Morphologically-rich Languages Research Articles

Efficient Use of Resources for Statistical Machine Translation

<div class="page" title="Page 1"><div class="layoutArea"><div class="column"><p><span>Machine translation has great potential to expand the audience for ever increasing digital collections. Success of data driven machine translation systems is governed by the volume of parallel data on which these systems are being modelled. The languages which do not have such resources in huge quantity, the optimum utilisation of them can only be assured through their quality. Morphologically rich language like Hindi poses further challenge, due to </span><span>having more number of orthographic inflections for a given word and presence of non-standard word spellings in </span><span>the corpus. This increases the chances of getting more number of words which are unseen in the training corpus. In this paper, the objective is to reduce redundancy of available corpus and utilise the other resources as well, to make best use of resources. Reduction in number of words unseen to the translation model is achieved through text noise removal, spell normalisation and utilising English WordNet (EWN). The test case presented here is for English-Hindi language pair. The results achieved are promising and set example for other morphological rich languages to optimise the resources to improve the performance of the translation system. </span></p></div></div></div>

Efficient Use of Resources for Statistical Machine Translation

<div class="page" title="Page 1"><div class="layoutArea"><div class="column"><p><span>Machine translation has great potential to expand the audience for ever increasing digital collections. Success of data driven machine translation systems is governed by the volume of parallel data on which these systems are being modelled. The languages which do not have such resources in huge quantity, the optimum utilisation of them can only be assured through their quality. Morphologically rich language like Hindi poses further challenge, due to </span><span>having more number of orthographic inflections for a given word and presence of non-standard word spellings in </span><span>the corpus. This increases the chances of getting more number of words which are unseen in the training corpus. In this paper, the objective is to reduce redundancy of available corpus and utilise the other resources as well, to make best use of resources. Reduction in number of words unseen to the translation model is achieved through text noise removal, spell normalisation and utilising English WordNet (EWN). The test case presented here is for English-Hindi language pair. The results achieved are promising and set example for other morphological rich languages to optimise the resources to improve the performance of the translation system. </span></p></div></div></div>

A Sentiment Treebank and Morphologically Enriched Recursive Deep Models for Effective Sentiment Analysis in Arabic

Accurate sentiment analysis models encode the sentiment of words and their combinations to predict the overall sentiment of a sentence. This task becomes challenging when applied to morphologically rich languages (MRL). In this article, we evaluate the use of deep learning advances, namely the Recursive Neural Tensor Networks (RNTN), for sentiment analysis in Arabic as a case study of MRLs. While Arabic may not be considered the only representative of all MRLs, the challenges faced and proposed solutions in Arabic are common to many other MRLs. We identify, illustrate, and address MRL-related challenges and show how RNTN is affected by the morphological richness and orthographic ambiguity of the Arabic language. To address the challenges with sentiment extraction from text in MRL, we propose to explore different orthographic features as well as different morphological features at multiple levels of abstraction ranging from raw words to roots. A key requirement for RNTN is the availability of a sentiment treebank; a collection of syntactic parse trees annotated for sentiment at all levels of constituency and that currently only exists in English. Therefore, our contribution also includes the creation of the first Arabic Sentiment Treebank (A r S en TB) that is morphologically and orthographically enriched. Experimental results show that, compared to the basic RNTN proposed for English, our solution achieves significant improvements up to 8% absolute at the phrase level and 10.8% absolute at the sentence level, measured by average F1 score. It also outperforms well-known classifiers including Support Vector Machines, Recursive Auto Encoders, and Long Short-Term Memory by 7.6%, 3.2%, and 1.6% absolute respectively, all models being trained with similar morphological considerations.

Social media text normalization for Turkish

AbstractText normalization is an indispensable stage in processing noncanonical language from natural sources, such as speech, social media or short text messages. Research in this field is very recent and mostly on English. As is known from different areas of natural language processing, morphologically rich languages (MRLs) pose many different challenges when compared to English. Turkish is a strong representative of MRLs and has particular normalization problems that may not be easily solved by a single-stage pure statistical model. This article introduces the first work on the social media text normalization of an MRL and presents the first complete social media text normalization system for Turkish. The article conducts an in-depth analysis of the error types encountered in Web 2.0 Turkish texts, categorizes them into seven groups and provides solutions for each of them by dividing the candidate generation task into separate modules working in a cascaded architecture. For the first time in the literature, two manually normalized Web 2.0 datasets are introduced for Turkish normalization studies. The exact match scores of the overall system on the provided datasets are 70.40 per cent and 67.37 per cent (77.07 per cent with a case insensitive evaluation).

