Universal Networking Language Research Articles

Machine translation using deep learning for universal networking language based on their structure

This paper presents a deep learning-based machine translation (MT) system that translates a sentence of subject-object-verb (SOV) structured language into subject-verb-object (SVO) structured language. This system uses recurrent neural networks (RNNs) and Encodings. Encode embedded RNNs generate a set of numbers from the input sentence, where the second RNNs generate the output from these sets of numbers. Three popular datasets of SOV structured language i.e., EMILLE corpus, Prothom-Alo corpus and Punjabi Monolingual Text Corpus ILCI-II are used as two different case-study to validate. In our experimental case-study 1, for the EMILLE corpus and Prothom-Alo corpus dataset, we have achieved 0.742, 4.11 and 0.18, respectively as Bilingual Evaluation Understudy (BLEU), NIST (metric) and tertiary entrance rank scores. Another case-study for Punjabi Monolingual Text Corpus ILCI-II dataset achieved a BLEU score of 0.75. Our results can be compared with the state-of-the-art results.

Sanskrit to universal networking language EnConverter system based on deep learning and context-free grammar

Machine Translation is a mechanism of transforming text from one language to another with the help of computer technology. Earlier in 2018, a machine translation system had been developed by the authors that translate Sanskrit text to Universal Networking Language expressions and was named as SANSUNL. The work presented in this paper is an extension of SANSUNL system by enhancing POS tagging, Sanskrit language processing and parsing. A Sanskrit stemmer having 23 prefixes and 774 suffixes with grammar rules are used for stemming the Sanskrit sentence in the proposed system. Bidirectional long short-term memory (Bi-LSTM) and stacked LSTM deep neural network models have been used for part of speech tagging of the input Sanskrit text. A tagged dataset of around 400 k entries for Sanskrit have been used for training and testing the neural network models. Proposed Sanskrit context-free grammar has been used with CYK parser to perform the parsing of the input sentence. Size of the Sanskrit-Universal Word dictionary has been increased from 15000 to 25000 entries. Approximately 1500 UNL generation rules have been used to resolve the 46 UNL relations. Four datasets UC-A1, UC-A2, Spanish server gold standard dataset, and 500 Sanskrit sentences taken from the general domain have been used for validating the system. The proposed system is evaluated on BLEU and Fluency score metrics and has reported an efficiency of 95.375%.

Detecting paraphrases in the Marathi language

Paraphrasing refers to writing that either differs in its textual content or is dissimilar in rearrangement of wordsbut conveys the same meaning. Identifying a paraphrase is exceptionally important in various real life applicationssuch as Information Retrieval, Plagiarism Detection, Text Summarization and Question Answering. A large amountof work in Paraphrase Detection has been done in English and many Indian Languages. However, there is noexisting system to identify paraphrases in Marathi. This is the first such endeavor in the Marathi Language.A paraphrase has differently structured sentences, and since Marathi is a semantically strong language, thissystem is designed for checking both statistical and semantic similarities of Marathi sentences. Statistical similaritymeasure does not need any prior knowledge as it is only based on the factual data of sentences. The factual datais calculated on the basis of the degree of closeness between the word-set, word-order, word-vector and word-distance. Universal Networking Language (UNL) speaks about the semantic significance in sentence without anysyntactic points of interest. Hence, the semantic similarity calculated on the basis of generated UNL graphs for twoMarathi sentences renders semantic equality of two Marathi sentences. The total paraphrase score was calculatedafter joining statistical and semantic similarity scores, which gives a judgment on whether there is paraphrase ornon-paraphrase about the Marathi sentences in question.

