English Adjectives and Estonian Nouns: Looking for Agreement?

This study investigated how speakers of Estonian as L1 with varying degree of proficiency in English judge grammaticality of bilingual constructions English adjective + Estonian noun from the point of view of adjective agreement. Estonian is rich in inflectional morphology, and adjectives agree with nouns in case and number. The empirical evidence from English-Estonian bilingual speech shows that agreement is not always the case even when an English adjective fits into Estonian declension system. It is hypothesized that the higher proficiency in/exposure to English is, the higher is the acceptability of bilingual adjective phrases, and (non-)agreement does not play a role. To test this, an experiment was designed where the test corpus of 108 sentences consisted of real and constructed examples, both in agreement and non-agreement condition. Real sentences came from fashion and beauty blogs and vlogs. The test was administered online and the participants were asked to rate adjective acceptability. The hypothesis was confirmed: increased proficiency in English, together with younger age, had a positive correlation with acceptability of all adjective types, independent of adjective (non-)agreement. Residence and birthplace had a small effect on acceptability of some adjective types. Whether sentences were real or constructed, had only a minor effect. Male participants tended to assess real sentences lower, probably because of the topics typical for female blogs. Monosyllabic consonant-ending adjectives were exceptional, as their assessment did not depend on any factor. All in all, the study demonstrated that grammaticality judgment among the native speakers of the same L1 differs because of different degrees of bilingualism, and structural factors, such as compatibility with Estonian declension system, are not decisive. Thus, it is not clear what an ideal native speaker is.

Graph Convolutional Network for Word Sense Disambiguation

Word sense disambiguation (WSD) is an important research topic in natural language processing, which is widely applied to text classification, machine translation, and information retrieval. In order to improve disambiguation accuracy, this paper proposes a WSD method based on the graph convolutional network (GCN). Word, part of speech, and semantic category are extracted from contexts of the ambiguous word as discriminative features. Discriminative features and sentence containing the ambiguous word are used as nodes to construct the WSD graph. Word2Vec tool, Doc2Vec tool, pointwise mutual information (PMI), and TF-IDF are applied to compute embeddings of nodes and edge weights. GCN is used to fuse features of a node and its neighbors, and the softmax function is applied to determine the semantic category of the ambiguous word. Training corpus of SemEval-2007: Task #5 is adopted to optimize the proposed WSD classifier. Test corpus of SemEval-2007: Task #5 is used to test the performance of WSD classifier. Experimental results show that average accuracy of the proposed method is improved.

Big Data, Big Noise: The Challenge of Finding Issue Networks on the Web

In this paper, we focus on noise in the sense of irrelevant information in a data set as a specific methodological challenge of web research in the era of big data. We empirically evaluate several methods for filtering hyperlink networks in order to reconstruct networks that contain only web pages that deal with a particular issue. The test corpus of web pages was collected from hyperlink networks on the issue of food safety in the United States and Germany. We applied three filtering strategies and evaluated their performance to exclude irrelevant content from the networks: keyword filtering, automated document classification with a machine-learning algorithm, and extraction of core networks with network-analytical measures. Keyword filtering and automated classification of web pages were the most effective methods for reducing noise whereas extracting a core network did not yield satisfying results for this case.

Beyond Differences: Assessing Effects of Shared Linguistic Features on L2 Writing Quality of Two Genres

This study explored the relationship between linguistic features and the rated quality of letters of application (LAs) and argumentative essays (AEs) composed in English by Chinese college-level English as a foreign language (EFL) learners. A corpus of 260 LAs and 260 AEs were analyzed via a confirmatory factor analysis. Latent variables were EFL writing quality, captured by writing scores, and lexical sophistication, syntactic complexity, and cohesion, each captured by different linguistic features in the two genres of writing. Results indicated that lexical decision times, moving average type-token ratio with a 50-word window, and complex nominals per clause explained 55.5 per cent of the variance in the holistic scores of both genres of writing. This pattern of predictivity was further validated with a test corpus of 110 LAs and 110 AEs, revealing that, albeit differing in genre, higher-rated LAs and AEs were likely to contain more sophisticated words and complex nominals and exhibit a higher type-token ratio with a 50-word window. These findings help enrich our understanding of the shared features of different genres of EFL writing and have potentially useful implications for EFL writing pedagogy and assessment.

The traceback method in child language acquisition research: identifying patterns in early speech

abstract This paper discusses the traceback method, which has been the basis of some influential papers on first language acquisition. The method sets out to demonstrate that many or even all utterances in a test corpus (usually the last two sessions of recording) can be accounted for with the help of recurrent fixed strings (like What’s that?) or frame-and-slot patterns (like [What’s X?]) that can also be identified in the remaining dataset (i.e., the previous sessions of recording). This is taken as evidence that language learning is much more item-based than previously assumed. In the present paper we sketch the development of the method over the last two decades, and discuss its relation to usage-based theory, as well as the cognitive plausibility of its components, and we highlight both its potential and its limitations.

Numerical Simulation of Ambiguity Resolution in Multiple Information Streams Based on Network Machine Translation

In natural language, the phenomenon of polysemy is widespread, which makes it very difficult for machines to process natural language. Word sense disambiguation is a key issue in the field of natural language processing. This paper introduces the more common statistical learning methods used in the field of word sense disambiguation. Using the naive Bayesian machine learning method and the feature vector set extracted and constructed by the Dice coefficient method, a semantic word disambiguation model based on semantics is realized. The results of comparative experiments show that the proposed method is better compared with known systems. This paper proposes a method for disambiguation of word segmentation in professional fields based on unsupervised learning. This method does not rely on professional domain knowledge and training corpus and only uses the frequency, mutual information, and boundary entropy information of the string in the test corpus to solve the problem of word segmentation ambiguity. The experimental results show that these three evaluation standards can solve the problem of word segmentation ambiguity in professional fields and improve the effect of word segmentation. Among them, the segmentation result using mutual information is the best, and the performance is stable.

Sentiment Mining Model for Opinionated Afaan Oromo Texts

Opinions are personal judgment on entity. This is not only true for individuals but also true for organizations. Opinion mining is a type of natural language processing for tracking the mood of the public about a particular product. The process of sentiment mining involves categorizing an opinionated document into predefined categories such as positive, negative or neutral based on the sentiment terms that appear within the opinionated document. For this study text document corpus is prepared by the researcher encompassing different movies ‘reviews and Various techniques of text pre-processing including tokenization, normalization, stop word removal and stemming are used for this system(sentiment mining model for opinionated afaan Oromo texts). The experiment shows that the performance is on the average 0.849(84.9%) precision and 0.887(88.7%) recall. The challenging tasks in the study are handling synonymy and inability of the stemmer algorithm to all word variants, and ambiguity of words in the language. The performance the system can be increased if stemming algorithm is improved, standard test corpus is used, and thesaurus is used to handle polysemy and synonymy words in the language.

Automatically detecting open academic review praise and criticism

PurposePeer reviewer evaluations of academic papers are known to be variable in content and overall judgements but are important academic publishing safeguards. This article introduces a sentiment analysis program, PeerJudge, to detect praise and criticism in peer evaluations. It is designed to support editorial management decisions and reviewers in the scholarly publishing process and for grant funding decision workflows. The initial version of PeerJudge is tailored for reviews from F1000Research's open peer review publishing platform.Design/methodology/approachPeerJudge uses a lexical sentiment analysis approach with a human-coded initial sentiment lexicon and machine learning adjustments and additions. It was built with an F1000Research development corpus and evaluated on a different F1000Research test corpus using reviewer ratings.FindingsPeerJudge can predict F1000Research judgements from negative evaluations in reviewers' comments more accurately than baseline approaches, although not from positive reviewer comments, which seem to be largely unrelated to reviewer decisions. Within the F1000Research mode of post-publication peer review, the absence of any detected negative comments is a reliable indicator that an article will be ‘approved’, but the presence of moderately negative comments could lead to either an approved or approved with reservations decision.Originality/valuePeerJudge is the first transparent AI approach to peer review sentiment detection. It may be used to identify anomalous reviews with text potentially not matching judgements for individual checks or systematic bias assessments.

Enabling Robust Grammatical Error Correction in New Domains: Data Sets, Metrics, and Analyses

Until now, grammatical error correction (GEC) has been primarily evaluated on text written by non-native English speakers, with a focus on student essays. This paper enables GEC development on text written by native speakers by providing a new data set and metric. We present a multiple-reference test corpus for GEC that includes 4,000 sentences in two new domains ( formal and informal writing by native English speakers) and 2,000 sentences from a diverse set of non-native student writing. We also collect human judgments of several GEC systems on this new test set and perform a meta-evaluation, assessing how reliable automatic metrics are across these domains. We find that commonly used GEC metrics have inconsistent performance across domains, and therefore we propose a new ensemble metric that is robust on all three domains of text.