Learning Morphological Normalization for Translation from and into Morphologically Rich Languages

Abstract When translating between a morphologically rich language (MRL) and English, word forms in the MRL often encode grammatical information that is irrelevant with respect to English, leading to data sparsity issues. This problem can be mitigated by removing from the MRL irrelevant information through normalization. Such preprocessing is usually performed in a deterministic fashion, using hand-crafted rules and yielding suboptimal representations. We introduce here a simple way to automatically compute an appropriate normalization of the MRL and show that it can improve machine translation in both directions.

Providing Morphological Information for SMT Using Neural Networks

Abstract Treating morphologically complex words (MCWs) as atomic units in translation would not yield a desirable result. Such words are complicated constituents with meaningful subunits. A complex word in a morphologically rich language (MRL) could be associated with a number of words or even a full sentence in a simpler language, which means the surface form of complex words should be accompanied with auxiliary morphological information in order to provide a precise translation and a better alignment. In this paper we follow this idea and propose two different methods to convey such information for statistical machine translation (SMT) models. In the first model we enrich factored SMT engines by introducing a new morphological factor which relies on subword-aware word embeddings. In the second model we focus on the language-modeling component. We explore a subword-level neural language model (NLM) to capture sequence-, word- and subword-level dependencies. Our NLM is able to approximate better scores for conditional word probabilities, so the decoder generates more fluent translations. We studied two languages Farsi and German in our experiments and observed significant improvements for both of them.

Text Classification Using Centroid Technique

During these days, till now there is no classifier used for classification of Punjabi documents. Here we have some Punjabi news article facts, which we have necessity to examine with the use of algorithms. Punjabi is an Indo Aryan language spoken in west Punjab (Pakistan) and East Punjab (India). So just a little work has been done in Text Classification for Punjabi language. In this work, assigns to the centroid technique for text News classification to construct structure for Punjabi Language. Huge News is accustoming for training and testing persistence of the classifiers. Language detailed pre-processing technique are beneficial for raw data to originate an identical or reduced-feature lexicon. Punjabi language is a morphological amusing language which makes those responsibilities complicated and hard to do. Statistical physical characteristics of bulk and lexicon are dignified which demonstrate satisfactory and appropriate results and consequences of text pre-processing of each distinct portion. We are residue able to get satisfactory and sufficient out comes or exertion using centroid Classifier. This research paper to reveal the centroid text News classification system to fix firmly for Punjabi Language. News quantity is accustomed for training and analysis purpose of the classifiers. Punjabi language is morphological rich language which originates those jobs and trustworthy complex. Statistical features of body and lexicon are measured which show acceptable results of text pre-processing section. We are able to get appropriate/competent out comes using centroid Classifier Algorithm

A Morphology-Aware Network for Morphological Disambiguation

Agglutinative languages such as Turkish, Finnish andHungarian require morphological disambiguation beforefurther processing due to the complex morphologyof words. A morphological disambiguator is usedto select the correct morphological analysis of a word.Morphological disambiguation is important because itgenerally is one of the first steps of natural languageprocessing and its performance affects subsequent analyses.In this paper, we propose a system that uses deeplearning techniques for morphological disambiguation.Many of the state-of-the-art results in computer vision,speech recognition and natural language processinghave been obtained through deep learning models.However, applying deep learning techniques to morphologicallyrich languages is not well studied. In this work,while we focus on Turkish morphological disambiguationwe also present results for French and German inorder to show that the proposed architecture achieveshigh accuracy with no language-specific feature engineeringor additional resource. In the experiments, weachieve 84.12 , 88.35 and 93.78 morphological disambiguationaccuracy among the ambiguous words forTurkish, German and French respectively.

Multilingual event extraction for epidemic detection.

