Mitamura, T. and Nyberg, E.: Controlled English for Knowledge-Based MT: Experience with the KANT System. In: Proceedings of TMI-95 (1995)  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... does not use a mathematical language to express abstract syntax and it seems not to be possible for the user to change the structure of what they call the knowledge base (our abstract syntax) Processing natural language within restricted domains makes GF related to the KANT translation system [10]. Kant Controlled English (KCE) put constraints on vocabulary, grammar and document structure in order to reduce the amount of ambiguity in the source text in the pre processing phase. The remaining ambiguities are resolved via interaction with the author in the authoring environment. The only ....

T. Mitamura and E. H. Nyberg. Controlled English for Knowledge-Based MT: Experience with the KANT system. In TMI, 1995.

....fast as you can speak . For details and reaction, see the home page of the American Translators Association at www.atanet.org. 26 continues on restricted domain MT, mostly involving technical documentation produced by large companies; an example is the KANT system developed for Caterpillar at CMU [96]. 2.1.3 Prospects for FAHQT Most work in combined human machine translation, including this thesis, is predicated on the assumption that unrestricted FAHQT is, if not impossible, at least a very long way o . Experts are almost unanimous about this [5, 57, 61, 70, 86, 95] The fundamental diculty ....

T. Mitamura and E. H. Nyberg. Controlled English for knowledge-based MT: experience with the kant system. In Proceedings of the 6th Conference on Theoretical and Methodological Issues in Machine Translation (TMI), pages 158-172, Leuven, Belgium, 1995.

....of categories as pronouns, quantifiers, participial forms and modal verbs. For grammar control, the main features of the system are the limitations imposed on ambiguous attachments both at the phrase and sentence level, coordinated structures, complex sentences and long distance dependencies [Mitamura 1995, Mitamura 1999, Nyberg and Mitamura 1996] The PACE system, an instantiation of a more general checker program, reported by [Douglas and Hurst 1996] detects errors at the character level (punctuation) illegal words, grammar errors such as agreement, and insertion and deletion of constituents. ....

Mitamura, T. Controlled English for Knowledge-Based MT: Experience with the KANT system. In Proceedings of 6th International Conference on Theoretical aaand Methodological Issues in Machine Translation, 1995. Available as http://www.lti.cs.cmu.edu/Research/PostScript/tmi-95-final.ps.

.... languages restrict the vocabulary and grammatical structure of full natural language in order to allow more accurate machine translation [9] Applications include interfaces to databases and knowledge bases [14, 15] and to theorem provers [4] multi lingual translation in technical domains [9, 10, 11, 12], and as specification languages for logic programs [2, 3] This paper reports the use of a controlled language to the problems of text to logic translation and knowledge fusion. We describe the techniques used to translate structured text into logical assertions in CycL, the representation ....

....around 2 of a typical COA text. In common with other controlled natural languages, the number of alternative parses of a sentence is greatly reduced. Typically, only one parse is possible. This is clearly very important in simplifying the interpretation task as has been noted by Mitamura [10] who reports an average reduction from 27 to 1 parses of a test suite of sentences through the use of structure constraints. The COA grammar does not restrict the nesting of COA phrases as many non terminals are recursively defined. Therefore the depth of parse trees is not limited. However, ....

[Article contains additional citation context not shown here]

Mitamura T. and Nyberg E. Controlled English for Knowledge-Based MT: Experience with the KANT System. Proceedings of TMI-95, 1995.

....paragraph level) As a result, the current version of KANT lacks a representational level which supports resolution of inter sentential phenomena (such as anaphora, ellipsis, definite descriptions, etc. While this design decision is justified by the lack of such phenomena in typical KANT domains (Mitamura and Nyberg, 1995; Nyberg and Mitamura, 1996) it does limit the broader applicability of KANT to less restrictive domains (as exemplified by the workshop text) In order to generate interlinguas for the sentences in the workshop text, we made the following extensions to the KANT analyzer knowledge sources: ffl ....