Detecting Paraphrases in Marathi Language

Paraphrasing refers to the sentences that either differs in their textual content or dissimilar in rearrangement of words but convey the same meaning. Identifying a paraphrase is exceptionally important in various real life applications such as Information Retrieval, Plagiarism Detection, Text Summarization and Question Answering. A large amount of work in Paraphrase Detection has been done in English and many Indian Languages. However, there is no existing system to identify paraphrases in Marathi. This is the first such endeavor in Marathi Language. A paraphrase has different structured sentences and Marathi being semantically strong language hence this system is designed for checking both statistical and semantic similarity of Marathi sentences. Statistical similarity measure does not need any prior knowledge as it is only based on the factual data of sentences. The factual data is calculated on the basis of the degree of closeness between the word-set, word-order, word-vector and word-distance. Universal Networking Language (UNL) speaks about the semantic significance in the sentence without any syntacticpointofinterest.Hence, these mantic similarity calculated on the basis of generated UNL graphs for two Marathi sentences renders semantic equality of two Marathi sentences. The total para phrases core was calculated after joining statistical and semantic similarity scores which gives the judgement of being paraphrase or non-paraphrase about the Marathi sentences.

Bangla DeConverter for Extraction of BanglaText from Universal Networking Language

The people in Bangladesh and two states (i.e., Tripura and West Bengal) in India, which is about 230 million of the world population, use Bengali as their first dialect. However, very few numbers of resources and tools are available for this language. This paper presents a Bangla DeConverter to extract Bangla texts from Universal Networking Language (UNL). It explains and illustrates the different phases of the proposed Bangla DeConverter. The syntactic linearization, the implementation of the results of the proposed Bangla DeConverter, and the extraction of a Bangla sentence from UNL expressions are presented in this paper. The Bangla DeConverter has been tested on UNL expressions of 300 Bangla sentences using a Russian and English Language Server. The proposed system generates 90% syntactically and semantically correct Bangla sentences with a UNL Bilingual Evaluation Understudy (BLEU) score of 0.76.

Bootstrapping of Semantic Relation Extraction for a Morphologically Rich Language

This article focuses on the use of a bootstrapping approach for the extraction of semantic relations that exist between two different concepts in a Tamil text. The proposed system, bootstrapping approach to semantic UNL relation extraction (BASURE) extracts generic relations that exist between different components of a sentence by exploiting the morphological richness of Tamil. Tamil is essentially a partially free word order language which means that semantic relations that exist between the concepts can occur anywhere in the sentence not necessarily in a fixed order. Here, the authors use Universal Networking Language (UNL), an Interlingua framework, to represent the word-based features and aim to define UNL semantic relations that exist between any two constituents in a sentence. The morphological suffix, lexical category and UNL semantic constraints associated with a word are defined as tuples of the pattern used for bootstrapping. Most systems define the initial set of seed patterns manually. However, this article uses a rule-based approach to obtain word-based features that form tuples of the patterns. A bootstrapping approach is then applied to extract all possible instances from the corpus and to generate new patterns. Here, the authors also introduce the use of UNL ontology to discover the semantic similarity between semantic tuples of the pattern, hence, to learn new patterns from the text corpus in an iterative manner. The use of UNL Ontology makes this approach general and domain independent. The results obtained are evaluated and compared with existing approaches and it has been shown that this approach is generic, can extract all sentence based semantic UNL relations and significantly increases the performance of the generic semantic relation extraction system.

SANSUNL: A Sanskrit to UNL Enconverter System

ABSTRACT Machine translation has become the most challenging as well as demanding field of research and as a result, to develop a fully automatic 100% correct machine translation system has not been achieved till date. More than 7000 of languages are being used worldwide for communication purposes and in India alone 22 languages have been recognized as official languages for communication purposes. Various efforts have been made to solve this problem by different researchers using different approaches. With the enhancement in technology and internet, automatic translation has become more important to share the knowledge. The authors in this paper described a mechanism of translation using machines for one of the ancient Indian Language Sanskrit to the latest computer understandable language Universal Networking Language (UNL). The proposed system uses two databases: one for the analysis purpose consists of more than 300 rules and the other for the generation of UNL which consists of about 1500 rules. For the performance evaluation, the authors selected 500 Sanskrit sentences in such a way that covers maximum UNL relations from various sources. The sentences are manually converted from English to Sanskrit language taken from Spanish Language center, Hindi to English language taken from IIT Bombay Hindi–UNL expressions and NCERT class 8th Sanskrit book titled “Ruchita Part 3”. The proposed system achieves a BLEU score of 0.85. The proposed system gives 93.18% efficiency in resolving UNL relation successfully.

