| { | |
| "paper_id": "C90-1007", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T12:36:05.298399Z" | |
| }, | |
| "title": "THALES: a Software Package for Plane Geometry Constructions with a Natural Language Interface", | |
| "authors": [ | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Fabricz", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "JATE University Szeged", | |
| "location": { | |
| "addrLine": "Egyetem u. 2", | |
| "postCode": "H-6722", | |
| "settlement": "Hungary" | |
| } | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "THALES is a software package for plane geometry constructions, supplied with a natural language interlace. Using THALES requires no knowledge of a programming language. The interface is capable of processing practically all kinds of instructions within the subset of plane geometry English. The \"static semantic\" module has been generated on the basis of a high-level attribute specification. Transportability, modifiability and generality-the key issues of natural language interface design-are investigated in the project note. The notion of specifiability is introduced to replace the three features mentioned above.", | |
| "pdf_parse": { | |
| "paper_id": "C90-1007", | |
| "_pdf_hash": "", | |
| "abstract": [ | |
| { | |
| "text": "THALES is a software package for plane geometry constructions, supplied with a natural language interlace. Using THALES requires no knowledge of a programming language. The interface is capable of processing practically all kinds of instructions within the subset of plane geometry English. The \"static semantic\" module has been generated on the basis of a high-level attribute specification. Transportability, modifiability and generality-the key issues of natural language interface design-are investigated in the project note. The notion of specifiability is introduced to replace the three features mentioned above.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Abstract", | |
| "sec_num": null | |
| } | |
| ], | |
| "body_text": [ | |
| { | |
| "text": "Natural language interfaces represent one of the most common applications of natural language processing. In the eighties, not only a considerable increase in natural language interface refinement has been achieved, but also methods for design and evaluation have been worked out [Mart 83, Schr 881. One might think that a natural language interface of the nineties would be properly described in terms ol throe parameters, viz., (i) transportability (ii) modifiability by the user Off) generality Our experience with THALES shows that none of these features is attainable in the near future. Rather, natural language interfaces based on welldefined subsets of languages and supplemented with possibly lull semantics appear to be real candidates for applications in the following years. Requirements for transportability and generality run counter to the need to supply the interface with as full semantics as possible. Modifiability by the user can be maintained at a price that is hardly worth paying.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction*", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "If one lets the user tailor the interface to his/her needs, that natural language interlace should provide this possibility by relaxing syntax or semantics. We do not consider this a real ahernative. Neither does modifiability in the form of substitution of one expression by another on the basis of their semantic identity seem to be a realization of what the term \"modifiability by the user\" suggests (Cf. [JAKE 88]).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction*", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "However, there appear to be different parts of natural language interlace design that may be reused in other applications. We shrill refer to this possibility by the term \"specifiability\" --a feature that merits thorough consideration.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Introduction*", | |
| "sec_num": "1." | |
| }, | |
| { | |
| "text": "THALES is a program package fbr plane geometry constructions. The constructions are carried out as series of instructions lormulated in English. Through a menu system, it is possible to save constructions and parts of constructions lot some later application, the constructions may be inserted into other applications. The instructions can be stored and edited as normal ASCII files. A number of examples covering the majority of tasks related to plane geometry are supplied to help the teacher or student in teaching or learning plane geometry. Manipulating objects during constructions is also supported without the risk of running into ambiguities in the course of reference resolution.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Overview of THALES", | |
| "sec_num": "2." | |
| }, | |
| { | |
| "text": "The natural language interface of THALES is briefly characterized as follows.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The natural language interface of THALES", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Linguistic processing is based on an attribute grammar description. From this specification, the PROF-LP language processor [Gyim 88 ] generates the parser and the evaluator. The lexicon consists of a fixed number of items covering the sublanguage of plane geometry. The parser contains a lexical and a syntactic analyzer. The former does morphological analysis as well in order to keep the size of the vocabulary at a minimum. The syntactic module represents a top-down, basically LL(1) algorithm augmented by attributes for initiating semantic processing. Semantic interpretation is based on a recta-level object description that provides a clue to the implementation of relations in the lorm of different steps in producing geometrical constructions [Ale 89].", | |
| "cite_spans": [ | |
| { | |
| "start": 124, | |
| "end": 132, | |
| "text": "[Gyim 88", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The natural language interface of THALES", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "Intersentential reference is resolved by building a symbol table. The range of syntactic and semantic coverage can be illustrated by the examples below. ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "The natural language interface of THALES", | |
| "sec_num": "3." | |
| }, | |
| { | |