....We added new compounds, more specific meanings for existing words, and some new verbs and verb alternations (see Section 5.1) ffl Extensions to the Interpreter rules, to handle new semantic roles for this domain. ffl Manual disambiguation of the text. Interactive disambiguation and SGML tagging (Mitamura and Nyberg, 1995) were used to disambiguate each sentence, so that the single preferred reading was produced. It should be noted that no procedural changes had to be made. The current results of adapting the KANT analyzer for the workshop text are available on line (see the final page of this paper for a pointer ....

Mitamura, T. and E. Nyberg (1995). "Controlled English for Knowledge-Based MT: Experience with the KANT System", " Proceedings of TMI-95.

....[10 bullet easy maintenance bullet ] 11 bullet multiple upgrading possibilities. bullet ] 5 See e.g. Zajac, this volume. 6 To stay within the field of controlled language applications: SGML tags inside sentences are also used in the KANT ClearCheck developments (see Mitamura and Nyberg 1995); see also Schachtl, this volume. The grammar then contains rules interpreting these tags (and possible variants: title , ttl , title1 , title2 , heading , heading1 , heading2 , head , subhead , and so on) as valid sentential material. If the tags are present, this information is ....

Mitamura, T. and E. H. Nyberg, 3rd (1995) - Controlled English for KnowledgeBased MT: Experiences with the KANT System. In Proceedings of the Sixth International Conference on Theoretical and Methodological Issues in Machine Translation (TMI95), Vol. 1, 158-172.

....and the generation phases become more complicated. This approach stresses the fact that meaning is language independent, and languages are encoding systems used by humans to present their view of world to each other. Among the systems conforming to the interlingua design are Ultra [8] Kant [26, 21], and Microcosmos [3, 18] The methodology that is utilized in this work is the interlingua approach [10, 22, 26, 23] It separates the analysis task from the generation task using an artificial meaning representation. Generally, the analysis step firstly extracts the syntactic structures of the ....

T. Mitamura and E. Nyberg. Controlled english for knowledge-based mt: Experience with the kant system. In Proceedings of 6th International Conference on Theoretical and Methodological Issues in Machine Translation, Leuven, Belgium, July 5-7, 1995. REFERENCES 107

....by writing either The customer enters a card that carries a numeric code. or The customer enters a card and a numeric code. Controlled, or simplified, English has been used since many years for technical documentation [Wojcik et al. 90; Adriaens Schreurs 92] for machine translation [Mitamura Nyberg 95] and as data base query language [Androutsopoulos 95] Pulman and Rayner are suggesting a computer processable controlled language that could be used for various purposes ranging from structured documentation over access to information to the control of devices [Pulman Rayner 94] However, ....

T. Mitamura, E. H. Nyberg, Controlled English for Knowledge-Based MT: Experience with the KANT System, in: Proc. of the 6th Int. Conf. on Theoretical and Methodological Issues in Machine Translation (TMI'95), pp. 158-172, 1995

....environment, Multra Developer s Tool, MDT [Beskow 1992] has been built. In this environment each component can be tested independently. In specific, MDT provides rich tracing facilities. 1 In implementing this approach, we find much inspiration in the achievements made in the Kant system [Mitamura Nyberg 1995] where such a strategy has been applied successfully. An example of translation in Multra Input: Satt upp vaxelladan i universalstativ: Put the gearbox on universal stand. Result of the parsing process: PHR.CAT : CL TYPE : IMP REG : V1.LEM : SATTA.VB PRED.LEM : SATTA.VB UPP.PL] PRED : ....

Mitamura, Teruko & Eric H. Nyberg 1995. Controlled English for Knowledge-Based MT: Experience with the KANT System In: Proceedings of the 6th International Conference on Theoretical and Methodological Issues in Machine Translation, Leuven, Belgium.