This paper presents a multilingual news surveillance system applied to tele-epidemiology. It has been shown that multilingual approaches improve timeliness in detection of epidemic events across the globe, eliminating the wait for local news to be translated into major languages. We present here a system to extract epidemic events in potentially any language, provided a Wikipedia seed for common disease names exists. The Daniel system presented herein relies on properties that are common to news writing (the journalistic genre), the most useful being repetition and saliency. Wikipedia is used to screen common disease names to be matched with repeated characters strings. Language variations, such as declensions, are handled by processing text at the character-level, rather than at the word level. This additionally makes it possible to handle various writing systems in a similar fashion. As no multilingual ground truth existed to evaluate the Daniel system, we built a multilingual corpus from the Web, and collected annotations from native speakers of Chinese, English, Greek, Polish and Russian, with no connection or interest in the Daniel system. This data set is available online freely, and can be used for the evaluation of other event extraction systems. Experiments for 5 languages out of 17 tested are detailed in this paper: Chinese, English, Greek, Polish and Russian. The Daniel system achieves an average F-measure of 82% in these 5 languages. It reaches 87% on BEcorpus, the state-of-the-art corpus in English, slightly below top-performing systems, which are tailored with numerous language-specific resources. The consistent performance of Daniel on multiple languages is an important contribution to the reactivity and the coverage of epidemiological event detection systems. Most event extraction systems rely on extensive resources that are language-specific. While their sophistication induces excellent results (over 90% precision and recall), it restricts their coverage in terms of languages and geographic areas. In contrast, in order to detect epidemic events in any language, the Daniel system only requires a list of a few hundreds of disease names and locations, which can actually be acquired automatically. The system can perform consistently well on any language, with precision and recall around 82% on average, according to this paper's evaluation. Daniel's character-based approach is especially interesting for morphologically-rich and low-resourced languages. The lack of resources to be exploited and the state of the art string matching algorithms imply that Daniel can process thousands of documents per minute on a simple laptop. In the context of epidemic surveillance, reactivity and geographic coverage are of primary importance, since no one knows where the next event will strike, and therefore in what vernacular language it will first be reported. By being able to process any language, the Daniel system offers unique coverage for poorly endowed languages, and can complete state of the art techniques for major languages.

Idioms-Proverbs Lexicon for Modern Standard Arabic and Colloquial Sentiment Analysis

Although, the fair amount of works in sentiment analysis (SA) and opinion mining (OM) systems in the last decade and with respect to the performance of these systems, but it still not desired performance, especially for morphologically-Rich Language (MRL) such as Arabic, due to the complexities and challenges exist in the nature of the languages itself. One of these challenges is the detection of idioms or proverbs phrases within the writer text or comment. An idiom or proverb is a form of speech or an expression that is peculiar to itself. Grammatically, it cannot be understood from the individual meanings of its elements and can yield different sentiment when treats as separate words. Consequently, In order to facilitate the task of detection and classification of lexical phrases for automated SA systems, this paper presents AIPSeLEX a novel idioms/ proverbs sentiment lexicon for modern standard Arabic (MSA) and colloquial. AIPSeLEX is manually collected and annotated at sentence level with semantic orientation (positive or negative). The efforts of manually building and annotating the lexicon are reported. Moreover, we build a classifier that extracts idioms and proverbs, phrases from text using n-gram and similarity measure methods. Finally, several experiments were carried out on various data, including Arabic tweets and Arabic microblogs (hotel reservation, product reviews, and TV program comments) from publicly available Arabic online reviews websites (social media, blogs, forums, e-commerce web sites) to evaluate the coverage and accuracy of AIPSeLEX.