UNLization of Punjabi text for natural language processing applications

During the last couple of years, in the field of Natural Language Processing, UNL (i.e., Universal Networking Language) immense research activities have been witnessed. This paper illustrates UNLization of Punjabi Natural Language for UC-A1, UGO-A1, and AESOP-A1 with IAN (i.e., Interactive Analyzer) tool using X-Bar approach. This paper also discusses the UNLization process in depth, step-by-step with the help of tree diagrams and tables.

Construction of Word Dictionary for Bangla Vowel Ended Roots and Its Verbal Inflexions in UNL Based Machine Translation Scheme

This paper focuses on the development of word dictionary of Bangla vowel ended roots and their verbal inflexions to incorporating them into an interlingua representation called Universal Networking Language (UNL) processors. A considerable amount of work has been done on the development of Bangla morphological analysis on verbs, nouns, prefixes and suffixes for machine translation. As far as the researchers are aware, no attempts, however, have been made to integrate the previous developments on Bangla vowel ended roots and inflexions for a concrete computational output. This paper attempts to bridge the gap on Bangla vowel ended roots and inflexions in the framework of UNL system aiming to produce a Bangla word dictionary for UNL. The paper analyzes the Bangla vowel ended roots and verbal inflexions and develops their formats in the UNL structure. Dictionary entries of all vowel ended roots and their inflexions are developed in order to generate associated verbs for sentences. These verbs can be used to convert Bangla native language sentences into UNL expressions using semantic rules, which are then converted into required native languages using the language specific generation rules. Conversion of a Bangla language sentence into UNL expression has also been shown in this paper.

Towards UNL-based machine translation for Moroccan Amazigh language

Amazigh languages, also called Berber, belong to the Afro-Asiatic languages (Hamito-Semitic) family. They are a family of similar and closely related languages and dialects indigenous to North Africa. They are spoken in Morocco, Algeria, and some populations in Libya, Tunisia, northern Mali, western and northern Niger, northern Burkina Faso, Mauritania, and in the Siwa Oasis of Egypt. Large Berber-speaking migrant communities have been living in Western Europe since the 1950s. In this paper, we study the Standard Moroccan Amazigh. It became a constitutionally official language of Morocco, in 2011. However, it is still considered as a less resourced language. So, it is time to develop linguistic resources and applications for processing automatically this language, in order to ensure its survival and promotion by integrating it into the new information and communication technologies (NICT). In this context and in the perspective to produce a Universal Networking Language (UNL) based machine translation system for this language, we have undertaken the creation of the Amazigh-UNL dictionary, as a first step of linguistic resources' development required by the UNL system to achieve translation. Thus, this paper is focused on presenting linguistic features' implementation such as morphological, syntactical and semantic information of the Amazigh language.

Exploring content selection strategies for Multilingual Multi-Document Summarization based on the Universal Network Language (UNL)

Abstract: Multilingual Multi-Document Summarization aims at ranking the sentences of a cluster with (at least) 2 news texts (1 in the user’s language and 1 in a foreign language), and select the top-ranked sentences for a summary in the user’s language. We explored three concept-based statistics and one superficial strategy for sentence ranking. We used a bilingual corpus (Brazilian Portuguese-English) encoded in UNL (Universal Network Language) with source and summary sentences aligned based on content overlap. Our experiment shows that “concept frequency normalized by the number of concepts in the sentence” is the measure that best ranks the sentences selected by humans. However, it does not outperform the superficial strategy based on the position of the sentences in the texts. This indicates that the most frequent concepts are not always contained in first sentences, usually selected by humans to build the summaries because they convey the main information of the collection.Keywords: content selection; concept; statistical measure; multilingual corpus; multi-document summarization.Keywords: content selection; concept; statistical measure; multilingualcorpus; multi-document summarization.Resumo: O objetivo da Sumarização Automática Multilíngue Multidocumento é ranquear as sentenças de uma coleção com ao menos duas notícias (1 na língua do usuário e 1 em língua estrangeira) e selecionar as mais bem pontuadas para compor um sumário na língua do usuário. Exploramos três estatísticas conceituais e uma estratégia superficial para criar um ranque das sentenças quanto à relevância. Para tanto, utilizamos um corpus bilíngue (português-inglês) anotado via UNL (Universal Network Language) e com textos-fonte e sumários alinhados em nível sentencial. A avaliação indica que a estatísticadenominada frequência de conceitos normalizada pelo número de conceitos da sentença é a que melhor reproduz o ranqueamento humano. Essa medida, entretanto, não supera a estratégia superficial baseada na posição das sentenças. Isso indica que os conceitos mais frequentes do cluster nem sempre estão contidos nas primeiras sentenças dos textosfonte, usualmente selecionadas pelos humanos para compor os sumários porque veiculam a informação principal da coleção.Palavras-chave: seleção de conteúdo; conceito; medida estatística; corpus multilíngue; sumarização multidocumento.

