With growing amounts of available textual data, development of algorithms capable of automatic analysis, categorization, and summarization of these data has become a necessity. In this research, we present a novel algorithm for keyword identification, that is, an extraction of one or multiword phrases representing key aspects of a given document, called Transformer-Based Neural Tagger for Keyword IDentification (TNT-KID). By adapting the transformer architecture for a specific task at hand and leveraging language model pretraining on a domain-specific corpus, the model is capable of overcoming deficiencies of both supervised and unsupervised state-of-the-art approaches to keyword extraction by offering competitive and robust performance on a variety of different datasets while requiring only a fraction of manually labeled data required by the best-performing systems. This study also offers thorough error analysis with valuable insights into the inner workings of the model and an ablation study measuring the influence of specific components of the keyword identification workflow on the overall performance.

Using natural language processing, it is possible to extract structured information from raw text in the electronic health record (EHR) at reasonably high accuracy. However, the accurate distinction between negated and non-negated mentions of clinical terms remains a challenge. EHR text includes cases where diseases are stated not to be present or only hypothesised, meaning a disease can be mentioned in a report when it is not being reported as present. This makes tasks such as document classification and summarisation more difficult. We have developed the rule-based EdIE-R-Neg, part of an existing text mining pipeline called EdIE-R (Edinburgh Information Extraction for Radiology reports), developed to process brain imaging reports, (https://www.ltg.ed.ac.uk/software/edie-r/) and two machine learning approaches; one using a bidirectional long short-term memory network and another using a feedforward neural network. These were developed on data from the Edinburgh Stroke Study (ESS) and tested on data from routine reports from NHS Tayside (Tayside). Both datasets consist of written reports from medical scans. These models are compared with two existing rule-based models: pyConText (Harkema et al. 2009. Journal of Biomedical Informatics42(5), 839–851), a python implementation of a generalisation of NegEx, and NegBio (Peng et al. 2017. NegBio: A high-performance tool for negation and uncertainty detection in radiology reports. arXiv e-prints, p. arXiv:1712.05898), which identifies negation scopes through patterns applied to a syntactic representation of the sentence. On both the test set of the dataset from which our models were developed, as well as the largely similar Tayside test set, the neural network models and our custom-built rule-based system outperformed the existing methods. EdIE-R-Neg scored highest on F1 score, particularly on the test set of the Tayside dataset, from which no development data were used in these experiments, showing the power of custom-built rule-based systems for negation detection on datasets of this size. The performance gap of the machine learning models to EdIE-R-Neg on the Tayside test set was reduced through adding development Tayside data into the ESS training set, demonstrating the adaptability of the neural network models.

Non-speaking people often rely on AAC (Augmentative and Alternative Communication) devices to assist them to communicate. These AAC devices are slow to operate, however, and as a result conversations can be very difficult and frequently break down. This is especially the case when the conversation partner is unfamiliar with this method of communication, and is a big obstacle to many people when they wish to conduct simple everyday transactions. A way of improving the performance of AAC devices by using scripts is discussed. A prototype system to test this idea was constructed, and a preliminary experiment performed with promising results. A practical AAC device which incorporates scripts was then developed, and is described.

We present a Weighted Finite State Transducer Translation Template Model for statistical machine translation. This is a source-channel model of translation inspired by the Alignment Template translation model. The model attempts to overcome the deficiencies of word-to-word translation models by considering phrases rather than words as units of translation. The approach we describe allows us to implement each constituent distribution of the model as a weighted finite state transducer or acceptor. We show that bitext word alignment and translation under the model can be performed with standard finite state machine operations involving these transducers. One of the benefits of using this framework is that it avoids the need to develop specialized search procedures, even for the generation of lattices or N-Best lists of bitext word alignments and translation hypotheses. We report and analyze bitext word alignment and translation performance on the Hansards French-English task and the FBIS Chinese-English task under the Alignment Error Rate, BLEU, NIST and Word Error-Rate metrics. These experiments identify the contribution of each of the model components to different aspects of alignment and translation performance. We finally discuss translation performance with large bitext training sets on the NIST 2004 Chinese-English and Arabic-English MT tasks.

