Dominican Republic

Large language models (LM) generate remarkably fluent text and can be efficiently adapted across NLP tasks. Measuring and guaranteeing the quality of generated text in terms of safety is imperative for deploying LMs in the real world; to this end, prior work often relies on automatic evaluation of LM toxicity. We critically discuss this approach, evaluate several toxicity mitigation strategies with respect to both automatic and human evaluation, and analyze consequences of toxicity mitigation in terms of model bias and LM quality. We demonstrate that while basic intervention strategies can effectively optimize previously established automatic metrics on the REALTOXICITYPROMPTS dataset, this comes at the cost of reduced LM coverage for both texts about, and dialects of, marginalized groups. Additionally, we find that human raters often disagree with high automatic toxicity scores after strong toxicity reduction interventions— highlighting further the nuances involved in careful evaluation of LM toxicity.

EMNLP 2021

Challenges in Detoxifying Language Models

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible. The experience for on-site participants would closely approximate a normal pre-COVID *ACL conference, with 5-6 thematically organized parallel sessions and live Q/A and interactive discussion immediately after the talks. Presentations by virtual participants will be equitably interleaved with those of on-site participants, projected on the auditorium screens as if on-site, and also followed immediately by live Q/A and interactive discussion at a time during reasonable waking hours for the virtual presenter. For all participants, on-site and virtual, who are unable to attend a session due to either time-zone issues or because they are participating in another session live, talk recordings and slides will be available online at a minimum after the live presentation (and in many cases before as well), and questions may be submitted in advance on session-specific discussion boards and answered live in session with the usual visual aids if desired.

<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/ceby/"  seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>

Please Note: The EMNLP registration system is not currently connected to the underline site as we are still in the process of building out EMNLP 2021. You will receive access instructions from underline the week of November 1st. 

Access is given only to EMNLP upon registration, if you have not registered please do so [here](https://2021.emnlp.org/registration).

Registered attendees will receive access the week of November 1st.

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible.

workshop paper

Natural language processing systems such as dialogue agents should be able to reason about other people's beliefs, intentions and desires. This capability, called theory of mind (ToM), is crucial, as it allows a model to predict and interpret the needs of users based on their mental states. A recent line of research evaluates the ToM capability of existing memory-augmented neural models through question-answering. These models perform poorly on false belief tasks where beliefs differ from reality, especially when the dataset contains distracting sentences. In this paper, we propose a new temporally informed approach for improving the ToM capability of memory-augmented neural models. Our model incorporates priors about the entities' minds and tracks their mental states as they evolve over time through an extended passage. It then responds to queries through textual time travel---i.e., by accessing the stored memory of an earlier time step. We evaluate our model on ToM datasets and find that this approach improves performance, particularly by correcting the predicted mental states to match the false belief.

Textual Time Travel: A Temporally Informed Approach to Theory of Mind

Jupyter notebook allows data scientists to write machine learning code together with its documentation in cells. 
In this paper, we propose a new task of code documentation generation (CDG) for computational notebooks. In contrast to the previous CDG tasks which focus on generating documentation for single code snippets, in a computational notebook, one documentation in a markdown cell often corresponds to multiple code cells, 
and these code cells have an inherent structure. We proposed a new model HAConvGNN that uses a hierarchical attention mechanism to consider the relevant code cells and the relevant code tokens information when generating the documentation. Tested on a new corpus constructed from well-documented Kaggle notebooks, we show that our model outperforms other baseline models.

HAConvGNN: Hierarchical Attention Based Convolutional Graph Neural Network for Code Documentation Generation in Jupyter Notebooks

Large pre-trained language models have repeatedly shown their ability to produce fluent text. Yet even when starting from a prompt, generation can continue in many plausible directions. Current decoding methods with the goal of controlling generation, e.g., to ensure specific words are included, either require additional models or fine-tuning, or work poorly when the task at hand is semantically unconstrained, e.g., story generation. In this work, we present a plug-and-play decoding method for controlled language generation that is so simple and intuitive, it can be described in a single sentence: given a topic or keyword, we add a shift to the probability distribution over our vocabulary towards semantically similar words. We show how annealing this distribution can be used to impose hard constraints on language generation, something no other plug-and-play method is currently able to do with SOTA language generators. Despite the simplicity of this approach, we see it works incredibly well in practice: decoding from GPT-2 leads to diverse and fluent sentences while guaranteeing the appearance of given guide words. We perform two user studies, revealing that (1) our method outperforms competing methods in human evaluations; and (2) forcing the guide words to appear in the generated text has no impact on the fluency of the generated text.

A Plug-and-Play Method for Controlled Text Generation

Online conversations include more than just text. Increasingly, image-based responses such as memes and animated gifs serve as culturally recognized and often humorous responses in conversation. However, while NLP has broadened to multimodal models, conversational dialog systems have largely focused only on generating text replies. Here, we introduce a new dataset of 1.56M text-gif conversation turns and introduce a new multimodal conversational model Pepe the King Prawn for selecting gif-based replies. We demonstrate that our model produces relevant and high-quality gif responses and, in a large randomized control trial of multiple models replying to real users, we show that our model replies with gifs that are significantly better received by the community.

