We present a novel way of injecting factual knowledge about entities into the pretrained BERT model (Devlin et al., 2019): We align Wikipedia2Vec entity vectors (Yamada et al., 2016) with BERT's native wordpiece vector space and use the aligned entity vectors as if they were wordpiece vectors. The resulting entity-enhanced version of BERT (called E-BERT) is similar in spirit to ERNIE (Zhang et al., 2019) and KnowBert (Peters et al., 2019), but it requires no expensive further pretraining of the BERT encoder. We evaluate E-BERT on unsupervised question answering (QA), supervised relation classification (RC) and entity linking (EL). On all three tasks, E-BERT outperforms BERT and other baselines. We also show quantitatively that the original BERT model is overly reliant on the surface form of entity names (e.g., guessing that someone with an Italian-sounding name speaks Italian), and that E-BERT mitigates this problem.

The behavior of deep neural networks (DNNs) is hard to understand. This makes it necessary to explore post hoc explanation methods. We conduct the first comprehensive evaluation of explanation methods for NLP. To this end, we design two novel evaluation paradigms that cover two important classes of NLP problems: small context and large context problems. Both paradigms require no manual annotation and are therefore broadly applicable. We also introduce LIMSSE, an explanation method inspired by LIME that is designed for NLP. We show empirically that LIMSSE, LRP and DeepLIFT are the most effective explanation methods and recommend them for explaining DNNs in NLP.

Domain adaptation of Pretrained LanguageModels (PTLMs) is typically achieved by unsupervised pretraining on target-domain text. While successful, this approach is expensive in terms of hardware, runtime and CO 2 emissions. Here, we propose a cheaper alternative: We train Word2Vec on target-domain text and align the resulting word vectors with the wordpiece vectors of a general-domain PTLM. We evaluate on eight English biomedical Named Entity Recognition (NER) tasks and compare against the recently proposed BioBERT model. We cover over 60% of the BioBERT -BERT F1 delta, at 5% of BioBERT's CO 2 footprint and 2% of its cloud compute cost. We also show how to quickly adapt an existing generaldomain Question Answering (QA) model to an emerging domain: the Covid-19 pandemic. 1

We address the task of unsupervised Semantic Textual Similarity (STS) by ensembling diverse pre-trained sentence encoders into sentence meta-embeddings. We apply, extend and evaluate different meta-embedding methods from the word embedding literature at the sentence level, including dimensionality reduction (Yin and Schütze, 2016), generalized Canonical Correlation Analysis (Rastogi et al., 2015) and cross-view auto-encoders (Bollegala and Bao, 2018). Our sentence metaembeddings set a new unsupervised State of The Art (SoTA) on the STS Benchmark and on the STS12-STS16 datasets, with gains of between 3.7% and 6.4% Pearson's r over singlesource systems.

Interpretability of machine learning (ML) models becomes more relevant with their increasing adoption. In this work, we address the interpretability of ML based question answering (QA) models on a combination of knowledge bases (KB) and text documents. We adapt post hoc explanation methods such as LIME and input perturbation (IP) and compare them with the self-explanatory attention mechanism of the model. For this purpose, we propose an automatic evaluation paradigm for explanation methods in the context of QA. We also conduct a study with human annotators to evaluate whether explanations help them identify better QA models. Our results suggest that IP provides better explanations than LIME or attention, according to both automatic and human evaluation. We obtain the same ranking of methods in both experiments, which supports the validity of our automatic evaluation paradigm.

Motivation: Here we investigate deep learning-based prediction of protein secondary structure from the protein primary sequence. We study the function of different features in this task, including one-hot vectors, biophysical features, protein sequence embedding (ProtVec), deep contextualized embedding (known as ELMo), and the Position Specific Scoring Matrix (PSSM). In addition to the role of features, we evaluate various deep learning architectures including the following models/mechanisms and certain combinations: Bidirectional Long Short-Term Memory (BiLSTM), convolutional neural network (CNN), highway connections, attention mechanism, recurrent neural random fields, and gated multi-scale CNN. Our results suggest that PSSM concatenated to one-hot vectors are the most important features for the task of secondary structure prediction. Results: Utilizing the CNN-BiLSTM network, we achieved an accuracy of 69.9% and 70.4% using ensemble top-k models, for 8-class of protein secondary structure on the CB513 dataset, the most challenging dataset for protein secondary structure prediction. Through error analysis on the best performing model, we showed that the misclassification is significantly more common at positions that undergo secondary structure transitions, which is most likely due to the inaccurate assignments of the secondary structure at the boundary regions. Notably, when ignoring amino acids at secondary structure transitions in the evaluation, the accuracy increases to 90.3%. Furthermore, the best performing model mostly mistook similar structures for one another, indicating that the deep learning model inferred high-level information on the secondary structure. Availability: The developed software called DeepPrime2Sec and the used datasets are available at

Input optimization methods, such as Google Deep Dream, create interpretable representations of neurons for computer vision DNNs. We propose and evaluate ways of transferring this technology to NLP. Our results suggest that gradient ascent with a gumbel softmax layer produces n-gram representations that outperform naive corpus search in terms of target neuron activation. The representations highlight differences in syntax awareness between the language and visual models of the Imaginet architecture.

We address the problem of Duplicate Question Detection (DQD) in low-resource domainspecific Community Question Answering forums. Our multi-view framework MV-DASE combines an ensemble of sentence encoders via Generalized Canonical Correlation Analysis, using unlabeled data only. In our experiments, the ensemble includes generic and domain-specific averaged word embeddings, domain-finetuned BERT and the Universal Sentence Encoder. We evaluate MV-DASE on the CQADupStack corpus and on additional low-resource Stack Exchange forums. Combining the strengths of different encoders, we significantly outperform BM25, all singleview systems as well as a recent supervised domain-adversarial DQD method.