While human annotation is crucial for many natural language processing tasks, it is often very expensive and time-consuming. Inspired by previous work on crowdsourcing, we investigate the viability of using non-expert labels instead of gold standard annotations from experts for a machine learning approach to automatic readability prediction. In order to do so, we evaluate two different methodologies to assess the readability of a wide variety of text material: A more traditional setup in which expert readers make readability judgments and a crowdsourcing setup for users who are not necessarily experts. To this purpose two assessment tools were implemented: a tool where expert readers can rank a batch of texts based on readability, and a lightweight crowdsourcing tool, which invites users to provide pairwise comparisons. To validate this approach, readability assessments for a corpus of written Dutch generic texts were gathered. By collecting multiple assessments per text, we explicitly wanted to level out readers' background knowledge and attitude. Our findings show that the assessments collected through both methodologies are highly consistent and that crowdsourcing is a viable alternative to expert labeling. This is a good news as crowdsourcing is more lightweight to use and can have access to a much wider audience of potential annotators. By performing a set of basic machine learning experiments using a feature set that mainly encodes basic lexical and morpho-syntactic information, we further illustrate how the collected data can be used to perform text comparisons or to assign an absolute readability score to an individual text. We do not focus on optimising the algorithms to achieve the best possible results for the learning tasks, but carry them out to illustrate the various possibilities of our data sets. The results on different data sets, however, show that our system outperforms the readability formulas and a baseline language modelling approach. We conclude that readability assessment by comparing texts is a polyvalent methodology, which can be adapted to specific domains and target audiences if required.

Question answering systems retrieve information from documents in response to queries. Most of the questions are who- and what-type questions that deal with named entities. A less common and more challenging question to deal with is the why -question. In this paper, we introduce Lemaza (Arabic for why), a system for automatically answering why -questions for Arabic texts. The system is composed of four main components that make use of the Rhetorical Structure Theory. To evaluate Lemaza, we prepared a set of why -question–answer pairs whose answer can be found in a corpus that we compiled out of Open Source Arabic Corpora. Lemaza performed best when the stop-words were not removed. The performance measure was 72.7%, 79.2% and 78.7% for recall, precision and c@1, respectively.

Efficiently exploiting all sources of information such as labeled instances, classes’ representation, and relations of them has a high impact on the performance of Multi-Label Text Classification (MLTC) systems. Most of the current approaches use labeled documents as the primary source of information for MLTC. We investigate the effectiveness of different sources of information— such as the labeled training data, textual labels of classes, and taxonomy relations of classes— for MLTC. More specifically, first, for each document–class pair, different features are extracted using different sources of information. The features reflect the similarity of classes and documents. Then, MLTC is considered to be a ranking problem, and a learning to rank (LTR) approach is used for ranking classes regarding documents and selecting labels of documents. An important characteristic of many MLTC instances is that documents can belong to multiple classes and there are implicit relations between classes. We apply score propagation on top of LTR to incorporate co-occurrence patterns of classes in labeled documents. Our main findings are the following. First, using an LTR approach integrating all features, we observe significantly better performance than previous systems for MLTC. Specifically, we show that simple classification approaches fail when there is a high number of classes. Second, the analysis of feature weights reveals the relative importance of various sources of evidence, also giving insight into the underlying classification problem. Interestingly, the results indicate that the titles of documents are more informative than all other sources of information. Third, a lean-and-mean system using only four features is able to perform at 96% of the large LTR model that we propose in this paper. Fourth, using the co-occurrence information of classes helps in classifying documents more accurately. Our results show that the co-occurrence information is more helpful when the underlying classifier has a poor performance.

It took a while, but natural language generation is now an established commercial software category. It’s commented upon frequently in both industry media and the mainstream press, and businesses are willing to pay hard cash to take advantage of the technology. We look at who’s active in the space, the nature of the technology that’s available today and where things might go in the future.

The distributional similarity methods have proven to be a valuable tool for the induction of semantic similarity. Until now, most algorithms use two-way co-occurrence data to compute the meaning of words. Co-occurrence frequencies, however, need not be pairwise. One can easily imagine situations where it is desirable to investigate co-occurrence frequencies of three modes and beyond. This paper will investigate tensor factorization methods to build a model of three-way co-occurrences. The approach is applied to the problem of selectional preference induction, and automatically evaluated in a pseudo-disambiguation task. The results show that tensor factorization, and non-negative tensor factorization in particular, is a promising tool for Natural Language Processing (nlp).

