Zastosowanie sztucznych sieci neuronowych w językoznawstwie

Author(s): Anna Obrębska / Language(s): Polish Issue: 44/2007

Artificial Neural Networks (ANN) are software systems inspired by biological models of the brain. They are capable of learning patterns and relations from input data. The ability to generalise knowledge enabling them to deal with previously unseen input data, so they can solve problems, that other software can’t. This paper presents a simple artificial neural network which classifies sentences into three classes: akceptable sentences, unakceptable sentences and trivial sentences. The main goal of this work is to show, how useful in linguistic Artificial Neural Networks could be.

Eestikeelsete veebitekstide automaatne liigitamine

Author(s): Kristiina Vaik,Kadri Muischnek / Language(s): Estonian Issue: 14/2018

Due to the size of the Internet and the multitude of traditional and new genres there has been an increasing interest in automatic genre classification. Labelling texts in natural language processing is essential because this allows us to select more appropriate language models for the analysis. The aim of the article is to describe and present the results of automatically classifying Estonian Web 2013 texts. We evalued the quality of different classification models on our training and manually labelled test set. Most of the research on automatic classification has focused on classifying multiple genres, while our objective was to do a binary classification. We set out to classify Estonian Web 2013 texts based on whether they are canonical or not. For training we used the Balanced Corpus to represent canonical language and the New Media Corpus to represent non-canonical language. Due to the non-availability of a binary labelled subcorpus of Estonian Web 2013 texts, we compiled it ourselves by manually labelling it. For classification we used different supervised machine learning algorithms and for features a simple Bag of Words method. The results obtained from the preliminary experiments show that neural networks outperformed other machine learning algorithms achieving over 0.7 on accuracy. The overall results of this study indicate that in order to increase the accuracy of the classifiers, new features should be added (e.g POS count, sentences per paragraph, words per sentence, uppercase and lowercase letters per sentence etc.). Our best model, the neural network classifier, achieved an accuracy of 0.99 on a training set but only a little over 0.74 on the test set. This suggests that future work requiers a bigger and more appropriate training set. The manually labelling task showed us that the transition from canonical to non-canonical is very smooth. Current models produce a score between 0 and 1, defining if the item belongs to a class or not. Therefore, the classification models must be programmed to be more predictive so that the predictions can be tuned by selecting a threshold.

The role of word frequencies in detecting unfamiliar terms and their effect on response quality

Author(s): Ana Slavec,Vasja Vehovar / Language(s): English Issue: 4/2015

Research into cognitive aspects of survey response has indicated unfamiliar terms as one of the psycholinguistic determinants of question comprehensibility problems. In this paper the estimates of wording familiarity based on text corpora for the English and Slovenian languages were used to detect potentially incomprehensible wordings in two web survey questionnaires for international exchange students at the University of Ljubljana, one for incoming (English) and the other for outgoing students (Slovenian). Two versions of the questionnaire were developed for each language, one with low-frequency (complex) and the other with high-frequency (improved) wordings, and compared in a split-ballot experiment. The results show a lower drop-out rate and a decreased subjective perception of difficulty for the improved language versions.

Using Amazon Mechanical Turk for linguistic research

Author(s): Tyler Schnoebelen,Victor Kuperman / Language(s): English Issue: 4/2010

Amazon’s Mechanical Turk service makes linguistic experimentation quick, easy, and inexpensive. However, researchers have not been certain about its reliability. In a series of experiments, this paper compares data collected via Mechanical Turk to those obtained using more traditional methods One set of experiments measured the predictability of words in sentences using the Cloze sentence completion task (Taylor, 1953). The correlation between traditional and Turk Cloze scores is high (rho=0.823) and both data sets perform similarly against alternative measures of contextual predictability. Five other experiments on the semantic relatedness of verbs and phrasal verbs (how much is “lift” part of “lift up”) manipulate the presence of the sentence context and the composition of the experimental list. The results indicate that Turk data correlate well between experiments and with data from traditional methods (rho up to 0.9), and they show high inter-rater consistency and agreement. We conclude that Mechanical Turk is a reliable source of data for complex linguistic tasks in heavy use by psycholinguists. The paper provides suggestions for best practices in data collection and scrubbing.

