) models for parsing tabular data. The comparison models included Min et al (2019) for WikiSQL, Wang et al. (2019) for WTQ and our own previous work for SQA (Mueller et al., 2019). For all datasets, we report the answer accuracy on the test sets for the weakly supervised training setup. For SQA and WIkiSQL we used our base model pre-trained on Wikipedia, while for WTQ, we found it beneficial to additionally pre-train on the SQA data. Our best models outperform the previous SOTA for SQA by more than 12 points, the previous SOTA for WTQ by more than 4 points and performs similarly to the best model published on WikiSQL.
Test answer accuracy for the weakly-supervised setup on three academic TableQA datasets.
Acknowledgements
This work was carried out by Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller, Francesco Piccinno and Julian Martin Eisenschlos in the Google AI language group in Zurich. We would like to thank Yasemin Altun, Srini Narayanan, Slav Petrov, William Cohen, Massimo Nicosia, Syrine Krichene, and Jordan Boyd-Graber for useful comments and suggestions.

Source: Google AI Blog