Unsupervised Authorship Analysis (UAA) aims to cluster documents by authorship without knowing the authorship of any documents. An important factor in UAA is the method for calculating the distance between documents. This choice of the authorship distance method is considered more critical to the end result than the choice of cluster analysis algorithm. One method for measuring the correlation between a distance metric and a labelling (such as class values or clusters) is the Silhouette Coefficient (SC). The SC can be leveraged by measuring the correlation between the authorship distance method and the true authorship, evaluating the quality of the distance method. However, we show that the SC can be severely affected by outliers. To address this issue, we introduce the Positive Silhouette Coefficient, given as the proportion of instances with a positive SC value. This metric is not easily altered by outliers and produces a more robust metric. A large number of authorship distance methods are then compared using the PSC, and the findings are presented. This research provides an insight into the efficacy of methods for UAA and presents a framework for testing authorship distance methods.

Sentiment analysis has attracted a lot of research interest in recent years, especially in the context of social media. While most of this research has focused on English, there is ample data and interest in the topic for many other languages, as well. In this article, we propose a comprehensive sentiment analysis system for Turkish. We cover different levels of sentiment analysis such as aspect, sentence, and document levels as well as some linguistic issues such as conjunction and intensification in Turkish sentiment analysis. Our system is evaluated on Turkish movie reviews and the obtained accuracies range from sixty per cent to seventy-nine per cent in ternary and binary classification tasks at different levels of analysis.

The Empirical Methods in Natural Language Processing (EMNLP) 2018 workshop BlackboxNLP was dedicated to resources and techniques specifically developed for analyzing and understanding the inner-workings and representations acquired by neural models of language. Approaches included: systematic manipulation of input to neural networks and investigating the impact on their performance, testing whether interpretable knowledge can be decoded from intermediate representations acquired by neural networks, proposing modifications to neural network architectures to make their knowledge state or generated output more explainable, and examining the performance of networks on simplified or formal languages. Here we review a number of representative studies in each category.

Sentence-level representations are necessary for various natural language processing tasks. Recurrent neural networks have proven to be very effective in learning distributed representations and can be trained efficiently on natural language inference tasks. We build on top of one such model and propose a hierarchy of bidirectional LSTM and max pooling layers that implements an iterative refinement strategy and yields state of the art results on the SciTail dataset as well as strong results for Stanford Natural Language Inference and Multi-Genre Natural Language Inference. We can show that the sentence embeddings learned in this way can be utilized in a wide variety of transfer learning tasks, outperforming InferSent on 7 out of 10 and SkipThought on 8 out of 9 SentEval sentence embedding evaluation tasks. Furthermore, our model beats the InferSent model in 8 out of 10 recently published SentEval probing tasks designed to evaluate sentence embeddings’ ability to capture some of the important linguistic properties of sentences.

An automated tool to assist in the understanding of legacy code components can be useful both in the areas of software reuse and software maintenance. Most previous work in this area has concentrated on functionally-oriented code. Whereas object-oriented code has been shown to be inherently more reusable than functionally-oriented code, in many cases the eventual reuse of the object-oriented code was not considered during development. A knowledge-based, natural language processing approach to the automated understanding of object-oriented code as an aid to the reuse of object-oriented code is described. A system, called the PATRicia system (Program Analysis Tool for Reuse) that implements the approach is examined. The natural language processing/information extraction system that comprises a large part of the PATRicia system is discussed and the knowledge-base of the PATRicia system, in the form of conceptual graphs, is described. Reports provided by natural language-generation in the PATRicia system are described.

Whilst Spoken Language Dialogue Systems (SLDSs) technology has made good progress in recent years, the issue of SLDS usability is still lagging behind both theoretically and in actual SLDS development and evaluation. However, as more products reach the market and competition intensifies, there is growing recognition of the importance of systematically understanding the factors which must be taken into account in order to optimise SLDS usability. Ideally, this understanding should be comprehensive (i.e. include all major human factors perspectives on SLDSs), and exhaustive (i.e. describe each perspective as it pertains to the detailed development and evaluation of any possible SLDS). This paper addresses the requirement of comprehensiveness by decomposing the complex space of SLDS usability best practice into eleven issues which should be considered by developers during specification, design, development and evaluation. The discussion of each issue is aimed to support the developer in building SLDSs which are likely to generate user satisfaction, which are perceived to be easy to understand and control, and which enable smooth user- system interaction. Based on the best practice issues discussed, criteria for evaluating SLDS usability are proposed. Several limits to our current understanding of SLDS usability are highlighted.

