This is an internal, incomplete preview of a proposed change to the ACL Anthology.
For efficiency reasons, we generate only three BibTeX files per volume, and the preview may be incomplete in other ways, or contain mistakes.
Do not treat this content as an official publication.
Large Language Models (LLMs) prompted to generate chain-of-thought (CoT) exhibit impressive reasoning capabilities. Recent attempts at prompt decomposition toward solving complex, multi-step reasoning problems depend on the ability of the LLM to simultaneously decompose and solve the problem. A significant disadvantage is that foundational LLMs are typically not available for fine-tuning, making adaptation computationally prohibitive. We believe (and demonstrate) that problem decomposition and solution generation are distinct capabilites, better addressed in separate modules, than by one monolithic LLM. We introduce DaSLaM, which uses a decomposition generator to decompose complex problems into subproblems that require fewer reasoning steps. These subproblems are answered by a solver. We use a relatively small (13B parameters) LM as the decomposition generator, which we train using policy gradient optimization to interact with a solver LM (regarded as black-box) and guide it through subproblems, thereby rendering our method solver-agnostic. Evaluation on multiple different reasoning datasets reveal that with our method, a 175 billion parameter LM (text-davinci-003) can produce competitive or even better performance, compared to its orders-of-magnitude larger successor, GPT-4. Additionally, we show that DaSLaM is not limited by the solver’s capabilities as a function of scale; e.g., solver LMs with diverse sizes give significant performance improvement with our solver-agnostic decomposition technique. Exhaustive ablation studies evince the superiority of our modular finetuning technique over exorbitantly large decomposer LLMs, based on prompting alone.
Existing Text-to-SQL generators require the entire schema to be encoded with the user text. This is expensive or impractical for large databases with tens of thousands of columns. Standard dense retrieval techniques are inadequate for schema subsetting of a large structured database, where the correct semantics of retrieval demands that we rank sets of schema elements rather than individual documents. In response, we propose a two-stage process for effective coverage during retrieval. First, we use an LLM to hallucinate a minimal DB schema that it deems adequate to answer the query. We use the hallucinated schema to retrieve a subset of the actual schema, by composing the results from multiple dense retrievals. Remarkably, hallucination — generally considered a nuisance — turns out to be actually useful as a bridging mechanism. Since no existing benchmarks exist for schema subsetting on large databases, we introduce two benchmarks: (1) A semi-synthetic dataset of 4502 schema elements, by taking a union of schema on the well-known SPIDER dataset, and (2) A real-life benchmark called SocialDB sourced from an actual large data warehouse comprising of 17844 schema elements. We show that our method leads to significantly higher recall than SOTA retrieval-based augmentation methods.
Multilingual language models (MLLMs) like mBERTpromise to extend the benefits of NLP research to low-resource languages (LRLs). However, LRL words are under-represented in the wordpiece/subword vocabularies of MLLMs. This leads to many LRL words getting replaced by UNK, or concatenated from morphologically unrelated wordpieces, leading to low task accuracy. (Pre)-training MLLMs after including LRL documents is resource-intensive in terms of both human inputs and computational resources. In response, we propose EVALM (entropy-based vocabulary augmented language model), which uses a new task-cognizant measurement to detect the most vulnerable LRL words, whose wordpiece segmentations are undesirable. EVALM then provides reasonable initializations of their embeddings, followed by limited fine-tuning using the small LRL task corpus. Our experiments show significant performance improvements and also some surprising limits to such vocabulary augmentation strategies in various classification tasks for multiple diverse LRLs, as well as code-mixed texts. We will release the code and data to enable further research.
