United States

Recent methods based on pre-trained language models have exhibited superior performance over tabular tasks (e.g. tabular NLI), despite showing inherent problems with reasoning over the tabular data (Gupta et al., 2021). In this work, we utilize Pattern-Exploiting Training (PET) to strengthen pre-existing knowledge
and reasoning abilities of these tabular reasoning models’. Compared to current baselines, our upgraded model exhibits a superior understanding of knowledge facts and tabular reasoning. Additionally, we demonstrate that such models are more effective for underlying downstream tasks of tabular inference on InfoTabS. Furthermore, we also show our model’s robustness against adversarial sets generated through various character and word level perturbations on InfoTabS

NAACL 2022

Masked Measurement Prediction: Learning to Jointly Predict Quantities and Units from Textual Context

## Welcome to NAACL 2022!
The Annual Meeting of the North American Chapter of the Association for Computational Linguistics (NAACL) is taking place July 10-15, 2022 as a hybrid event, in Seattle, WA and online. We are happy to welcome all of you to conference! 
 
The main conference program features oral presentations, in-person and virtual posters and demo sessions, a plenary session for our best paper presentations and awards, keynote presentations, a plenary panel on the place of linguistics and symbolic structures in NLP, and an Industrial Track panel on Careers in NLP.
 
Posters (including Findings of NAACL 2022) and demos are grouped by areas for both the in-person and the virtual sessions. For the virtual component, the talks will be on Zoom and the posters and the demos will be in GatherTown. The Student Research Workshop and will have an oral session and a poster session.
 
The program also features 6 Tutorials and 26 Workshops.
 
 We wish you a wonderful conference!
 
[The NAACL 2022 Organizing Committee](https://2022.naacl.org/committees/organization/)

2022 Annual Conference of the North American Chapter of the Association for Computational Linguistics

poster

Within the emerging research efforts to combine structured and unstructured knowledge, many approaches incorporate factual knowledge, e.g., available in form of structured knowledge graphs (KGs), into pre-trained language models (PLMs) and then apply the knowledge-enhanced PLMs to downstream NLP tasks. However, (1) they typically only consider \textit{static} factual knowledge, whereas, e.g., knowledge graphs (KGs) also contain \textit{temporal facts} or \textit{events} indicating evolutionary relationships among entities at different timestamps. (2) PLMs cannot be directly applied to many KG tasks, such as temporal KG completion. In this paper, we focus on \textbf{e}nhancing temporal knowledge embeddings with \textbf{co}ntextualized \textbf{la}nguage representations (ECOLA). We align structured knowledge, contained in temporal knowledge graphs, with their textual descriptions extracted from news articles, and propose a novel knowledge-text prediction task to inject the abundant information from descriptions into temporal knowledge embeddings. ECOLA jointly optimizes the knowledge-text prediction objective and the temporal knowledge embeddings, which can simultaneously take full advantage of textual and knowledge information. The proposed fusion method is model-agnostic and can be combined with potentially any temporal KG model. For training ECOLA, we introduce three temporal KG datasets with aligned textual descriptions. Experimental results on the temporal knowledge graph completion task show that ECOLA outperforms state-of-the-art temporal KG models by a large margin. The proposed datasets can serve as new temporal KG benchmarks and facilitate future research on structured and unstructured knowledge integration.

Multi-Hop Open-Domain Question Answering over Structured and Unstructured Knowledge

Recent work has shown that predicate-argument label representations from semantic role labeling (SRL) can be concatenated with BERT representations to improve natural language understanding tasks such as natural language inference (NLI) and reading comprehension. Two natural questions that arise are whether infusing SRL representations with BERT 1) improves model performance and 2) increases the model’s linguistic awareness. This paper aims at answering both questions with a case study on the NLI task. We start by analyzing whether and how infusing SRL information helps BERT learn linguistic knowledge. We compare model performance on two benchmark datasets, SNLI and MNLI. We also conduct in-depth analysis on two probing datasets, Breaking NLI and HANS, which contain abundant examples where SRL information is expected to be helpful. We found that combining SRL representations with BERT representations does outperform BERT-only representations in general, with better awareness in lexical meaning and world knowledge but not in logic knowledge. We also found that infusing SRL information via predicate-wise concatenation with BERT word representations followed by an interaction layer is more effective than sentence-wise concatenation.

Is Semantic-aware BERT more Linguistically Aware? A Case Study on Natural Language Inference

The semantics of relations are crucial for comprehending and analyzing multi-relational data. Polysemous relations between different types of entities (i.e. those that represent multiple semantics) are common in real-world relational datasets represented by knowledge graphs. For numerous use cases, such as entity type classification, question answering and knowledge graph completion, the correct semantic interpretation of these relations is necessary. In this work, we provide a method for discovering the different semantics associated with abstract relations and deriving many sub-relations with fine-grained meaning. To do this, we leverage the types of the entities associated with the relations and cluster the vector representations of entities and relations. The suggested method automatically discovers the best number of sub-relations for a polysemous relation and determine their semantic interpretation, according to our empirical evaluation.

Semantics-driven Refinement for Relational Data

Our collective attention span is shortened by the flood of online information. We address the need for automated claim validation based on the unstructured, aggregated information, continuously derived from multiple online news sources. We introduce an entity-centric reasoning framework in which latent connections between events, actions, or statements are revealed via entity mentions and are structurally integrated in a graph database. Using entity linking and semantic similarity, we collect and combine information from diverse sources, in order to generate evidence relevant to the user's claim. We then leverage textual entailment recognition to quantitatively determine whether this assertion is credible, based on the created evidence. Our approach fills the gap in automated claim validation for less-resourced languages, while being showcased for the Greek language and is complemented by the training of STS and NLI models which are evaluated on translated versions of common benchmarks.