Generation of UNL Attributes and Resolving Relations for Bangla EnConverter

Conversion of Bangla language to another native language and another language to Bangla language using Universal Networking Language (UNL) is highly demanding due to rapidly increasing the usage of Internet-based applications. UNL has been used by various researchers as an inter-lingual approach for an Automated Machine Translation (AMT) scheme. This article presents a novel work on construction of EnConverter for Bangla language with a special focus on generation of UNL attributes and resolving relations of Bangla text. The architecture of Bangla EnConverter, algorithms for understanding the Bangla input sentence; resolution of UNL relations; and attributes for Bangla text/language are also explained in this article. This article highlights the analysis rules for EnConverter and indicates its usage in generation of UNL expressions. This article presents the results of implementation of Bangla EnConverter and compares these with the system available at Russian and English Language Server.

Query Focused Summary Generation System using Unique Discourse Structure

In this paper, the authors propose a query focussed summary generation system which is constructed on top of a unique language-independent discourse structure. The discourse structure is comprised of three text representation techniques, namely, Universal Networking Language (UNL), Rhetorical Structure Theory (RST) and san?gatis. The discourse structure is indexed based on a concept called sutra. Both sutra and san?gatis have been used in ancient Indian literatures. The proposed approach is tested using Forum for Information Retrieval (FIRE) corpus and a performance comparison has been done with the one of the state-of-art approaches.

English to Tamil machine translation system using universal networking language

This paper proposes English to Tamil machine translation system, using the universal networking language (UNL) as the intermediate representation. The UNL approach is a hybrid approach of the rule and knowledge-based approaches to machine translation. UNL is a declarative formal language, specifically designed to represent semantic data extracted from a natural language text. The input English sentence is converted to UNL (enconversion), which is then converted to a Tamil sentence (deconversion) by ensuring that the meaning of the input sentence is preserved. The representation of UNL was modified to suit the translation process. A new sentence formation algorithm was also proposed to rearrange the translated Tamil words to sentences. The translation system was evaluated using bilingual evaluation understudy (BLEU) score. A BLEU score of 0.581 was achieved, which is an indication that most of the information in the input sentence is retained in the translated sentence. The scores obtained using the UNL based approach were compared with existing approaches to translation, and it can be concluded that the UNL is a more suited approach to machine translation.

Templates of Dictionary Entries of Bangla Repetition Words for Universal Networking Language

This paper presents the development of dictionary entries of Bangla repetition words to integrate them into an interlingua representation called universal networking language (UNL) processors. Here, we particularly emphasize on categorization and construction of dictionary for the repetition words with the goal to prepare a Bangla-UNL dictionary. We have outlined the format of dictionary entries of different types of repetition words that come from verb roots and assigned head words and attributes for those words based on the framework of UNL provided by the UNL center.