A standard tag set expounding traditional morphological features for Arabic language part-of-speech tagging

The SALMA Morphological Features Tag Set (SALMA, Sawalha Atwell Leeds Morphological Analysis tag set for Arabic) captures long-established traditional morphological features of grammar and Arabic, in a compact yet transparent notation. First, we introduce Part-of-Speech tagging and tag set standards for English and other European languages, and then survey Arabic Part-of-Speech taggers and corpora, and long-established Arabic traditions in analysis of morphology. A range of existing Arabic Part-of-Speech tag sets are illustrated and compared; and we review generic design criteria for corpus tag sets. For a morphologically-rich language like Arabic, the Part-of-Speech tag set should be defined in terms of morphological features characterizing word structure. We describe the SALMA Tag Set in detail, explaining and illustrating each feature and possible values. In our analysis, a tag consists of 22 characters; each position represents a feature and the letter at that location represents a value or attribute of the morphological feature; the dash ‘-’ represents a feature not relevant to a given word. The first character shows the main Parts of Speech, from: noun, verb, particle, punctuation, and Other (residual); these last two are an extension to the traditional three classes to handle modern texts. ‘Noun’ in Arabic subsumes what are traditionally referred to in English as ‘noun’ and ‘adjective’. The characters 2, 3, and 4 are used to represent subcategories; traditional Arabic grammar recognizes 34 subclasses of noun (letter 2), 3 subclasses of verb (letter 3), 21 subclasses of particle (letter 4). Others (residuals) and punctuation marks are represented in letters 5 and 6 respectively. The next letters represent traditional morphological features: gender (7), number (8), person (9), inflectional morphology (10) case or mood (11), case and mood marks (12), definiteness (13), voice (14), emphasized and non-emphasized (15), transitivity (16), rational (17), declension and conjugation (18). Finally there are four characters representing morphological information which is useful in Arabic text analysis, although not all linguists would count these as traditional features: unaugmented and augmented (19), number of root letters (20), verb root (21), types of nouns according to their final letters (22). The SALMA Tag Set is not tied to a specific tagging algorithm or theory, and other tag sets could be mapped onto this standard, to simplify and promote comparisons between and reuse of Arabic taggers and tagged corpora.

Parsing Morphologically Rich Languages: Introduction to the Special Issue

Parsing is a key task in natural language processing. It involves predicting, for each natural language sentence, an abstract representation of the grammatical entities in the sentence and the relations between these entities. This representation provides an interface to compositional semantics and to the notions of “who did what to whom.” The last two decades have seen great advances in parsing English, leading to major leaps also in the performance of applications that use parsers as part of their backbone, such as systems for information extraction, sentiment analysis, text summarization, and machine translation. Attempts to replicate the success of parsing English for other languages have often yielded unsatisfactory results. In particular, parsing languages with complex word structure and flexible word order has been shown to require non-trivial adaptation. This special issue reports on methods that successfully address the challenges involved in parsing a range of morphologically rich languages (MRLs). This introduction characterizes MRLs, describes the challenges in parsing MRLs, and outlines the contributions of the articles in the special issue. These contributions present up-to-date research efforts that address parsing in varied, cross-lingual settings. They show that parsing MRLs addresses challenges that transcend particular representational and algorithmic choices.

Portuguese text generation using factored language models

Abstract As in many other natural language processing (NLP) fields, the use of statistical methods is now part of mainstream natural language generation (NLG). In the development of systems of this kind, however, there is the issue of data sparseness, a problem that is particularly evident in the case of morphologically-rich languages such as Portuguese. This work presents a shallow surface realisation system that makes use of factored language models (FLMs) of Portuguese to overcome some of these difficulties. The system combines FLMs trained on a large corpus with a number of NLP resources that have been made publicly available by the Brazilian NLP research community in recent years, such as corpora, dictionaries, thesauri and others. Our FLM-based approach to surface realisation has been successfully applied to the generation of Brazilian newspapers headlines, and the results are shown to outperform a number of statistical and non-statistical baseline systems alike.

Graph-Based Dependency Parsing Using Dynamic Features in Korean

In this paper, we propose a graph-based dependency parsing model using dynamic features. The dynamic features, which are derived from the partial tree during bottom-up chart parsing, play an important role in finding the correct heads. Based on the beneficial aspects of the head-final property of Korean, we suggest a variant of the CYK parsing algorithm with an O(n3) complexity which has the ability to search for the maximum spanning tree (MST) from all projective trees. Compared to Eisner's algorithm, this algorithm is time-efficient in parsing all other head-final languages. We also explain why the graph-based approaches are seldom attempted in parsing morphologically rich languages (MRLs) and suggest a morpheme-level dependency representation as a solution. Finally, we evaluate the proposed method on the KAIST corpus using two discriminative online learning methods (the averaged perceptron algorithm and passive-aggressive algorithm) and show that the trained parser achieves the best reported parsing accuracy.

