Search

This project examined whether there is a lexicon of messages commonly used by vocational rehabilitation (VR) counselors with deaf clients that might be stored digitally on a CD-ROM disk with the eventual purpose of developing a voice-to-sign translation system to facilitate communication between hearing and deaf individuals. Surveys of counselors and reception personnel from 34 VR offices were analyzed, and a set of commonly used messages was identified. Messages were evaluated for length, complexity, and evidence of enabling/disabling language. Ninety-three percent of the total messages (N=1080) identified contained no references to enabling or disabling language. Examination of the remaining 88 messages revealed only three instances of "disabling language". Calculations indicated that it is feasible to create a computer program to translate voice or text input to manual signs. Results suggest the inclusion in such a disk of 125 commonly used messages and 120 individually signed words. Development of computer strategies to implement recommendations is discussed. (Contains 16 references.) (DB)

The sixth session of IT@EDU98 consisted of seven papers on the topic of the learning machine--Vietnamese based human-computer interface, and was chaired by Phan Viet Hoang (Informatics College, Singapore). "Knowledge Based Approach for English Vietnamese Machine Translation" (Hoang Kiem, Dinh Dien) presents the knowledge base approach, which consists of concepts such as things, actions, relations, and attributes, organized on the structure of inheritance hierarchy. "A Learning Algorithm for Feature Selection Based on Genetic Approach" (Nguyen Dinh Thuc, Le Hoai Bac) presents a genetic algorithm that chooses relevant features from a set of given features, for feature selection based on the correlation among the features and between every feature and given target curve. "Artificial Neural Network for Color Classification" (Tran Cong Toai) examines several neural network models, their learning schemes, and their effectiveness in color classification. "Synthesizing and Recognizing Vietnamese Speech" (Hoang Kiem, Nguyen Minh Triet, Vo Tuan Kiet, Thai Hung Van, Luu Duc Hien, Bui Tien Len) presents algorithms applied successfully in Vietnamese isolated word recognition and Vietnamese synthesis. "On-Line Character Recognition" (Nguyen Thanh Phuong) presents a real-time handwriting character recognition system based on a structural approach. "Data Mining and Knowledge Acquisition from a Database" (Hoang Kiem, Do Phuc) considers how to use multi-dimensional data model (MDDM) for mining rules in a large database. "Genetic Algorithm for Initiative of Neural Networks" (Nguyen Dinh Thuc, Tan Quang Sang, Le Ha Thanh, Tran Thai Son) describes a procedure for initiative of neural networks based on genetic algorithms, based on the correlation between every weight and error function. (SWC)

This paper describes an attempt to combine the advantages of both example-based translation and stochastic translation methods in an attempt to develop a method for inferring symbolic transfer functions from a bilingual corpus. By formalizing the translation process and applying standard optimization techniques, a system can be developed that will identify grammatical categories and produce coherent transfer functions between languages. The validity of this approach is demonstrated in a prototype system that can learn transfer functions between English, French, and Urdu. Contains 13 references. (Author)

Currently, no machine translation (MT) system is capable of successfully imitating the behavior of a human translator, and there exists no formal description of what an MT system is supposed to do. The biggest problem in practice is disambiguation. However, various types of existing systems do help reduce language barriers, even if they are poor imitations of the human translator. Users of such systems determine the utility of the systems. Three groups of users are distinguished: big companies and institutions; professional and occasional translators; and monolingual users. Big MT systems are becoming more cost-effective. Smaller systems providing high-quality output for restricted tasks and domains and multilingual systems based on spoken dialogue are being developed. Whereas market developments show a trend toward more specialized and restricted systems, the trend in research is to widen the scope of machine translation, moving from sentence to text, focusing on disambiguation based on domain restrictions, and using non-traditional statistical methods. Work is needed to provide lexicological and terminological resources for East European languages, and to create large collections of real-life data, including monolingual and bilingual text corpora. (MSE)

Trends and developments in computer applications in Chinese language research are described, focusing on these areas: input of Chinese characters and Chinese corpus; automatic segmentation of Chinese written text in corpus; development of a grammar knowledge base for Chinese words to be used as a resource for text segmentation and corpus annotation; automatic part-of-speech tagging for the corpus; automatic phrase bracketing and syntactic annotation for the corpus; creation of specialized terminology data banks; and machine translation systems. Specific projects, cooperative efforts, and resulting resources are noted. (MSE)

As part of the development of a completely sub-symbolic machine translation system, a method for automatically identifying German compounds was developed. Given a parallel bilingual corpus, German compounds are identified along with their English word groupings by statistical processing alone. The underlying principles and the design process are described here. Design began with small-scale word-alignment experiments, using 2,543 English words and 1,898 German words that yielded unique lexical items in each language. A technique for decreasing reliance on one-to-one word correspondences was then applied, resulting in a distinct ability to capture relationships between compounds and non-compounded expressions. Statistical analysis of these relationships provides data on which to base machine translation operations. It is concluded that the method used is effective on identifying cross-language lexical fertility to establish translation units for re-combination within the example-based translation process. (MSE)

