Mikolov et al. (2013a) observed that continuous bag-of-words (CBOW) word embeddings tend to underperform Skip-gram (SG) embeddings, and this finding has been reported in subsequent works. We find that these observations are driven not by fundamental
differences in their training objectives, but more likely on faulty negative sampling CBOW implementations in popular libraries such as the official implementation, word2vec.c, and Gensim. We show that after correcting a bug in the CBOW gradient update, one can learn CBOW word embeddings that are fully competitive with SG on various intrinsic and extrinsic tasks, while being many times faster to train.
Language representations are known to carry stereotypical biases and, as a result, lead to biased predictions in downstream tasks. While existing methods are effective at mitigating biases by linear projection, such methods are too aggressive: they n
ot only remove bias, but also erase valuable information from word embeddings. We develop new measures for evaluating specific information retention that demonstrate the tradeoff between bias removal and information retention. To address this challenge, we propose OSCaR (Orthogonal Subspace Correction and Rectification), a bias-mitigating method that focuses on disentangling biased associations between concepts instead of removing concepts wholesale. Our experiments on gender biases show that OSCaR is a well-balanced approach that ensures that semantic information is retained in the embeddings and bias is also effectively mitigated.
This paper investigates how to correct Chinese text errors with types of mistaken, missing and redundant characters, which are common for Chinese native speakers. Most existing models based on detect-correct framework can correct mistaken characters,
but cannot handle missing or redundant characters due to inconsistency between model inputs and outputs. Although Seq2Seq-based or sequence tagging methods provide solutions to the three error types and achieved relatively good results in English context, they do not perform well in Chinese context according to our experiments. In our work, we propose a novel alignment-agnostic detect-correct framework that can handle both text aligned and non-aligned situations and can serve as a cold start model when no annotation data are provided. Experimental results on three datasets demonstrate that our method is effective and achieves a better performance than most recent published models.
This paper presents Self-correcting Encoding (Secoco), a framework that effectively deals with noisy input for robust neural machine translation by introducing self-correcting predictors. Different from previous robust approaches, Secoco enables NMT
to explicitly correct noisy inputs and delete specific errors simultaneously with the translation decoding process. Secoco is able to achieve significant improvements over strong baselines on two real-world test sets and a benchmark WMT dataset with good interpretability. We will make our code and dataset publicly available soon.
Generating paragraphs of diverse contents is important in many applications. Existing generation models produce similar contents from homogenized contexts due to the fixed left-to-right sentence order. Our idea is permuting the sentence orders to imp
rove the content diversity of multi-sentence paragraph. We propose a novel framework PermGen whose objective is to maximize the expected log-likelihood of output paragraph distributions with respect to all possible sentence orders. PermGen uses hierarchical positional embedding and designs new procedures for training, and decoding in the sentence-permuted generation. Experiments on three paragraph generation benchmarks demonstrate PermGen generates more diverse outputs with a higher quality than existing models.
Recently, it has been argued that encoder-decoder models can be made more interpretable by replacing the softmax function in the attention with its sparse variants. In this work, we introduce a novel, simple method for achieving sparsity in attention
: we replace the softmax activation with a ReLU, and show that sparsity naturally emerges from such a formulation. Training stability is achieved with layer normalization with either a specialized initialization or an additional gating function. Our model, which we call Rectified Linear Attention (ReLA), is easy to implement and more efficient than previously proposed sparse attention mechanisms. We apply ReLA to the Transformer and conduct experiments on five machine translation tasks. ReLA achieves translation performance comparable to several strong baselines, with training and decoding speed similar to that of the vanilla attention. Our analysis shows that ReLA delivers high sparsity rate and head diversity, and the induced cross attention achieves better accuracy with respect to source-target word alignment than recent sparsified softmax-based models. Intriguingly, ReLA heads also learn to attend to nothing (i.e. switch off') for some queries, which is not possible with sparsified softmax alternatives.
Sarcasm detection is important for several NLP tasks such as sentiment identification in product reviews, user feedback, and online forums. It is a challenging task requiring a deep understanding of language, context, and world knowledge. In this pap
er, we investigate whether incorporating commonsense knowledge helps in sarcasm detection. For this, we incorporate commonsense knowledge into the prediction process using a graph convolution network with pre-trained language model embeddings as input. Our experiments with three sarcasm detection datasets indicate that the approach does not outperform the baseline model. We perform an exhaustive set of experiments to analyze where commonsense support adds value and where it hurts classification. Our implementation is publicly available at: https://github.com/brcsomnath/commonsense-sarcasm.
State-of-the-art approaches to spelling error correction problem include Transformer-based Seq2Seq models, which require large training sets and suffer from slow inference time; and sequence labeling models based on Transformer encoders like BERT, wh
ich involve token-level label space and therefore a large pre-defined vocabulary dictionary. In this paper we present a Hierarchical Character Tagger model, or HCTagger, for short text spelling error correction. We use a pre-trained language model at the character level as a text encoder, and then predict character-level edits to transform the original text into its error-free form with a much smaller label space. For decoding, we propose a hierarchical multi-task approach to alleviate the issue of long-tail label distribution without introducing extra model parameters. Experiments on two public misspelling correction datasets demonstrate that HCTagger is an accurate and much faster approach than many existing models.
Recent transformer-based approaches to NLG like GPT-2 can generate syntactically coherent original texts. However, these generated texts have serious flaws: global discourse incoherence and meaninglessness of sentences in terms of entity values. We a
ddress both of these flaws: they are independent but can be combined to generate original texts that will be both consistent and truthful. This paper presents an approach to estimate the quality of discourse structure. Empirical results confirm that the discourse structure of currently generated texts is inaccurate. We propose the research directions to correct it using discourse features during the fine-tuning procedure. The suggested approach is universal and can be applied to different languages. Apart from that, we suggest a method to correct wrong entity values based on Web Mining and text alignment.
We develop a minimally-supervised model for spelling correction and evaluate its performance on three datasets annotated for spelling errors in Russian. The first corpus is a dataset of Russian social media data that was recently used in a shared tas
k on Russian spelling correction. The other two corpora contain texts produced by learners of Russian as a foreign language. Evaluating on three diverse datasets allows for a cross-corpus comparison. We compare the performance of the minimally-supervised model to two baseline models that do not use context for candidate re-ranking, as well as to a character-level statistical machine translation system with context-based re-ranking. We show that the minimally-supervised model outperforms all of the other models. We also present an analysis of the spelling errors and discuss the difficulty of the task compared to the spelling correction problem in English.