Dealing With Complex Linguistic Annotations Within a Language Processing Framework

In this paper we present AWA, a general purpose Annotation Web Architecture for representing, storing, and accessing the information produced by different linguistic processors. The objective of AWA is to establish a coherent and flexible representation scheme that will be the basis for the exchange and use of linguistic information. In morphologically-rich languages as Basque it is necessary to represent and provide easy access to complex phenomena such as intraword structure, declension, derivation and composition features, constituent discontinousness (in multiword expressions) and so on. AWA provides a well-suited schema to deal with these phenomena. The annotation model relies on XML technologies for data representation, storage and retrieval. Typed feature structures are used as a representation schema for linguistic analyses. A consistent underlying data model, which captures the structure and relations contained in the information to be manipulated, has been identified and implemented. AWA is integrated into LPAF, a multilayered Language Processing and Annotation Framework, whose goal is the management and integration of diverse NLP components and resources. Moreover, we introduce EULIA, an annotation tool which exploits and manipulates the data created by the linguistic processors. Two real corpora have been processed and annotated within this framework.

Multi-Alignment Templates Induction

This paper examins approaches for translation between English and morphology-rich languages. Experiment with English-Russian and English-Lithuanian revels that “pure” statistical approaches on 10 million word corpus gives unsatisfactory translation. Then, several Web-available linguistic resources are suggested for translation. Syntax parsers, bilingual and semantic dictionaries, bilingual parallel corpus and monolingualWeb-based corpus are integrated in one comprehensive statistical model. Multi-abstraction language representation is used for statistical induction of syntactic and semantic transformation rules called multi-alignment templates. The decodingmodel is described using the feature functions, a log-linear modeling approach and A* search algorithm. An evaluation of this approach is performed on the English-Lithuanian language pair. Presented experimental results demonstrates that the multi-abstraction approach and hybridization of learning methods can improve quality of translation.

Meibomian Gland Dysfunction in Computer Vision Syndrome

This paper presents a multilingual news surveillance system applied to tele-epidemiology. It has been shown that multilingual approaches improve timeliness in detection of epidemic events across the globe, eliminating the wait for local news to be translated into major languages. We present here a system to extract epidemic events in potentially any language, provided a Wikipedia seed for common disease names exists.The Daniel system presented herein relies on properties that are common to news writing (the journalistic genre), the most useful being repetition and saliency. Wikipedia is used to screen common disease names to be matched with repeated characters strings. Language variations, such as declensions, are handled by processing text at the character-level, rather than at the word level. This additionally makes it possible to handle various writing systems in a similar fashion.As no multilingual ground truth existed to evaluate the Daniel system, we built a multilingual corpus from the Web, and collected annotations from native speakers of Chinese, English, Greek, Polish and Russian, with no connection or interest in the Daniel system. This data set is available online freely, and can be used for the evaluation of other event extraction systems.Experiments for 5 languages out of 17 tested are detailed in this paper: Chinese, English, Greek, Polish and Russian. The Daniel system achieves an average F-measure of 82% in these 5 languages. It reaches 87% on BEcorpus, the state-of-the-art corpus in English, slightly below top-performing systems, which are tailored with numerous language-specific resources. The consistent performance of Daniel on multiple languages is an important contribution to the reactivity and the coverage of epidemiological event detection systems.Most event extraction systems rely on extensive resources that are language-specific. While their sophistication induces excellent results (over 90% precision and recall), it restricts their coverage in terms of languages and geographic areas. In contrast, in order to detect epidemic events in any language, the Daniel system only requires a list of a few hundreds of disease names and locations, which can actually be acquired automatically. The system can perform consistently well on any language, with precision and recall around 82% on average, according to this paper's evaluation. Daniel's character-based approach is especially interesting for morphologically-rich and low-resourced languages. The lack of resources to be exploited and the state of the art string matching algorithms imply that Daniel can process thousands of documents per minute on a simple laptop. In the context of epidemic surveillance, reactivity and geographic coverage are of primary importance, since no one knows where the next event will strike, and therefore in what vernacular language it will first be reported. By being able to process any language, the Daniel system offers unique coverage for poorly endowed languages, and can complete state of the art techniques for major languages.