| "text": "The experience gained from designing and implementing THALES has led to the following conclusions.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Generating natural language interfaces", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "A natural language interface cstablishes connection between the user and some other language or set of user-defined procedures. In constructing a natural language interface, the primary aim is to provide a means for the translation of natural language sentences into sentences of that other language or into procedure calls. The latter define the semantics of execution lor the input sentence(s).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "I. Evaluation of natural language interfaces", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "When the value of a natural language interface is being assessed, transportability, modifiability and generality seem to be false starting points. These features would make a natural language interface truly valuable, were it not tot the fact that they cannot be taken prima facie. A transportable natural language interface means some kind of over-generalization. For a natural language interface to be transportable, it is necessary that we have a sublanguage with too wide a coverage to carry specific infmrnation encoded in its semantics. That is why transportability is achieved only in highly identical domains (characteristically, interfaces to databases). A natural language interface can bc modified by the user according to his/her specific purposes exceptionally in the lexicon. Hence, the semantics is prewired, the syntax is fixed, and what you gain is hardly more than a rephtcement of one or two mnemonics with some other character strings that you might prefer. The notion of generality of a natural language interface is too vague to deserve special attention. It might refer to syntactic and/or semantic coverage, it might imply the existence of a fairly large word stock, but, we should think, it would hardly ever be takcn to mean \"general\" in the sense that it could be applied in different domains.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "I. Evaluation of natural language interfaces", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "A basic question in designing natural language interfaces of the THALES type is that of assessing which parts can and which cannot be generated. An answer to this question would throw light on the problem of transportability, modifiability and generality. We argue that those parts of a natural language interlace which can be generated from some high level specification appeal\" to closely satisfy the demand lor transportability, modifiability and generality. It seems to be the case that the lexical, syntactic and static semantic analyzers are suitable for generation on the basis of an attribute grammar description.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Specifiability of natural language interface architecture", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "It should be noted that the lexical analyzer also contains the morphological module. The syntactic parser processes a CF grammar. Dynamic semantics is also suitable lot analysis by an attribute grammar description, but there are too many application-specific elements whose generation does not seem to be a realistic goal for the time being.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Specifiability of natural language interface architecture", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "Of crucial importance is the notion of static (compile time) semantics in the case of natural language interfaces like TIJALES. In a conventional programming language, we define its static semantics as the set of context-sensitive properties that can bc evaluated, checked or calculated If one compares static semantic analysis in the case ot' natural language interfaces like THALES, it appears that basic types correspond to objects with a hierarchic structure. E.g. a triangle is represented as having vertices, bisectors, altitudes etc. Again, e.g. altitudes have endpoints, a midpoint, length etc. Endpoints, in turn, have coordinates expressed as numbers. Such hierarchic structure can be ascribed to other natural language interfaces as well. A specification language [Ale 8911 is used for type definition. In the course of specification, other data can also be incorporated (adjectival, prepositional co-occurrence restrictions etc). On the basis of this specification it is possible to generate procedures for symbol table manipulation. The symbol table can then serve as a tool for deciding type compatibility. For example, one can refcr to the midpoint of an altitude due to the fact that an altitude is of segment type, and segments do have a midpoint.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Specifiability of natural language interface architecture", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "In the differentiation of defining and applied occurrences, determiners play a crucial role. While in conventional programming languages identification is resolved by a relatively simple algorithm, this problem in THALES-like natural language interfaces is much more complicated. Below we give a sample of the identification methodology used in THALES. This methodology can be extended on the specificational level or, vice versa, the application of some of the methods can be prohibited. The abbreviations ID, OBJ, and ADJ stand lot identitier (a unique name of an object, denoted by '-' in THALES ), object (the name of an object, possibly a part of another object), and adjective, resp. OBJ 1D -an object with a unique name to be lbund unambiguously, cf.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Specifiability of natural language interface architecture", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "Draw the radius of circle ~C. OBJ -search for the first occurrence of the object with the given type by tracking the symbol table backward, cf.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Specifiability of natural language interface architecture", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "Pick a point in the circle. ADJ OBJ -a search for the first occurrence of that type of object specified by the adjective, cf.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Specifiability of natural language interface architecture", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "Draw a circle inside the acute triangle.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Specifiability of natural language interface architecture", | |