Entity-centric Reasoning and Claim Validation based on Dynamic Textually Represented Environments

We propose a method to teach an automated agent to learn how to search for multi-hop paths of relations between entities in an open domain. The method learns a policy for directing existing information retrieval and machine reading resources to focus on relevant regions of a corpus. The approach formulates the learning problem as a Markov decision process with a state representation that encodes the dynamics of the search process and a reward structure that minimizes the number of documents that must be processed while still finding multi-hop paths. We implement the method in an actor-critic reinforcement learning algorithm
and evaluate it on a dataset of search problems derived from a subset of English Wikipedia.
The algorithm finds a family of policies that succeeds in extracting the desired information while processing fewer documents compared to several baseline heuristic algorithms.

Learning Open Domain Multi-hop Search Using Reinforcement Learning

Tables are an important form of structured data for both human and machine readers alike, providing answers to questions that cannot, or cannot easily, be found in texts. Recent work has designed special models and training paradigms for table-related tasks such as table-based question answering and table retrieval. Though effective, they add complexity in both modeling and data acquisition compared to generic text solutions and obscure which elements are truly beneficial. In this work, we focus on the task of table retrieval, and ask: “is table-specific model design necessary for table retrieval, or can a simpler text-based model be effectively used to achieve a similar result?” First, we perform an analysis on a table-based portion of the Natural Questions dataset (NQtable), and find that structure plays a negligible role in more than 70% of the cases. Based on this, we experiment with a general Dense Passage Retriever (DPR) based on text and a specialized Dense Table Retriever (DTR) that uses table-specific model designs. We find that DPR performs well without any table-specific design and training, and even achieves superior results compared to DTR when fine-tuned on properly linearized tables. We then experiment with three modules to explicitly encode table structures, namely auxiliary row/column embeddings, hard attention masks, and soft relation-based attention biases. However, none of these yielded significant improvements, suggesting that table-specific model design may not be necessary for table retrieval.

Table Retrieval May Not Necessitate Table-specific Model Design

Multi-hop question answering (QA) combines multiple pieces of evidence to search for the correct answer. Reasoning over a text corpus (TextQA) and/or a knowledge base (KBQA) has been extensively studied and led to distinct system architectures. However, knowledge transfer between such two QA systems has been under-explored. Research questions like what knowledge is transferred or whether the transferred knowledge can help answer over one source using another one, are yet to be answered. In this work, therefore, we study the knowledge transfer of multi-hop reasoning between structured and unstructured sources. We first propose a unified QA framework named SimultQA to enable knowledge transfer and bridge the distinct supervisions from KB and text sources. Then, we conduct extensive analyses to explore how knowledge is transferred by leveraging the pre-training and fine-tuning paradigm. We focus on the low-resource fine-tuning to show that pre-training SimultQA on one source can substantially improve its performance on the other source. More fine-grained analyses on transfer behaviors reveal the types of transferred knowledge and transfer patterns. We conclude with insights into how to construct better QA datasets and systems to exploit knowledge transfer for future work.

Knowledge Transfer between Structured and Unstructured Sources for Complex Question Answering

Recent methods based on pre-trained language models have exhibited superior performance over tabular tasks (e.g. tabular NLI), despite showing inherent problems with reasoning over the tabular data (Gupta et al., 2021). In this work, we utilize Pattern-Exploiting Training (PET) to strengthen pre-existing knowledge
and reasoning abilities of these tabular reasoning models’. Compared to current baselines, our upgraded model exhibits a superior understanding of knowledge facts and tabular reasoning. Additionally, we demonstrate that such models are more effective for underlying downstream tasks of tabular inference on InfoTabS. Furthermore, we also show our model’s robustness against adversarial sets generated through various character and word level perturbations on InfoTabS

Enhancing Tabular Reasoning with Pattern Exploiting Training

Prior work on constructing challenging tabular inference data centered primarily on human annotation or automatic synthetic generation. Both techniques have their own set of issues. Human annotation, despite its diversity and superior reasoning, struggles from scaling concerns. Synthetic data, on the other hand, despite its scalability, suffers from lack of linguistic and reasoning diversity. In this paper, we address both of these concerns by presenting a recasting approach that semi-automatically generates tabular NLI instances. We transform the table2text dataset ToTTo (Parikh et al., 2020) into a tabular NLI dataset using our proposed
framework. We demonstrate the use of our recasted data as an evaluation benchmark as well as augmentation data to improve performance on TabFact (Chen et al., 2020b). Furthermore, we test the effectiveness of models trained on
our data on the TabFact benchmark in the zeroshot scenario.

Framework for Recasting Table-to-Text Generation Datafor Tabular Inference

When humans conceive how to perform a particular task, they do so hierarchically: splitting higher-level tasks into smaller sub-tasks. How- ever, in the literature on natural language (NL) command of situated agents, most works have treated the procedures to be executed as flat sequences of simple actions, or any hierarchies of procedures have been shallow at best. In this paper, we propose a formalism of procedures as programs, a powerful yet intuitive method of representing hierarchical procedural knowledge for agent command and control. We further propose a modeling paradigm of hierarchical modular networks, which consist of a planner and reactors that convert NL intents to predictions of executable programs and probe the environment for information necessary to complete the program execution. We instantiate this framework on the IQA and ALFRED datasets for NL instruction following. Our model outperforms reactive baselines by a large margin on both datasets. We also demonstrate that our framework is more data-efficient, and that it allows for fast iterative development.

Masked Measurement Prediction: Learning to Jointly Predict Quantities and Units from Textual Context

Downloads

Next from NAACL 2022

Multi-Hop Open-Domain Question Answering over Structured and Unstructured Knowledge

Similar lecture

Reasoning Like Program Executors