Discusses issues related to natural language processing, including theoretical developments; natural language understanding; tools and techniques; natural language text processing systems; abstracting; information extraction; information retrieval; interfaces; software; Internet, Web, and digital library applications; machine translation for multilingual retrieval; and evaluation. (Contains 136 references.) (LRW)

A review of literature on technology and second language (L2) instruction focused on the use of seven different technologies: television/videotapes/tape recordings; hand computers; computer conferencing; electronic mail (e-mail); multimedia and hypermedia; computer-assisted language learning; and machine translation. A secondary purpose of the review was to integrate the findings concerning each technology into a holistic framework. Seventeen articles reported here were found to provide a cross-section of technological tools applied to L2 learning. The findings of those articles are analyzed, with attention to both current realities and future promise, in three areas: summary of the major contributions of each technology to L2 learning; major issues or concerns related to implementation; and research methods used in the studies. Future research is proposed. In addition, the Internet and emerging technologies are considered briefly. Contains 50 references. (Author/MSE)

Discusses the relative merits of feature versus model based semantics for the interpretation of verb phrases in English, French, and German. The article concludes that the simplicity afforded by features is offset by the depth of analysis achieved with event models that are additionally able to support a sophisticated approach to machine translation. (19 references) (Author/CK)

Overviews the development of corpus linguistics, reviews the use of corpora in modern lexicography, and presents central issues in ongoing work aimed at broadening the scope of lexicographical use of corpus data. Focuses on how the field has developed in relation to the production of new monolingual English dictionaries by major British publishers. (50 references) (CK)

Natural language processing (NLP) is concerned with getting computers to do useful things with natural language. Major applications include machine translation, text generation, information retrieval, and natural language interfaces. Reviews important developments since 1987 that have led to advances in NLP; current NLP applications; and problems still awaiting solution. (Contains 110 references.) (PEN)

Describes the evaluation of Computronics Corporations'"Targumatic" (Hebrew into English), a PC-based machine translation system running under DOS, and shows how each problem and potential obstruction to learning can be overcome through discovery procedures using a set of tools and procedures called the "learning algorithm." (22 references) (Author/CK)

Discusses the development of phraseological machine translation based on the theory that in natural languages, phraseological units cannot always be reduced to words. Describes the phraseological translation process of computer software and machine dictionary systems that translate Russian into English and English into Russian. (JMV)

Examines recent trends in the fields of translation and interpretation, focusing on translation and interpretation theory and practice, language-specific challenges, computer-assisted translation, machine translation, subtitling, and translator and interpreter training. An annotated bibliography discusses seven important works in the field. (112 references) (MDM)

This paper examines the development and testing of a machine-translation program designed to translate English text into Portuguese. A comparison of machine- and human-translated texts, using Brazilian secondary school students as readers, found that they were equally understood, both in terms of main ideas and details. (six references) (MDM)

The use of a model for treatment of English modals in a machine translation system being developed in Europe is discussed. The Eurotra system, sponsored by the Commission of the European Communities, is designed to translate into the nine official languages of the European Economic Community. The report first examines modality, the different ways of expressing modality, and why it is necessary to have a semantic representation of modal expressions for each language. The Palmer/Davidsen-Nielsen grid for representation of English modal meanings is then outlined. The Eurotra organization and translation system are also described. Eurotra's stratified framework of analysis and synthesis at four levels (morphological, configurational, relational, and interface) and the process of transfer from the source to the target language are explained, and its model for classifying modality type is presented, giving attention to the method used for identifying modality types in a text. Danish and English examples are referred to, and structural and processing problems are noted. (MSE)

The investigation and use of machine translation as an adjunct to human translation at the Soviet Union's All-Union Translation Center, a centralized scientific and technical translation facility, is examined. The issues of error correction, the variety of translation types needed, and special lexical considerations are discussed. (MSE)

The machine translation system SYSTRAN was assessed for translation quality and system productivity. The test was carried out on translations from English to French dealing with food science and technology. Machine translations were compared to manual translations of the same texts. SYSTRAN was found to be a useful system of information processing, but resulting translations still required much human revision. An assessment of machine translation quality put average intelligibility at 78%. It was found that, while fewer than 5% of the words in the manual translations required revision, over one-quarter of all words in the machine translations had to be corrected. The methodology and results of the evaluation are described in detail, with much data presented in tabular form. (JB)

The rapid growth of the need for technical translations in recent years has led specialists to utilize computer technology to improve the efficiency and quality of translation. The two approaches considered were automatic translation and terminology banks. Since the results of fully automatic translation were considered unsatisfactory by various reports, significant efforts were made to construct automatized terminology banks. A critical comparison of the main existing banks suggests that these banks are the first step toward interactive translation systems. Exchange problems in networks of banks and terminological cooperation between international organizations will necessitate standardization of terms, symbols, and definitions. It is proposed that automatized terminology banks be designed from a different viewpoint by making greater use of well known techniques in computer systems and artificial intelligence, based on evaluation of present and future developmental issues in data base organization, translation research, interactive translation systems, and man-computer communication. (Author/CWM)