| "sec_num": "4.2." | |
| }, | |
| { | |
| "text": "From the experience with THALES we can conclude that the notion of static semantics can be successfully applied to the group of natural language interfaces represented by THALES. The static analyzer in THALES appears to satisfy the requirement for transportability, modifiability and generality, although in a modified and more realistic sense. That is, we do not think that there is a one-to-one correspondence between a generated static analyzer and the requirements. Rather, the static analyzer can be specified in a high-level language, and that specification can meet the transportability, modifiability and generality requirements.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions", | |
| "sec_num": "4.3." | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "The authors gratefully acknowledge the financial support of Cogito Ltd., Philadelphia.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "acknowledgement", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": { | |
| "BIBREF0": { | |
| "ref_id": "b0", | |
| "title": "CONSTRUCTOR: A Natural Language Interface Based on Attribute Grammar", | |
| "authors": [ | |
| { | |
| "first": "Z", | |
| "middle": [], | |
| "last": "Alcxin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "K", | |
| "middle": [], | |
| "last": "Ffibricz", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Gyim6thy", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Horv~ith", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": 1989, | |
| "venue": "Proceedings of the First Finnish-Hungarian Workshop on Programming Languages and Software Tools", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "135--145", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Alcxin, Z., Ffibricz, K., Gyim6thy, T., Horv~ith, T. CONSTRUCTOR: A Natural Lan- guage Interface Based on Attribute Grammar. In: T. Gyim6thy (ed.) Proceedings of the First Finnish-Hungarian Workshop on Programming Languages and Software Tools, Szeged, 1989, pp. 135-145.", | |
| "links": null | |
| }, | |
| "BIBREF1": { | |
| "ref_id": "b1", | |
| "title": "Natural Language Querying of Historical Databases", | |
| "authors": [ | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Clifford", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "Computational Linguistics", | |
| "volume": "14", | |
| "issue": "4", | |
| "pages": "10--34", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Clifford, J. Natural Language Querying of Historical Databases. Computational Lin- guistics 14 (4), 10-34.", | |
| "links": null | |
| }, | |
| "BIBREF2": { | |
| "ref_id": "b2", | |
| "title": "Incremental Algorithms in PROF-LP", | |
| "authors": [ | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Gyim6thy", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "T", | |
| "middle": [], | |
| "last": "Horvfith", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Kocsis", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "J", | |
| "middle": [], | |
| "last": "Toczki", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "Lecture Notes in Computer Science", | |
| "volume": "371", | |
| "issue": "", | |
| "pages": "93--103", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "88] Gyim6thy, T., Horvfith, T., Kocsis, F., Toczki, J. Incremental Algorithms in PROF-LP. Lecture Notes in Computer Science, 371. pp. 93-103.", | |
| "links": null | |
| }, | |
| "BIBREF3": { | |
| "ref_id": "b3", | |
| "title": "The Application-Independent Natural Language User Interface", | |
| "authors": [], | |
| "year": 1988, | |
| "venue": "", | |
| "volume": "", | |
| "issue": "", | |
| "pages": "", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "[JAKE 8811 JAKE. The Application-Independent Natural Language User Interface. English Knowledge Systems, Inc., Scotts Valley, Cali- lomia, 1988.", | |
| "links": null | |
| }, | |
| "BIBREF4": { | |
| "ref_id": "b4", | |
| "title": "Transportability and Generality in a Natural Language Interface System", | |
| "authors": [ | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Martin", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "P", | |
| "middle": [], | |
| "last": "Appelt", | |
| "suffix": "" | |
| }, | |
| { | |
| "first": "F", | |
| "middle": [], | |
| "last": "Pereira", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "Proceedings of HCAI-83", | |
| "volume": "1", | |
| "issue": "", | |
| "pages": "573--581", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "Martin, P., Appelt, P., Pereira, F. Trans- portability and Generality in a Natural Lan- guage Interface System. Proceedings of HCAI- 83, 1, pp. 573-581.", | |
| "links": null | |
| }, | |
| "BIBREF5": { | |
| "ref_id": "b5", | |
| "title": "Evaluating User Utterances in Natural Language Interfaces to Databases", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Schrtider", | |
| "suffix": "" | |
| } | |
| ], | |
| "year": null, | |
| "venue": "Computers and Artificial Intelligence", | |
| "volume": "7", | |
| "issue": "4", | |
| "pages": "317--337", | |
| "other_ids": {}, | |
| "num": null, | |
| "urls": [], | |
| "raw_text": "M. Schrtider. Evaluating User Utterances in Natural Language Interfaces to Databases. Computers and Artificial Intelligence, 7 (4), pp. 317-337.", | |
| "links": null | |
| } | |
| }, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "text": "during compile time. The most important properties el' this kind are a) the dil]erentiation between defining and applied occm'rences of variables b) the assessment of the scope of defining occurrences c) the problem of identification: finding the defining occurrences that match the applied occurrences d) the question of type compatibility Conventional compilers contain a symbol table that will serve as a basis for the investigation of the questions raised above. Elements of the symbol table are built from objects of the types recognized by the language, supplemented with auxiliary information. During compilation the symbol table undergoes continuous modification, problems related to static semantics are solved on the basis of the actual state of the symbol table.", | |
| "type_str": "figure", | |
| "uris": null, | |
| "num": null | |
| } | |
| } | |
| } | |
| } |