An animated picture says at least a thousand words: Selecting Gif-based Replies in Multimodal Dialog

The ability to detect Out-of-Domain (OOD) inputs has been a critical requirement in many real-world NLP applications. For example, intent classification in dialogue systems. The reason is that the inclusion of unsupported OOD inputs may lead to catastrophic failure of systems. However, it remains an empirical question whether current methods can tackle such problems reliably in a realistic scenario where zero OOD training data is available. In this study, we propose ProtoInfoMax, a new architecture that extends Prototypical Networks to simultaneously process in-domain and OOD sentences via Mutual Information Maximization (InfoMax) objective. Experimental results show that our proposed method can substantially improve performance up to 20% for OOD detection in low resource settings of text classification. We also show that ProtoInfoMax is less prone to typical overconfidence errors of Neural Networks, leading to more reliable prediction results.

ProtoInfoMax: Prototypical Networks with Mutual Information Maximization for Out-of-Domain Detection

While state-of-the-art Dialogue State Tracking (DST) models show promising results, all of them rely on a traditional cross-entropy loss function during the training process, which may not be optimal for improving the joint goal accuracy. Although several approaches recently proposed augmenting the training set by copying user utterances and replacing the real slot values with other possible or even similar values, they are not effective at improving the performance of existing DST models. To address these challenges, we propose a Turn-based Loss Function (TLF) that penalises the model if it inaccurately predicts a slot value at the early turns more so than in later turns in order to improve joint goal accuracy. We also propose a simple but effective Sequential Data Augmentation (SDA) algorithm to generate more complex user utterances and system responses to effectively train existing DST models. Experimental results on two standard DST benchmark collections demonstrate that our proposed TLF and SDA techniques significantly improve the effectiveness of the state-of-the-art DST model by approximately 7-8% relative reduction in error and achieves a new state-of-the-art joint goal accuracy with 59.50 and 54.90 on MultiWOZ2.1 and MultiWOZ2.2, respectively.

Improving Dialogue State Tracking with Turn-based Loss Function and Sequential Data Augmentation

Most progress in machine learning has been measured according to gains in top line metrics. However, top-line metrics are often not correlated with other design objectives that we would like our models to fulfill. How can we explain how deep neural networks arrive at decisions and audit when there are generalization failures? This talk explores some open challenges in deep neural network interpretability and will discuss desirable properties methods should fulfill in order to build trust between humans and algorithm.

Through the Looking-Glass: Understanding Model Behavior

I survey work on linking Large Language Models to brain imaging data. I argue 'interpretability' is crucial for using this work to reveal more about how the human brain processes language. I then survey some of our contributions to interpretability methods, and discuss the relevance of these methods for the big questions from cognitive neuroscience.

Language, Brains & Interpretability

Human explanations are responses to "why P rather than Q?" where P is an observed event and Q an imagined, counterfactual event that did not occur. Since people assign human-like traits to AI models, they expect explanations of models' behavior to follow the same conceptual framework used to explain human behavior. The consequence of not meeting this expectation is that people using AI technology do not have agency and hence one of the XAI objectives is not achieved. NLP researchers are accepting the perspective of the social and behavioral sciences as evident by the number of papers that have been published in the last and this year on contrastive explanations. In this talk, I will present a categorization of methods for producing contrastive explanations of NLP models and go into details of one such method.

Contrastive Explanations of NLP Models

Research in financial economics utilizing natural language processing has modeled how U.S. companies compete across product markets, how they innovate, and how supply chains are organized. This presentation will summarize these models, how they have evolved to the present day, and new work in progress. A key technical foundation is that these concepts of market structure, which were developed by economic theorists, have relatedness and overlap as their very intellectual foundation. Hence NLP methods (new and old) that focus on textual relatedness make up an ideal toolkit for analyzing these economic concepts, which are also central to many applications aimed at understanding and predicting business success.
One of the most compelling properties of an NLP approach is that market structures can be updated and redrawn every year at low cost as textual corpora are updated. This observation indicates the use of NLP to measure innovation. This has led to many insights on how companies manage growth, whether and when they go public, and how they manage the risk of disruption. This work continues to evolve with a focus on innovation, disaggregating complex product portfolios, and to build NLP measures for private firms through their websites. Among other economic findings, this work reveals that U.S. companies are becoming increasingly complex and broad in their product offerings, a result that is not seen using standard numerical databases in the public domain. Work using these same methods has also been deployed in related settings such as modeling systemic risk.

Challenges in Detoxifying Language Models

Downloads

Next from EMNLP 2021

Textual Time Travel: A Temporally Informed Approach to Theory of Mind

Similar lecture

"I don't know who she is": Discourse and Knowledge Driven Coreference Resolution

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Challenges in Detoxifying Language Models

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from EMNLP 2021

Textual Time Travel: A Temporally Informed Approach to Theory of Mind

Similar lecture

"I don't know who she is": Discourse and Knowledge Driven Coreference Resolution

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads