Feature importance is commonly used to explain machine predictions. While feature importance can be derived from a machine learning model with a variety of methods, the consistency of feature importance via different methods remains understudied. In this work, we systematically compare feature importance from built-in mechanisms in a model such as attention values and post-hoc methods that approximate model behavior such as LIME. Using text classification as a testbed, we find that 1) no matter which method we use, important features from traditional models such as SVM and XGBoost are more similar with each other, than with deep learning models; 2) post-hoc methods tend to generate more similar important features for two models than built-in methods. We further demonstrate how such similarity varies across instances. Notably, important features do not always resemble each other better when two models agree on the predicted label than when they disagree.

Recently, end-to-end ASR based either on sequence-to-sequence networks or on the CTC objective function gained a lot of interest from the community, achieving competitive results over traditional systems using robust but complex pipelines. One of the main features of end-to-end systems, in addition to the ability to free themselves from extra linguistic resources such as dictionaries or language models, is the capacity to model acoustic units such as characters, subwords or directly words; opening up the capacity to directly translate speech with different representations or levels of knowledge depending on the target language. In this paper we propose a review of the existing end-to-end ASR approaches for the French language. We compare results to conventional state-of-the-art ASR systems and discuss which units are more suited to model the French language.

A quality abstractive summary should not only copy salient source texts as summaries but should also tend to generate new conceptual words to express concrete details. Inspired by the popular pointer generator sequence-to-sequence model, this paper presents a concept pointer network for improving these aspects of abstractive summarization. The network leverages knowledge-based, context-aware conceptualizations to derive an extended set of candidate concepts. The model then points to the most appropriate choice using both the concept set and original source text. This joint approach generates abstractive summaries with higher-level semantic concepts. The training model is also optimized in a way that adapts to different data, which is based on a novel method of distantly-supervised learning guided by reference summaries and testing set. Overall, the proposed approach provides statistically significant improvements over several state-of-the-art models on both the DUC-2004 and Gigaword datasets. A human evaluation of the model's abstractive abilities also supports the quality of the summaries produced within this framework.

Query-based open-domain NLP tasks require information synthesis from long and diverse web results. Current approaches extractively select portions of web text as input to Sequence-to-Sequence models using methods such as TF-IDF ranking. We propose constructing a local graph structured knowledge base for each query, which compresses the web search information and reduces redundancy. We show that by linearizing the graph into a structured input sequence, models can encode the graph representations within a standard Sequence-to-Sequence setting. For two generative tasks with very long text input, long-form question answering and multi-document summarization, feeding graph representations as input can achieve better performance than using retrieved text portions.

One of the basic tasks of computational language documentation (CLD) is to identify word boundaries in an unsegmented phonemic stream. While several unsupervised monolingual word segmentation algorithms exist in the literature, they are challenged in real-world CLD settings by the small amount of available data. A possible remedy is to take advantage of glosses or translation in a foreign, well-resourced, language, which often exist for such data. In this paper, we explore and compare ways to exploit neural machine translation models to perform unsupervised boundary detection with bilingual information, notably introducing a new loss function for jointly learning alignment and segmentation. We experiment with an actual under-resourced language, Mboshi, and show that these techniques can effectively control the output segmentation length.

Deep pre-training and fine-tuning models (such as BERT and OpenAI GPT) have demonstrated excellent results in question answering areas. However, due to the sheer amount of model parameters, the inference speed of these models is very slow. How to apply these complex models to real business scenarios becomes a challenging but practical problem. Previous model compression methods usually suffer from information loss during the model compression procedure, leading to inferior models compared with the original one. To tackle this challenge, we propose a Two-stage Multi-teacher Knowledge Distillation (TMKD for short) method for web Question Answering system. We first develop a general Q\&A distillation task for student model pre-training, and further fine-tune this pre-trained student model with multi-teacher knowledge distillation on downstream tasks (like Web Q\&A task, MNLI, SNLI, RTE tasks from GLUE), which effectively reduces the overfitting bias in individual teacher models, and transfers more general knowledge to the student model. The experiment results show that our method can significantly outperform the baseline methods and even achieve comparable results with the original teacher models, along with substantial speedup of model inference.