We present an Augmented Template-Based approach to text realization that addresses the requirements of real-time, interactive systems such as a dialog system or an intelligent tutoring system. Template-based approaches are easier to implement and use than traditional approaches to text realization. They can also generate texts more quickly. However traditional template-based approaches with rigid templates are inflexible and difficult to reuse. Our approach augments traditional template-based approaches by adding several types of declarative control expressions and an attribute grammar-based mechanism for processing missing or inconsistent slot fillers. Therefore, augmented templates can be made more general than traditional ones, yielding templates that are more flexible and reusable across applications.

This paper presents a novel approach for word sense disambiguation. The underlying algorithm has two main components: (1) pattern learning from available sense-tagged corpora (SemCor), from dictionary definitions (WordNet) and from a generated corpus (GenCor); and (2) instance based learning with automatic feature selection, when training data is available for a particular word. The ideas described in this paper were implemented in a system that achieves excellent performance on the data provided during the SENSEVAL-2 evaluation exercise, for both English all words and English lexical sample tasks.

We address the problem of clustering words (or constructing a thesaurus) based on co-occurrence data, and conducting syntactic disambiguation by using the acquired word classes. We view the clustering problem as that of estimating a class-based probability distribution specifying the joint probabilities of word pairs. We propose an efficient algorithm based on the Minimum Description Length (MDL) principle for estimating such a probability model. Our clustering method is a natural extension of that proposed in Brown, Della Pietra, deSouza, Lai and Mercer (1992). We next propose a syntactic disambiguation method which combines the use of automatically constructed word classes and that of a hand-made thesaurus. The overall disambiguation accuracy achieved by our method is 88.2%, which compares favorably against the accuracies obtained by the state-of-the-art disambiguation methods.

GPT-3 is a large-scale natural language model developed by OpenAI that can perform many different tasks, including topic classification. Although researchers claim that it requires only a small number of in-context examples to learn a task, in practice GPT-3 requires these training examples to be either of exceptional quality or a higher quantity than easily created by hand. To address this issue, this study teaches GPT-3 to classify whether a question is related to data science by augmenting a small training set with additional examples generated by GPT-3 itself. This study compares two augmented classifiers: the Classification Endpoint with an increased training set size and the Completion Endpoint with an augmented prompt optimized using a genetic algorithm. We find that data augmentation significantly increases the accuracy of both classifiers, and that the embedding-based Classification Endpoint achieves the best accuracy of about 76%, compared to human accuracy of 85%. In this way, giving large language models like GPT-3 the ability to propose their own training examples can improve short text classification performance.

The recent breakthroughs in deep neural architectures across multiple machine learning fields have led to the widespread use of deep neural models. These learners are often applied as black-box models that ignore or insufficiently utilize a wealth of preexisting semantic information. In this study, we focus on the text classification task, investigating methods for augmenting the input to deep neural networks (DNNs) with semantic information. We extract semantics for the words in the preprocessed text from the WordNet semantic graph, in the form of weighted concept terms that form a semantic frequency vector. Concepts are selected via a variety of semantic disambiguation techniques, including a basic, a part-of-speech-based, and a semantic embedding projection method. Additionally, we consider a weight propagation mechanism that exploits semantic relationships in the concept graph and conveys a spreading activation component. We enrich word2vec embeddings with the resulting semantic vector through concatenation or replacement and apply the semantically augmented word embeddings on the classification task via a DNN. Experimental results over established datasets demonstrate that our approach of semantic augmentation in the input space boosts classification performance significantly, with concatenation offering the best performance. We also note additional interesting findings produced by our approach regarding the behavior of term frequency - inverse document frequency normalization on semantic vectors, along with the radical dimensionality reduction potential with negligible performance loss.