Language and linguistics: Aims, perspectives, and duties of linguistics Ed. by Pierre Swiggers (review)

178 LANGUAGE, VOLUME 75, NUMBER 1 (1999) deal with acquisition of case, agreement, negation, questions; its applicability to morphologically-rich languages; and its explanatory adequacy. From the full-competence camp, an ambitious paper by Nina Hyams derives a common source for children's null subjects, optional infinitives, and determinerless DPs in the underspecification of FCs. On the other hand, Harald Clahsen, Sonja Eisenbeiss, and Martina Penke claim that learning the lexicon drives growth of syntactic knowledge: For example, acquisition of verbal inflection leads children to develop the phrase structure underlying German V2 facts. Zvi Penner and Jürgen Weissenborn counter with data from Swiss versus Standard German. They find no correlation between acquisition ofinflection and ofthe properties of DPs, instead attributing the two dialects' different acquisitional paths to canonical versus noncanonical triggers for parameter setting. Jürgen M. Meisel and Maria-Jose Ezeizabarrena report that in Basque acquisition of subject agreement precedes thatofobject agreement. This challenges the assumption that the order of acquisition of FCs reflects phrase-structural position since AGR(S) dominates AGR(O). Maria Teresa Guasti shows that unlike children learning English, children learning Italian produce adult-like negative questions, which fact she identifies with different features of CP in the two languages. A paper by Liliane Haegeman similarly brings acquisitional data to bear on cross-linguistic differences. She claims that object clitics appear in early child French but not Dutch because of their different landing sites in the two languages. The topic of clitics is carried forward in two more papers. Cornelia Hamann, Luigi Rizzi, and Uli H. Frauenfelder find that French-learning children place clitics correctly, though subject clitics are acquired before object clitics and rarely appear with untensed verbs. Lydia White demonstrates that English -speaking children learning French as a second language both distinguish clitics from pronouns and distribute them appropriately despite the presumed lack of a native-language model for their phrase structure. Two additional papers wrestle with the application of the minimalist program to studies of acquisition: Christer Platzack conceptualizes acquisition as a process of recognizing which features of FCs are strong and so must be checked before spell-out, testing his proposal against various data including some on acquisition of Swedish. Thomas Roeper speculates about how merger theory may make sense of diverse empirical facts, assuming that children create unique maximal projections both in the sense that they are unlike those in adult grammar and they are specific to particular lexical items. The editor gives the last word to Martin Atkinson , who closes the text with a feisty attack on several theoretical, methodological, and data-interpretive assumptions running through the other papers . Atkinson's objections deserve attention, and raising them only strengthens the volume, as to end on this critical note is a tribute to the self-confidence of generative perspectives on language acquisition. [Margaret Thomas, Boston College.] Language and linguistics: Aims, perspectives , and duties of linguistics. Ed. by Pierre Swiggers. Leuven: Peeters, 1997. Pp. xvii, 90. This intriguing book consists of transcriptions of three interviews conducted in 1993-96 by the Belgian linguist Pierre Swiggers. All three interviews appeared in the journal Orbis but are assembled here with a preface in English and French. The interviewees are André-Georges Haudricourt (1911-96), Henry M. Hoenigswald (b. 1915), and Robert H. Robins (b. 1921). S's aim is to record the personal experiences and professional opinions ofthree 'privileged witnesses of the development of linguistics in the past half-century' (x). In justifying the choice of interviewees , S cites as common threads in their work regard for linguistic data, systematicity of approach, and concern for basic disciplinary issues. Granted that linguists (should) generally share these virtues, Haudricourt, Hoenigswald, and Robins are further distinguished in that each has worked on both familiar European and 'exotic' languages (Haudricourt in the Far East, Hoenigswald with Etruscan, and Robins with Yurok and Sundanese). S also attributes to each a leading role in modern linguistics and 'outright refusal of all kinds of dogmatism' (vii)—though perhaps the accuracy of the latter claim should be left to the reader's judgment. What one could add to the list of shared characteristics is that all three men, in conversation with...