Deep Neural Networks (DNN) have been widely employed in industry to address various Natural Language Processing (NLP) tasks. However, many engineers find it a big overhead when they have to choose from multiple frameworks, compare different types of models, and understand various optimization mechanisms. An NLP toolkit for DNN models with both generality and flexibility can greatly improve the productivity of engineers by saving their learning cost and guiding them to find optimal solutions to their tasks. In this paper, we introduce NeuronBlocks\footnote{Code: \url{https://github.com/Microsoft/NeuronBlocks}} \footnote{Demo: \url{https://youtu.be/x6cOpVSZcdo}}, a toolkit encapsulating a suite of neural network modules as building blocks to construct various DNN models with complex architecture. This toolkit empowers engineers to build, train, and test various NLP models through simple configuration of JSON files. The experiments on several NLP datasets such as GLUE, WikiQA and CoNLL-2003 demonstrate the effectiveness of NeuronBlocks.

We show state-of-the-art word representation learning methods maximize an objective function that is a lower bound on the mutual information between different parts of a word sequence (i.e., a sentence). Our formulation provides an alternative perspective that unifies classical word embedding models (e.g., Skip-gram) and modern contextual embeddings (e.g., BERT, XLNet). In addition to enhancing our theoretical understanding of these methods, our derivation leads to a principled framework that can be used to construct new self-supervised tasks. We provide an example by drawing inspirations from related methods based on mutual information maximization that have been successful in computer vision, and introduce a simple self-supervised objective that maximizes the mutual information between a global sentence representation and n-grams in the sentence. Our analysis offers a holistic view of representation learning methods to transfer knowledge and translate progress across multiple domains (e.g., natural language processing, computer vision, audio processing).

Session-based recommender systems have attracted much attention recently. To capture the sequential dependencies, previous sequential recommendation models resort either to data augmentation techniques or a left-to-right style autoregressive training approach. While effective, an obvious drawback is that future user behaviors are always missing during model training. In this paper, we argue that users' future action signals can be exploited to boost the recommendation quality. We present GRec, a simple Gap-filling based encoder-decoder Recommendation framework to generative modelling using both past and future contexts. GfedRec encodes a partially-complete item sequence with blank masks, and autoregressively reconstructs the missing item distributions. In contrast with the typical encoder-decoder paradigm used in the computer vision and NLP domains, GfedRec does not have the data leakage problem when jointly training the encoder and decoder conditioned on the same user action sequence. Experiments on real-word datasets with short-, medium- and long-range user sessions show that GRec largely exceeds the performance of its left-to-right counterparts. Empirical evidence confirms that training sequential recommendation models with future contexts is a promising way to offer better recommendation accuracy.

We present a corpus of sentence-aligned triples of German audio, German text, and English translation, based on German audio books. The corpus consists of over 100 hours of audio material and over 50k parallel sentences. The audio data is read speech and thus low in disfluencies. The quality of audio and sentence alignments has been checked by a manual evaluation, showing that speech alignment quality is in general very high. The sentence alignment quality is comparable to well-used parallel translation data and can be adjusted by cutoffs on the automatic alignment score. To our knowledge, this corpus is to date the largest resource for end-to-end speech translation for German.