Recent years have witnessed interest in Temporal Question Answering over Knowledge Graphs (TKGQA), resulting in the development of multiple methods. However, these are highly engineered, thereby limiting their generalizability, and they do not automatically discover relevant parts of the KG during multi-hop reasoning. Relational graph convolutional networks (RGCN) provide an opportunity to address both of these challenges – we explore this direction in the paper. Specifically, we propose a novel, intuitive and interpretable scheme to modulate the messages passed through a KG edge during convolution based on the relevance of its associated period to the question. We also introduce a gating device to predict if the answer to a complex temporal question is likely to be a KG entity or time and use this prediction to guide our scoring mechanism. We evaluate the resulting system, which we call TwiRGCN, on a recent challenging dataset for multi-hop complex temporal QA called TimeQuestions. We show that TwiRGCN significantly outperforms state-of-the-art models on this dataset across diverse question types. Interestingly, TwiRGCN improves accuracy by 9–10 percentage points for the most difficult ordinal and implicit question types.
Question answering (QA) over tables and linked text, also called TextTableQA, has witnessed significant research in recent years, as tables are often found embedded in documents along with related text. HybridQA and OTT-QA are the two best-known TextTableQA datasets, with questions that are best answered by combining information from both table cells and linked text passages. A common challenge in both datasets, and TextTableQA in general, is that the training instances include just the question and answer, where the gold answer may match not only multiple table cells across table rows but also multiple text spans within the scope of a table row and its associated text. This leads to a noisy multi-instance training regime. We present MITQA, a transformer-based TextTableQA system that is explicitly designed to cope with distant supervision along both these axes, through a multi-instance loss objective, together with careful curriculum design. Our experiments show that the proposed multi-instance distant supervision approach helps MITQA get sate-of-the-art results beating the existing baselines for both HybridQA and OTT-QA, putting MITQA at the top of HybridQA leaderboard with best EM and F1 scores on a held out test set.
Automated completion of open knowledge bases (Open KBs), which are constructed from triples of the form (subject phrase, relation phrase, object phrase), obtained via open information extraction (Open IE) system, are useful for discovering novel facts that may not be directly present in the text. However, research in Open KB completion (Open KBC) has so far been limited to resource-rich languages like English. Using the latest advances in multilingual Open IE, we construct the first multilingual Open KBC dataset, called mOKB6, containing facts from Wikipedia in six languages (including English). Improvingthe previous Open KB construction pipeline by doing multilingual coreference resolution andkeeping only entity-linked triples, we create a dense Open KB. We experiment with several models for the task and observe a consistent benefit of combining languages with the help of shared embedding space as well as translations of facts. We also observe that current multilingual models struggle to remember facts seen in languages of different scripts.
Table Question Answering (Table QA) systems have been shown to be highly accurate when trained and tested on open-domain datasets built on top of Wikipedia tables. However, it is not clear whether their performance remains the same when applied to domain-specific scientific and business documents, encountered in industrial settings, which exhibit some unique characteristics: (a) they contain tables with a much more complex layout than Wikipedia tables (including hierarchical row and column headers), (b) they contain domain-specific terms, and (c) they are typically not accompanied by domain-specific labeled data that can be used to train Table QA models. To understand the performance of Table QA approaches in this setting, we introduce AIT-QA; a domain-specific Table QA test dataset. While focusing on the airline industry, AIT-QA reflects the challenges that domain-specific documents pose to Table QA, outlined above. In this work, we describe the creation of the dataset and report zero-shot experimental results of three SOTA Table QA methods. The results clearly expose the limitations of current methods with a best accuracy of just 51.8%. We also present pragmatic table pre-processing steps to pivot and project complex tables into a layout suitable for the SOTA Table QA models. Finally, we provide data-driven insights on how different aspects of this setting (including hierarchical headers, domain-specific terminology, and paraphrasing) affect Table QA methods, in order to help the community develop improved methods for domain-specific Table QA.