....and processability [Allen 91, Pulman 94] Controlled or simplified English is not a new idea. It has been used for quite some time for technical documentation [Wojcik et al. 90, Adriaens Schreurs 92, AECMA 95] and lately both for knowledge based machine translation in the KANT system [Mitamura Nyberg 95] and as data base query language [Androutsopoulos 95] Additionally, a general computer processable controlled language has been suggested that could be used for various purposes ranging from structured documentation over access to information to the control of devices 4 [Pulman Rayner 94] ....

T. Mitamura, E. H. Nyberg, 3rd, Controlled English for Knowledge-Based MT: Experience with the KANT System, Center for Machine Translation, Carnegie Mellon University, Pittsburgh, 1995

....can be understood by all parties concerned. Controlled or simplified English is not a new idea. It has been used for quite some time for technical documentation [Wojcik et al. 90, Adriaens Schreurs 92, AECMA 95] and recently both for knowledge based machine translation in the KANT system [Mitamura Nyberg 95] and as data base query language [Androutsopoulos 95] Additionally, a general computer processable controlled language has been suggested that could be used for various purposes ranging from structured documentation over access to information to the control of devices [Pulman Rayner 94] ....

....[Kamp Reyle 93] Some English verbs are constructed with the help of a particle, e.g. turn on. Particles usually overlap with prepositions leading to ambiguous readings for the same sentence. A strategy to remedy this ambiguous reading is to choose a one word verb, e.g. start instead of turn on [Mitamura Nyberg 95] Modal verbs are banished from ACE because they can be used to express a vague degree of certainty about facts. The scale of vagueness extends from greatest uncertainty (might) to the greatest certainty (must) Modal verbs in a specification text are an indicator that the writer does not have a ....

T. Mitamura, E. H. Nyberg, 3rd, Controlled English for KnowledgeBased MT: Experience with the KANT System, Center for Machine Translation, Carnegie Mellon University, Pittsburgh, 1995

No context found.

Mitamura, T. and Nyberg, E.: Controlled English for Knowledge-Based MT: Experience with the KANT System. In: Proceedings of TMI-95 (1995)

No context found.

Mitamura, T. and Nyberg, E.: Controlled English for Knowledge-Based MT: Experience with the KANT System. In: Proceedings of TMI-95 (1995)

....generate knowledge representation languages as part of a knowledge extraction system. The inputs to the system are short texts describing critical infrastructures (e.g. financial markets, electrical power transmission) Where necessary, the texts are rewritten to conform to a controlled language [4]. Then each text is analyzed to produce an interlingua representation (IR) for its sentences. The interlinguas are merged together into a single representation for the paragraph. The merged representation is then generated into different output languages. In this study, the outputs were not ....

....Analysis The general goals of Controlled Language (CL) are to achieve consistent au thoring of source texts and to encourage clear and direct writing. Controlled language is also used to improve the quality of translation output through the use of short, concise and unambiguous sentences [4, 2, 8]. Although the goal of this study was to generate knowledge representations, rather than translated texts, this task also requires an accurate, unambiguous meaning representation. 2.1 Controlled Language in KANT In KANT Controlled English (KCE) we explicitly encode domain lexemes along with their ....

[Article contains additional citation context not shown here]

Mitamura, T. and Nyberg, E.: Controlled English for Knowledge-Based MT: Expe- rience with the KANT System. In: Proceedings of TMI-95 (1995)

....2 through 6, we describe the core servers and tools in the KANTOO architecture 1 . In Section 7 we conclude with some remarks about ongoing and future KANTOO research topics. 1 Space limitations preclude a discussion of a) the Controlled Language Checker, which has been discussed at length in [3], and b) the Batch Translator, which is a simple piece of driver code that uses the KANTOO servers to translate entire documents. 2 2. Analyzer Module The Analyzer module performs tokenization, morphological processing, lexical lookup, syntactic parsing with a unification grammar, and semantic ....