Vector based semantic analysis reveals absence of competition among related senses

Author(s): Dušica Filipović-Đurđević,Đorđe Đurđević,Aleksandar Kostić / Language(s): English Issue: 1/2009

Previous research demonstrated that processing time was facilitated by number of related word senses (polysemy) and inhibited by number of unrelated word meanings (homonymy). The starting point of this research were the findings described by Moscoso del Prado Martín and colleagues, who offered a unique account of processing of two forms of lexical ambiguity. By applying the techniques they proposed, for the set of strictly polysemous Serbian nouns we calculated ambiguity measures they introduced. Based on the covariance matrix of the context vectors, we derived entropy of equivalent Gaussian distribution, and based on the context vectors probability density function, we derived differential entropy. Negentropy was calculated as the difference between the two. Based on interpretation that entropy of equivalent Gaussian mirrors sense cooperation, or polysemy, while negentropy mirrors meaning competition, or homonymy, we predicted that in the set of strictly polysemous nouns, negentropy effect would disappear. In accordance with our predictions, entropy of equivalent Gaussian distribution accounted for significant proportion of processing latencies variance. Negentropy did not affect reaction time. This finding is in accordance with the hypothesis that entropy of equivalent Gaussian distribution, as a measure of general width of activation in semantic space, reflects polysemy, that is, the existence of related senses. Therefore, polysemy advantage could be the result of the wide-spread activation in semantic space and reduced competition among overlapping Gaussians.

Stability of the syntagmatic probability distributions

Author(s): Strahinja Dimitrijević,Aleksandar Kostić,Petar Milin / Language(s): English Issue: 1/2009

The aim of the present study is to establish criteria for the optimal size of a corpus that can provide stable conditional probabilities of morphological and/or syntagmatic types. The optimality of corpus size is defined in terms of the smallest sample that generates probability distribution equal to distribution derived from the large sample that generates stable probabilities. The latter distribution we refer to as “target distribution”. In order to establish the above criteria we varied the sample size, the word sequence size (bigrams and trigrams), sampling procedure (randomly chosen words and continuous text) and position of the target word in a sequence. The obtained distributions of conditional probabilities derived from smaller samples have been correlated with target distributions. Sample size at which probability distribution reaches maximal correlation (r=1) with the target distribution was taken as being optimal. The research was done on Corpus of Serbian language. In case of bigrams the optimal sample size for random word selection is 65.000 words, and 281.000 words for trigrams. In contrast, continuous text sampling requires much larger samples to reach stability: 810.000 words for bigrams and 868.000 words for trigrams. The factors that caused these differences remain unclear and need additional empirical investigation.

Aproksimacija verovatnoća i optimalna veličina jezičkog uzorka

Author(s): Aleksandar Kostić,Svetlana Ilić,Petar Milin / Language(s): Serbian Issue: 1/2008

Reliable language corpus implies a text sample of size n that provides stable probability distributions of linguistic phenomena. The question is what is the minimal (i.e. the optimal) text size at which probabilities of linguistic phenomena become stable. Specifically, we were interested in probabilities of grammatical forms. We started with an a priori assumption that text size of 1.000.000 words is sufficient to provide stable probability distributions. Text of this size we treated as a "quasi-population". Probability distribution derived from the "quasi-population" was then correlated with probability distribution obtained on a minimal sample size (32 items) for a given linguistic category (e.g. nouns). Correlation coefficient was treated as a measure of similarity between the two probability distributions. The minimal sample was increased by geometrical progression, up to the size where correlation between distribution derived from the quasi-population and the one derived from an increased sample reached its maximum (r=1). Optimal sample size was established for grammatical forms of nouns, adjectives and verbs. General formalism is proposed that allows estimate of an optimal sample size from minimal sample (i.e. 32 items).

Machine Languages as Media Infrastructures

Author(s): Mujie Li / Language(s): English Issue: 10/2017

This paper argues that through affording communications between machines and humans, machine languages become structures of media. To examine the argument, the paper selects a virtual city model of the Alphatown as a case study and draws on Roman Jakobson’s linguistic theory, at first analyses how the metalanguage functions via the mediums of listing and navigation in the Alphatown, then looks at how the affects of boredom, being lost and strangeness arise from the media infrastructures based upon metalanguage, and analyses where the poetic function lies in the transmission of affects and how the sensual and the material aspects of machine languages emerge thus functioning as poetic. With analysing the metalingual and poetic functions in a diagrammatic thought of machine languages, different levels of machine languages can be considered as media structures, while media structures manifest themselves as material and processual in the meantime.

GOOD FOOD, CLEAN ROOMS AND FRIENDLY STAFF: IMPLICATIONS OF USER-GENERATED CONTENT FOR SLOVENIAN SKIING, SEA AND SPA HOTELS’ MANAGEMENT

Author(s): Uroš Godnov,Tjaša Redek / Language(s): English Issue: 1/2018

The paper studies user-generated evaluations of hotels in three types of destinations in Slovenia (skiing, sea and spa). Using a broad dataset of user-generated evaluations for 28 different hotels and a combination of text mining and standard statistical methods we show how this data provides rich decision-making information. Although numerical evaluations of different destination types can hardly be directly compared due to different guest structure, the results show that guests in general evaluate in their texts primarily the “basics” (room, food/drink, staff). Using a combination of sentiment and novel aspect- based sentiment, hotels can monitor their competitiveness in time, across different types or brands, and use content analysis to further determine sources of competitive advantages in order to enhance performance. The article is the first comprehensive evaluation of Slovenian tourism using on-line peer reviews and provides a toolkit for similar applied analyses.

eTranslation TermBank – naujas tarptautinis projektas Europos Komisijos automatinio vertimo kokybei pagerinti

Author(s): Eglė Pesliakaitė / Language(s): English,Lithuanian Issue: 24/2017

Building a LVCSR System for Romanian: Methods and Challenges

Author(s): Paula Georgiana Zălhan / Language(s): English Issue: 10/2016

The aim of this paper is to present a brief survey in the field of Automatic Speech Recognition (ASR) and major advances made in the last decades of research in order to highlight the fundamental progress that has been made so far. After years of development, the accuracy of Large Vocabulary Continuous Speech Recognition (LVCSR) systems remains the most important research challenge due to several degrading factors such as variations of the context, speakers, and environment. Various methods adopted by research workers when building Large Vocabulary Speech Recognition systems are presented based on the architectural components of such systems. Challenges existing in LVCSR for Romanian language and various techniques to deal these challenges have been presented in chronological order.

Translation, Meaning, Corpora

Author(s): Temenuzhka Seizova-Nankova / Language(s): English,Bulgarian Issue: 1/2014

The paper deals with problems of translation based on the assumptions that ‘translation should preserve meaning’ and that meaning is unconscious (Jackendoff 2012:47). The center of attention are some types of meaning. Corpus data is used to provide empirical evidence and facilitate Bulgarian learners of English in the translation process and the decision making. Solutions are not easy to come by. Some issues are raised with sentence meaning, complementation patterns of behavior, collocational meaning, meaning of compounds and reflexivity, etc. Lexemes are in principle polysemous. But it is lexical items that exhibit different collocational properties and these pose problems for foreign language learning. Two types of valency, qualitative and quantitative, are discussed. Some concrete examples are given as an illustration.

Negative Imperatives With Let’s In The Context Of Corpus Linguistics

Author(s): Petranka Ivanova / Language(s): English Issue: 1/2014

The paper gives short information about imperatives gradually drawing the attention to one of the ways for expressing negative imperative meaning. Nevertheless, two possibilities are in the centre of the research. As let’s seems to occur rarely in negative structures, the questionnaire given to a group of college students shows that they have doubts which of the variants is the correct one. The answer could be found in grammars and the results from the corpus linguistics investigation done with the help of BNC, COCA, and WordSmith Tool confirm it successfully.

EFFECT OF FACULTY ON RESEARCH COOPERATION AND PUBLICATION: EMPLOYING NATURAL LANGUAGE PROCESSING

Author(s): Nitza Davidovitch,Eyal Eckhaus / Language(s): English Issue: 4/2018

This study continues a series of studies on the effectiveness of scientific conferences. This topic has not been sufficiently investigated although it receives large funds, assuming that these conferences have added value for staff members' academic-professional development. Predicated on questionnaires filled by 96 academic staff members from 17 different departments, we found that when choosing conferences to attend, the type of faculty affect the search for cooperation. Moreover, staff members from the Faculty of Natural Sciences attribute more significance to conferences that result in publications than staff members from the Faculty of Health. The Faculty of Engineering creates negative mediation in the correlation between gender and cooperation. Namely, the Faculty of Engineering does not urge cooperation and even has a negative effect, but its effect is evident mainly among men. This finding complements prior research findings showing that women are more inclined to cooperation (Eckhaus & Davidovitch, 2018a). The current findings show that the inclination to cooperation is not related only to gender issues rather the faculty has an effect as well. The current findings might have a contribution to the significance of the faculty as an influential factor of conferences on cooperation – and in fact on the professional development of staff members.

Evaluation System for Online Terminological Databases

Author(s): Dóra Mária Tamás,Eszter Sermann / Language(s): English Issue: 26/2019

This paper attempts to create an extensive classification and evaluation system to offer objective viewpoints for the comprehension, description, classification, evaluation and review of central online termbases of larger organisations. Such a comprehensive classification already exists for lexicographical genres, i.e. printed and online dictionaries, but it is still missing for online termbases. The research is based on previous international and Hungarian literature, the study of several online examples, of existing partial classifications and our previous research. The experimental categorisation with a schematic version in the appendix offers a contribution to a deeper comprehension, the possibility of a thorough description and more attentive evaluation, and ultimately, a growing awareness of the use of online termbases. Providing a new insight may also inspire the development of these beneficial tools which facilitate communication in the field of specialised languages.