Knowledge Graph Completion (KGC) predicts missing facts in an incomplete Knowledge Graph (KG). Multilingual KGs associate entities and relations with surface forms written in different languages. An entity or relation may be associated with distinct IDs in different KGs, necessitating entity alignment (EA) and relation alignment (RA). Many effective algorithms have been proposed for completion and alignment as separate tasks. Here we show that these tasks are synergistic and best solved together. Our multitask approach starts with a state-of-the-art KG embedding scheme, but adds a novel relation representation based on sets of embeddings of (subject, object) entity pairs. This representation leads to a new relation alignment loss term based on a maximal bipartite matching between two sets of embedding vectors. This loss is combined with traditional KGC loss and optionally, losses based on text embeddings of entity (and relation) names. In experiments over KGs in seven languages, we find that our system achieves large improvements in KGC compared to a strong completion model that combines known facts in all languages. It also outperforms strong EA and RA baselines, underscoring the value of joint alignment and completion.
Progress with supervised Open Information Extraction (OpenIE) has been primarily limited to English due to the scarcity of training data in other languages. In this paper, we explore techniques to automatically convert English text for training OpenIE systems in other languages. We introduce the Alignment-Augmented Constrained Translation (AACTrans) model to translate English sentences and their corresponding extractions consistently with each other — with no changes to vocabulary or semantic meaning which may result from independent translations. Using the data generated with AACTrans, we train a novel two-stage generative OpenIE model, which we call Gen2OIE, that outputs for each sentence: 1) relations in the first stage and 2) all extractions containing the relation in the second stage. Gen2OIE increases relation coverage using a training data transformation technique that is generalizable to multiple languages, in contrast to existing models that use an English-specific training loss. Evaluations on 5 languages — Spanish, Portuguese, Chinese, Hindi and Telugu — show that the Gen2OIE with AACTrans data outperforms prior systems by a margin of 6-25% in F1.
Relation classification (sometimes called ‘extraction’) requires trustworthy datasets for fine-tuning large language models, as well as for evaluation. Data collection is challenging for Indian languages, because they are syntactically and morphologically diverse, as well as different from resource-rich languages like English. Despite recent interest in deep generative models for Indian languages, relation classification is still not well-served by public data sets. In response, we present IndoRE, a dataset with 39K entity- and relation-tagged gold sentences in three Indian languages, plus English. We start with a multilingual BERT (mBERT) based system that captures entity span positions and type information and provides competitive monolingual relation classification. Using this system, we explore and compare transfer mechanisms between languages. In particular, we study the accuracy-efficiency tradeoff between expensive gold instances vs. translated and aligned ‘silver’ instances.
Temporal Knowledge Graphs (Temporal KGs) extend regular Knowledge Graphs by providing temporal scopes (start and end times) on each edge in the KG. While Question Answering over KG (KGQA) has received some attention from the research community, QA over Temporal KGs (Temporal KGQA) is a relatively unexplored area. Lack of broad coverage datasets has been another factor limiting progress in this area. We address this challenge by presenting CRONQUESTIONS, the largest known Temporal KGQA dataset, clearly stratified into buckets of structural complexity. CRONQUESTIONS expands the only known previous dataset by a factor of 340x. We find that various state-of-the-art KGQA methods fall far short of the desired performance on this new dataset. In response, we also propose CRONKGQA, a transformer-based solution that exploits recent advances in Temporal KG embeddings, and achieves performance superior to all baselines, with an increase of 120% in accuracy over the next best performing method. Through extensive experiments, we give detailed insights into the workings of CRONKGQA, as well as situations where significant further improvements appear possible. In addition to the dataset, we have released our code as well.