Multilingual Automatic Translation Based on UNL: A Case Study for the Vietnamese Language

In the field of natural language processing, Universal Networking Language (UNL) has been used by various researchers as an inter-lingual approach to automatic machine translation. The UNL system consists of two main components, namely, EnConverter for converting text from a source language to UNL, and DeConverter for converting from UNL to a target language. Currently, many projects are researching how to apply UNL to different languages. In this paper, we introduce the tools that are UNL’s applications and discuss how to reuse them to encode a Vietnamese sentence into UNL expressions and decode UNL expressions into a Vietnamese sentence. The testing was done with about 1,000 Vietnamese sentences (a dictionary that includes 4573 entries and 3161 rules). In addition, we compare the proportion of sentences translated based on a direct method (Google Translator) and another one based on UNL.

A Multilingual Cross-Domain Client Application Prototype for UNL-ization and NL-ization for NLP Applications

Ever since the Universal Networking Language (UNL) programme started in 1996, researchers and computational linguists across the globe have actively participated in this. English, Chinese, Georgian, Greek, Hindi, Punjabi, Portuguese, Russian, Ukrainian, Vietnamese, Slovenian, etc., are some of the languages which are part of the UNL programme. Although previous years have witnessed significant interest and application development in UNL, UNL can be exploited for several different tasks in natural language engineering, such as multilingual document generation, summarization, text simplification, information retrieval, and semantic reasoning. Although UNL is language-independent artificial language, yet, existing UNL-based applications are standalone in the sense that they are language specific which supports the local natural language of the application developer. This article illustrates the cross-domain client application prototype that has been built to support the development of multilingual UNL-based applications where researchers, developers, or computational linguists across the globe can use this common application for UNL-ization and NL-ization. Output of the proposed system can further be utilized by any other UNL-based application. The proposed system is evaluated on the basis of two measures, i.e. accuracy and correctness. The accuracy of the proposed system is 100%. However, correctness of the proposed system depends upon the F-Measure of UNL-ization and NL-ization modules which has been consistently greater than 0.95 (on a scale of 0–1) for Punjabi language.

Translation Challenges and Universal Networking Language

This paper presents challenges being faced in designing automatic translation software. There are many approaches to automatic translation like Direct, Rule based, Transfer based, Statistical based and Interlingua. A brief description of all the approaches and their advantages and drawbacks are discussed. Universal Networking Language (UNL) based on Interlingua approach can be used especially for translation among multiple languages because it requires knowledge of UNL and of the language which user wants UNL to support. User can then get translated text in any of the languages supported by UNL without even being oblivious to any other language. It is less expensive approach also. This paper also gives brief introduction to UNL and how it can overcome many of the challenges in translation.

Towards UNL based machine translation for Moroccan Amazigh language

Amazigh languages, also called Berber, belong to the Afro-Asiatic languages (Hamito-Semitic) family. They are a family of similar and closely related languages and dialects indigenous to North Africa. They are spoken in Morocco, Algeria, and some populations in Libya, Tunisia, northern Mali, western and northern Niger, northern Burkina Faso, Mauritania, and in the Siwa Oasis of Egypt. Large Berber-speaking migrant communities have been living in Western Europe since the 1950s. In this paper, we study the Standard Moroccan Amazigh. It became a constitutionally official language of Morocco, in 2011. However, it is still considered as a less resourced language. So, it is time to develop linguistic resources and applications for processing automatically this language, in order to ensure its survival and promotion by integrating it into the new information and communication technologies (NICT). In this context and in the perspective to produce a Universal Networking Language (UNL) based machine translation system for this language, we have undertaken the creation of the Amazigh-UNL dictionary, as a first step of linguistic resources' development required by the UNL system to achieve translation. Thus, this paper is focused on presenting linguistic features' implementation such as morphological, syntactical and semantic information of the Amazigh language.

Representation of Universal Quantifier in Bulgarian Language with Universal Networking Language

The paper presents interpretation of universal quantifier in Bulgarian language with Universal Networking Language (UNL). It analyzes semantic and linguistic properties of universal quantifier and describes the parts of speech which function as quantifiers. In UNL frameworks they are presented as a lexical database with related grammar and semantic features using idea of synonymic semantic representation. The formal approach presented may be used also to language learning and understanding.