Majority of the text modelling techniques yield only point-estimates of document embeddings and lack in capturing the uncertainty of the estimates. These uncertainties give a notion of how well the embeddings represent a document. We present Bayesian subspace multinomial model (Bayesian SMM), a generative log-linear model that learns to represent documents in the form of Gaussian distributions, thereby encoding the uncertainty in its co-variance. Additionally, in the proposed Bayesian SMM, we address a commonly encountered problem of intractability that appears during variational inference in mixed-logit models. We also present a generative Gaussian linear classifier for topic identification that exploits the uncertainty in document embeddings. Our intrinsic evaluation using perplexity measure shows that the proposed Bayesian SMM fits the data better as compared to the state-of-the-art neural variational document model on Fisher speech and 20Newsgroups text corpora. Our topic identification experiments show that the proposed systems are robust to over-fitting on unseen test data. The topic ID results show that the proposed model is outperforms state-of-the-art unsupervised topic models and achieve comparable results to the state-of-the-art fully supervised discriminative models.

Newspapers are a popular form of written discourse, read by many people, thanks to the novelty of the information provided by the news content in it. A headline is the most widely read part of any newspaper due to its appearance in a bigger font and sometimes in colour print. In this paper, we suggest and implement a method for computing inferences from English news headlines, excluding the information from the context in which the headlines appear. This method attempts to generate the possible assumptions a reader formulates in mind upon reading a fresh headline. The generated inferences could be useful for assessing the impact of the news headline on readers including children. The understandability of the current state of social affairs depends greatly on the assimilation of the headlines. As the inferences that are independent of the context depend mainly on the syntax of the headline, dependency trees of headlines are used in this approach, to find the syntactical structure of the headlines and to compute inferences out of them.

For conversational AI and virtual assistants to communicate with humans in a realistic way, they must exhibit human characteristics such as expression of emotion and personality. Current attempts toward constructing human-like dialogue agents have presented significant difficulties. We propose Human Level Attributes (HLAs) based on tropes as the basis of a method for learning dialogue agents that can imitate the personalities of fictional characters. Tropes are characteristics of fictional personalities that are observed recurrently and determined by viewers' impressions. By combining detailed HLA data with dialogue data for specific characters, we present a dataset that models character profiles and gives dialogue agents the ability to learn characters' language styles through their HLAs. We then introduce a three-component system, ALOHA (which stands for Artificial Learning On Human Attributes), that combines character space mapping, character community detection, and language style retrieval to build a character (or personality) specific language model. Our preliminary experiments demonstrate that ALOHA, combined with our proposed dataset, can outperform baseline models at identifying correct dialogue responses of any chosen target character, and is stable regardless of the character's identity, genre of the show, and context of the dialogue.

Context modeling has a pivotal role in open domain conversation. Existing works either use heuristic methods or jointly learn context modeling and response generation with an encoder-decoder framework. This paper proposes an explicit context rewriting method, which rewrites the last utterance by considering context history. We leverage pseudo-parallel data and elaborate a context rewriting network, which is built upon the CopyNet with the reinforcement learning method. The rewritten utterance is beneficial to candidate retrieval, explainable context modeling, as well as enabling to employ a single-turn framework to the multi-turn scenario. The empirical results show that our model outperforms baselines in terms of the rewriting quality, the multi-turn response generation, and the end-to-end retrieval-based chatbots.

Online customer reviews on large-scale e-commerce websites, represent a rich and varied source of opinion data, often providing subjective qualitative assessments of product usage that can help potential customers to discover features that meet their personal needs and preferences. Thus they have the potential to automatically answer specific queries about products, and to address the problems of answer starvation and answer augmentation on associated consumer Q & A forums, by providing good answer alternatives. In this work, we explore several recently successful neural approaches to modeling sentence pairs, that could better learn the relationship between questions and ground truth answers, and thus help infer reviews that can best answer a question or augment a given answer. In particular, we hypothesize that our neural domain adaptation-based approach, due to its ability to additionally learn domain-invariant features from a large number of unlabeled, unpaired question-review samples, would perform better than our proposed baselines, at answering specific, subjective product-related queries using reviews. We validate this hypothesis using a small gold standard dataset of question-review pairs evaluated by human experts, significantly surpassing our chosen baselines. Moreover, our approach, using no labeled question-review sentence pair data for training, gives performance at par with another method utilizing labeled question-review samples for the same task.