Weakly-supervised table question-answering (TableQA) models have achieved state-of-art performance by using pre-trained BERT transformer to jointly encoding a question and a table to produce structured query for the question. However, in practical settings TableQA systems are deployed over table corpora having topic and word distributions quite distinct from BERT’s pretraining corpus. In this work we simulate the practical topic shift scenario by designing novel challenge benchmarks WikiSQL-TS and WikiTable-TS, consisting of train-dev-test splits in five distinct topic groups, based on the popular WikiSQL and WikiTable-Questions datasets. We empirically show that, despite pre-training on large open-domain text, performance of models degrades significantly when they are evaluated on unseen topics. In response, we propose T3QA (Topic Transferable Table Question Answering) a pragmatic adaptation framework for TableQA comprising of: (1) topic-specific vocabulary injection into BERT, (2) a novel text-to-text transformer generator (such as T5, GPT2) based natural language question generation pipeline focused on generating topic-specific training data, and (3) a logical form re-ranker. We show that T3QA provides a reasonably good baseline for our topic shift benchmarks. We believe our topic split benchmarks will lead to robust TableQA solutions that are better suited for practical deployment
While traditional systems for Open Information Extraction were statistical and rule-based, recently neural models have been introduced for the task. Our work builds upon CopyAttention, a sequence generation OpenIE model (Cui et. al. 18). Our analysis reveals that CopyAttention produces a constant number of extractions per sentence, and its extracted tuples often express redundant information. We present IMoJIE, an extension to CopyAttention, which produces the next extraction conditioned on all previously extracted tuples. This approach overcomes both shortcomings of CopyAttention, resulting in a variable number of diverse extractions per sentence. We train IMoJIE on training data bootstrapped from extractions of several non-neural systems, which have been automatically filtered to reduce redundancy and noise. IMoJIE outperforms CopyAttention by about 18 F1 pts, and a BERT-based strong baseline by 2 F1 pts, establishing a new state of the art for the task.
Research on temporal knowledge bases, which associate a relational fact (s,r,o) with a validity time period (or time instant), is in its early days. Our work considers predicting missing entities (link prediction) and missing time intervals (time prediction) as joint Temporal Knowledge Base Completion (TKBC) tasks, and presents TIMEPLEX, a novel TKBC method, in which entities, relations and, time are all embedded in a uniform, compatible space. TIMEPLEX exploits the recurrent nature of some facts/events and temporal interactions between pairs of relations, yielding state-of-the-art results on both prediction tasks. We also find that existing TKBC models heavily overestimate link prediction performance due to imperfect evaluation mechanisms. In response, we propose improved TKBC evaluation protocols for both link and time prediction tasks, dealing with subtle issues that arise from the partial overlap of time intervals in gold instances and system predictions.
A recent state-of-the-art neural open information extraction (OpenIE) system generates extractions iteratively, requiring repeated encoding of partial outputs. This comes at a significant computational cost. On the other hand,sequence labeling approaches for OpenIE are much faster, but worse in extraction quality. In this paper, we bridge this trade-off by presenting an iterative labeling-based system that establishes a new state of the art for OpenIE, while extracting 10x faster. This is achieved through a novel Iterative Grid Labeling (IGL) architecture, which treats OpenIE as a 2-D grid labeling task. We improve its performance further by applying coverage (soft) constraints on the grid at training time. Moreover, on observing that the best OpenIE systems falter at handling coordination structures, our OpenIE system also incorporates a new coordination analyzer built with the same IGL architecture. This IGL based coordination analyzer helps our OpenIE system handle complicated coordination structures, while also establishing a new state of the art on the task of coordination analysis, with a 12.3 pts improvement in F1 over previous analyzers. Our OpenIE system - OpenIE6 - beats the previous systems by as much as 4 pts in F1, while being much faster.
State-of-the-art NLP inference uses enormous neural architectures and models trained for GPU-months, well beyond the reach of most consumers of NLP. This has led to one-size-fits-all public API-based NLP service models by major AI companies, serving millions of clients. They cannot afford traditional fine tuning for individual clients. Many clients cannot even afford significant fine tuning, and own little or no labeled data. Recognizing that word usage and salience diversity across clients leads to reduced accuracy, we initiate a study of practical and lightweight adaptation of centralized NLP services to clients. Each client uses an unsupervised, corpus-based sketch to register to the service. The server modifies its network mildly to accommodate client sketches, and occasionally trains the augmented network over existing clients. When a new client registers with its sketch, it gets immediate accuracy benefits. We demonstrate the proposed architecture using sentiment labeling, NER, and predictive language modeling.