ЛИНГВОЭКОЛОГИЯ КАК ФАКТОР ИНФОРМАЦИОННО-КОММУНИКАТИВНОЙ БЕЗОПАСНОСТИ В СЕТИ ИНТЕРНЕТ

Author(s): Ekaterina Sergeevna Palekha / Language(s): Russian Issue: 5-6/2019

The problem of speech deviation and language abuse prevention in the mass-media communication was discussed. The purpose of the research was to identify the potential of linguoecology in struggling with linguistic violence and extremism on the Internet. A discourse description of the text content on the Internet was carried out. Typical language behavior patterns with signs of verbal aggression and deviation on the Internet were analyzed. Examples from the linguistic and criminological practice were provided. The problems faced during the analysis of texts on the Internet were described. A program has been proposed to reduce speech aggression by collaboration of researchers and experts. To solve the problems under study, the methods of complex linguistic and communicative analysis, content and intent analysis, conceptual analysis, discourse analysis, etc. were used. As a result of the study, it was concluded that texts modeling the concepts of anti-values (aggression, violence, death, hopelessness, permissiveness) and used as a means of explicit and hidden speech influence on the addressee should be attributed to dangerous network texts. There is the need to introduce the concept of “linguistic security” of a personality on the Internet. A number of unsolved problems in the field of scientific expertology, as well as socio- and psycholinguistics, were raised. Measures to prevent the practice of linguistic violence were proposed. An important result of the study is the suggestion to expand the practice of applying the concept of “linguoecology” and to use the data of this area of linguistics for developing the concept of “linguistic security”. The results of the study are important for strengthening the linguistic security of a person on the Internet, increasing the comfort and environmental friendliness of the web content, and preventing linguistic toxicity of the Internet environment.

Encoding Textual Variants of the Early Modern Slovenian Poetic Texts in TEI

Author(s): Nina Ditmajer,Matija Ogrin,Tomaž Erjavec / Language(s): English Issue: 1/2019

The paper deals with the problem of encoding the verses and textual variants in the critical edition of Foglar’s Manuscript, a Styrian Baroque hymn book from the mid-eighteenth century. We first show the diplomatic transcript of the verse in selected problematic cases, after which we present the method applied to produce a critical apparatus for approaching textual variants. The base text, i.e. Foglar’s Manuscript, is compared with versions in eight other manuscripts and prints from the eighteenth and early nineteenth centuries. Variants are encoded with XML elements according to the TEI Guidelines as units of the critical apparatus. We highlight some examples of the detailed encoding of rhymes, feet, verse replacements, and textual variants on the spelling, vocabulary and lexical levels of the language. To conclude, we present a number of possibilities for the online display of the electronic diplomatic transcript. The need for the adaptability of these tools to the Slovenian literary tradition is evident.

Explaining IT programming concepts using NSM explications: The case of ‘variable’ and ‘constant’

Author(s): Bartłomiej Biegajło / Language(s): English Issue: 5/2019

The paper seeks to explore a practical application of Natural Semantic Metalanguage in defining two core concepts in computer programming, i.e. the concept of a variable and the concept of a constant. The investigation of both programming concepts is carried out with reference to Apple’s Swift programming language, which is now the dominant language in creating applications designed for Apple’s devices. The explications of a variable and a constant developed in this paper are tentative definitions of the most fundamental functionalities behind the two programming concepts. They are meant to ease the learning experience of programming enthusiasts who are at the early stages of their learning path.

Automatic Identification of Domain Terms: An Approach for Italian

Author(s): Maria Teresa Artese,Isabella Gagliardi / Language(s): English Issue: X/2020

The problem of creating a fully automated specific-domain thesaurus is very topical. The paper presents a novel method to address this problem in the Italian language. The main feature of this approach is the integration of different methods: machine learning classification methods working on the semantic representation of candidate terms, word embed-dings models, able to capture the semantics of words, and a computation of the degree of specialization of a term. The work is in progress and results obtained so far are promising.

Preserving Linguistic Heritage for Generations to Come!

Author(s): Adarsh Appannagari,Manideep Chittineni,Sathvik Jetty,Zinnia Sarkar,Vinod Eslavath,Raj Ramachandran,Emmanuel Ogunshile / Language(s): English Issue: X/2020

Speech to Text is the ability to convert spoken word to text. There are many speech to text conversion applications available for different languages. Tamil is an ancient classical language which is vastly spoken in southern parts of India, Sri Lanka, Malaysia, and Singapore. The speech to text application is de-signed from an indigenous perspective to enable the native speakers to preserve their linguistic heritage. The application was developed using agile methodology and the testing of the application suggested that the existing API do not support in the notion of preserving the language in its original form.