Large language models have shown impressive abilities to reason over input text, however, they are prone to hallucinations. On the other hand, end-to-end knowledge graph question answering (KGQA) models output responses grounded in facts, but they still struggle with complex reasoning, such as comparison or ordinal questions. In this paper, we propose a new method for complex question answering where we combine a knowledge graph retriever based on an end-to-end KGQA model with a language model that reasons over the retrieved facts to return an answer. We observe that augmenting language model prompts with retrieved KG facts improves performance over using a language model alone by an average of 83%. In particular, we see improvements on complex questions requiring count, intersection, or multi-hop reasoning operations.

We introduce Mintaka, a complex, natural, and multilingual dataset designed for experimenting with end-to-end question-answering models. Mintaka is composed of 20,000 question-answer pairs collected in English, annotated with Wikidata entities, and translated into Arabic, French, German, Hindi, Italian, Japanese, Portuguese, and Spanish for a total of 180,000 samples. Mintaka includes 8 types of complex questions, including superlative, intersection, and multi-hop questions, which were naturally elicited from crowd workers. We run baselines over Mintaka, the best of which achieves 38% hits@1 in English and 31% hits@1 multilingually, showing that existing models have room for improvement. We release Mintaka at https://github.com/amazon-research/mintaka.

Speech disfluencies are prevalent in spontaneous speech. The rising popularity of voice assistants presents a growing need to handle naturally occurring disfluencies. Semantic parsing is a key component for understanding user utterances in voice assistants, yet most semantic parsing research to date focuses on written text. In this paper, we investigate semantic parsing of disfluent speech with the ATIS dataset. We find that a state-of-the-art semantic parser does not seamlessly handle disfluencies. We experiment with adding real and synthetic disfluencies at training time and find that adding synthetic disfluencies not only improves model performance by up to 39% but can also outperform adding real disfluencies in the ATIS dataset.

Recently, end-to-end (E2E) trained models for question answering over knowledge graphs (KGQA) have delivered promising results using only a weakly supervised dataset. However, these models are trained and evaluated in a setting where hand-annotated question entities are supplied to the model, leaving the important and non-trivial task of entity resolution (ER) outside the scope of E2E learning. In this work, we extend the boundaries of E2E learning for KGQA to include the training of an ER component. Our model only needs the question text and the answer entities to train, and delivers a stand-alone QA model that does not require an additional ER component to be supplied during runtime. Our approach is fully differentiable, thanks to its reliance on a recent method for building differentiable KGs (Cohen et al., 2020). We evaluate our E2E trained model on two public datasets and show that it comes close to baseline models that use hand-annotated entities.

End-to-end question answering using a differentiable knowledge graph is a promising technique that requires only weak supervision, produces interpretable results, and is fully differentiable. Previous implementations of this technique (Cohen et al, 2020) have focused on single-entity questions using a relation following operation. In this paper, we propose a model that explicitly handles multiple-entity questions by implementing a new intersection operation, which identifies the shared elements between two sets of entities. We find that introducing intersection improves performance over a baseline model on two datasets, WebQuestionsSP (69.6% to 73.3% Hits@1) and ComplexWebQuestions (39.8% to 48.7% Hits@1), and in particular, improves performance on questions with multiple entities by over 14% on WebQuestionsSP and by 19% on ComplexWebQuestions.

Speech disfluencies have been hypothesized to occur before words that are less predictable and therefore more cognitively demanding. In this paper, we revisit this hypothesis by using OpenAI’s GPT-2 to calculate predictability of words as language model perplexity. Using the Switchboard corpus, we find that 51% of disfluencies occur at the highest, second highest, or within one token of the highest perplexity, and this distribution is not random. We also show that disfluencies precede words with significantly higher perplexity than fluent contexts. Based on our results, we offer new evidence that disfluencies are more likely to occur before less predictable words.

Collecting training data for semantic parsing is a time-consuming and expensive task. As a result, there is growing interest in industry to reduce the number of annotations required to train a semantic parser, both to cut down on costs and to limit customer data handled by annotators. In this paper, we propose uncertainty and traffic-aware active learning, a novel active learning method that uses model confidence and utterance frequencies from customer traffic to select utterances for annotation. We show that our method significantly outperforms baselines on an internal customer dataset and the Facebook Task Oriented Parsing (TOP) dataset. On our internal dataset, our method achieves the same accuracy as random sampling with 2,000 fewer annotations.

While models have reached superhuman performance on popular question answering (QA) datasets such as SQuAD, they have yet to outperform humans on the task of question answering itself. In this paper, we investigate if models are learning reading comprehension from QA datasets by evaluating BERT-based models across five datasets. We evaluate models on their generalizability to out-of-domain examples, responses to missing or incorrect data, and ability to handle question variations. We find that no single dataset is robust to all of our experiments and identify shortcomings in both datasets and evaluation methods. Following our analysis, we make recommendations for building future QA datasets that better evaluate the task of question answering through reading comprehension. We also release code to convert QA datasets to a shared format for easier experimentation at https://github.com/amazon-research/qa-dataset-converter