Recent years have seen increasingly complex question-answering on knowledge bases (KBQA) involving logical, quantitative, and comparative reasoning over KB subgraphs. Neural Program Induction (NPI) is a pragmatic approach toward modularizing the reasoning process by translating a complex natural language query into a multi-step executable program. While NPI has been commonly trained with the ‘‘gold’’ program or its sketch, for realistic KBQA applications such gold programs are expensive to obtain. There, practically only natural language queries and the corresponding answers can be provided for training. The resulting combinatorial explosion in program space, along with extremely sparse rewards, makes NPI for KBQA ambitious and challenging. We present Complex Imperative Program Induction from Terminal Rewards (CIPITR), an advanced neural programmer that mitigates reward sparsity with auxiliary rewards, and restricts the program space to semantically correct programs using high-level constraints, KB schema, and inferred answer type. CIPITR solves complex KBQA considerably more accurately than key-value memory networks and neural symbolic machines (NSM). For moderately complex queries requiring 2- to 5-step programs, CIPITR scores at least 3× higher F1 than the competing systems. On one of the hardest class of programs (comparative reasoning) with 5–10 steps, CIPITR outperforms NSM by a factor of 89 and memory networks by 9 times.
Given a small corpus D_T pertaining to a limited set of focused topics, our goal is to train embeddings that accurately capture the sense of words in the topic in spite of the limited size of D_T. These embeddings may be used in various tasks involving D_T. A popular strategy in limited data settings is to adapt pretrained embeddings E trained on a large corpus. To correct for sense drift, fine-tuning, regularization, projection, and pivoting have been proposed recently. Among these, regularization informed by a word’s corpus frequency performed well, but we improve upon it using a new regularizer based on the stability of its cooccurrence with other words. However, a thorough comparison across ten topics, spanning three tasks, with standardized settings of hyper-parameters, reveals that even the best embedding adaptation strategies provide small gains beyond well-tuned baselines, which many earlier comparisons ignored. In a bold departure from adapting pretrained embeddings, we propose using D_T to probe, attend to, and borrow fragments from any large, topic-rich source corpus (such as Wikipedia), which need not be the corpus used to pretrain embeddings. This step is made scalable and practical by suitable indexing. We reach the surprising conclusion that even limited corpus augmentation is more useful than adapting embeddings, which suggests that non-dominant sense information may be irrevocably obliterated from pretrained embeddings and cannot be salvaged by adaptation.
Multilingual writers and speakers often alternate between two languages in a single discourse. This practice is called “code-switching”. Existing sentiment detection methods are usually trained on sentiment-labeled monolingual text. Manually labeled code-switched text, especially involving minority languages, is extremely rare. Consequently, the best monolingual methods perform relatively poorly on code-switched text. We present an effective technique for synthesizing labeled code-switched text from labeled monolingual text, which is relatively readily available. The idea is to replace carefully selected subtrees of constituency parses of sentences in the resource-rich language with suitable token spans selected from automatic translations to the resource-poor language. By augmenting the scarce labeled code-switched text with plentiful synthetic labeled code-switched text, we achieve significant improvements in sentiment labeling accuracy (1.5%, 5.11% 7.20%) for three different language pairs (English-Hindi, English-Spanish and English-Bengali). The improvement is even significant in hatespeech detection whereby we achieve a 4% improvement using only synthetic code-switched data (6% with data augmentation).
State-of-the-art knowledge base completion (KBC) models predict a score for every known or unknown fact via a latent factorization over entity and relation embeddings. We observe that when they fail, they often make entity predictions that are incompatible with the type required by the relation. In response, we enhance each base factorization with two type-compatibility terms between entity-relation pairs, and combine the signals in a novel manner. Without explicit supervision from a type catalog, our proposed modification obtains up to 7% MRR gains over base models, and new state-of-the-art results on several datasets. Further analysis reveals that our models better represent the latent types of entities and their embeddings also predict supervised types better than the embeddings fitted by baseline models.