Mitamura, Teruko & Eric Nyberg: 1995, `Controlled English for Knowledge-Based MT: Experience with the KANT System', in Proceedings of TMI-95, Leuven.

No context found.

Mitamura, T., and Nyberg, E.: Controlled English for knowledge-based MT: Experience with the KANT system. In: Proceedings of TMI-95 (1995)

.... system (Knowledge based, Accurate Natural language Translation) is a set of software tools for automatic and interactive analysis of source text and generation of target text (Mitamura, et al. 1991) It has been primarily targeted towards the translation of technical text in controlled subdomains (Mitamura and Nyberg, 1995). Initially an outgrowth of research ideas following the completion of the KBMT 89 system at CMU (Goodman and Nirenburg, 1990) KANT has been scaled up for multilingual document production in an industrial setting. In this paper, we discuss the facets of KANT which are of potential interest to the ....

Mitamura, T. and E. Nyberg (1995). "Controlled English for Knowledge-based MT: Experience with the KANT System," Proceedings of TMI-95, Leuven, Belgium.

No context found.

Mitamura and Nyberg (1995). "Controlled English for Knowledge-Based MT: Experience with the KANT System", Proceedings of TMI-95.

....in the KANT machine translation system. 1 Introduction Research and development has shown that knowledge based machine translation, combined with the use of controlled language in well defined technical domains, can achieve very high accuracy in translation (Nyberg Mitamura 1992; Mitamura Nyberg 1995; Kamprath et al. 1998) Detailed knowledge bases often include semantic preferences for disambiguating structural attachments (Baker et al. 1994) However, the efficacy of knowledge based MT has often been questioned because of the significant time and effort required to build semantic knowledge ....

....used in KANT for resolution of structural (attachment) ambiguity. During interactive grammar checking, KANT takes the following steps to analyze each sentence in the document: 1 For a full discussion of the types of ambiguity in technical text and how they are handled by the KANT system, see (Mitamura Nyberg 1995). 1. Morphological analysis is performed, and the set of possible lexical entries for each input token is retrieved; 2. A unification grammar is used to produce the legal set of grammatical functional structures (f structures) for the input tokens; 3. If there is more than one possible ....

Mitamura, Teruko & Eric Nyberg: 1995, `Controlled English for Knowledge-Based MT: Experience with the KANT System', in Proceedings of TMI-95, Leuven.

....of the pilot phase, the grammar was refined and the usage examples in the authoring interface were improved. 4.2. CTE Maintenance Support for the ongoing use of CTE in day to day production required the following activities: 3 More details regarding design issues and examples are discussed in (Mitamura and Nyberg, 1995; Nyberg and Mitamura, 1996) ffl Problem Reporting. CAT developed in house problem reporting software and a process for author requests for new terms and grammar updates. CGI and CMU provided a problem report database, built using the GNATS software (Osier, 1993) which is used for problem ....

....machine translation, so we extended the design of the grammar checker to include an interactive disambiguation module. When ambiguous terms are identified, the system asks the author to specify the intended meaning, which is then preserved in the input text using an unobtrusive SGML marker (Mitamura and Nyberg, 1995). ffl Incremental IE Updates. When reusing one IE in several different documents, it is often necessary to make only minor incremental revisions. However, the original CTE system required that the entire IE be re checked sentence by sentence if a change was made to a previously checked i.e. To ....

Mitamura, T. and E. Nyberg (1995). "Controlled English for Knowledge-Based MT: Experience with the KANT System", " Proceedings of TMI-95.

No context found.

Teruko Mitamura and Eric Nyberg. 1995. Controlled English for Knowledge Based MT: Experience with the KANT System. In Proceedings of Sixth International Conference on Theoretical and Methodological Issues in Machine Translation, Leuven, Belgium, pp.158--172.

No context found.

Mitamura, T. and Nyberg, E. H.: Controlled English for Knowledge-Based MT: Experience with the KANT System, Center for Machine Translation, Carnegie Mellon University, Pittsburgh, 1995

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC