{ "paper_id": "A88-1004", "header": { "generated_with": "S2ORC 1.0.0", "date_generated": "2023-01-19T02:04:01.515559Z" }, "title": "THE SEMSYN GENERATION SYSTEM: INGREDIENTS, APPLICATIONS, PROSPECTS", "authors": [ { "first": "Dietmar", "middle": [], "last": "Rssner", "suffix": "", "affiliation": { "laboratory": "", "institution": "Universit\u00a3t Stuttgart", "location": { "addrLine": "Herdweg 51", "postCode": "D-7000", "settlement": "Stuttgart 1", "country": "West Germany" } }, "email": "" }, { "first": "Projekt", "middle": [], "last": "Semsyn", "suffix": "", "affiliation": { "laboratory": "", "institution": "Universit\u00a3t Stuttgart", "location": { "addrLine": "Herdweg 51", "postCode": "D-7000", "settlement": "Stuttgart 1", "country": "West Germany" } }, "email": "semsyn@ifistg.uucp" } ], "year": "", "venue": null, "identifiers": {}, "abstract": "We report about the current status of the SEM-SYN generation system. This system-initially implemented within a Japanese to German MT project-has been applied to a variety of generation tasks both within MT and text generation. We will work out how these applications enhanced the system's capacities. In addition to the paper we will give a demo of both the German and a recently implemented English version of the system.", "pdf_parse": { "paper_id": "A88-1004", "_pdf_hash": "", "abstract": [ { "text": "We report about the current status of the SEM-SYN generation system. This system-initially implemented within a Japanese to German MT project-has been applied to a variety of generation tasks both within MT and text generation. We will work out how these applications enhanced the system's capacities. In addition to the paper we will give a demo of both the German and a recently implemented English version of the system.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Abstract", "sec_num": null } ], "body_text": [ { "text": "The SEMSYN project began in 1983 with an MT application as starting point: We had to investigate the possibility of using semantic structures derived from Japanese as input to a generator for German. The semantic structures were produced from Japanese titles of papers in information technology by the parser of Fujitsu's ATLAS/II-system, the German generator had to be designed and implemented by SEMSYN. A first prototype was presented at the end of 1984, further enhanced versions of this Japanese/German system were demonstrated at various meetings, e.g. ].", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The starting point", "sec_num": "1" }, { "text": "When we designed and implemented the initial SEMSYN generator we tried to base this work on guidelines like the following [Rfsner 86b]:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Design guidelines", "sec_num": "2" }, { "text": "\u2022 The implemented system should not be confined by its/irst application.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Design guidelines", "sec_num": "2" }, { "text": "Since we had only little influence on the format of the output of our partner's parser for Japanese we designed a more general frame language that served as input to our system.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Design guidelines", "sec_num": "2" }, { "text": "The generator should be easily extensible and transportable to other applications. This gave strong motivation for choosing an object-oriented implementation technique.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Design guidelines", "sec_num": "2" }, { "text": "The generator should be integrated in an environment of software tools supporting ~every day\" work (e.g. testing, debuggi~zg, experimentation) with the system. This led to the implementation of a whole system of utilities: e. g. menu-based tools for lexicon update and maintenance or an interactive editor for semantic structures.", "cite_spans": [ { "start": 99, "end": 142, "text": "(e.g. testing, debuggi~zg, experimentation)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Design guidelines", "sec_num": "2" }, { "text": "The SEMSYN generator is organized into two major modules:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The basic machinery", "sec_num": "3" }, { "text": "\u2022 the generator kernel or 'realization component' and", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The basic machinery", "sec_num": "3" }, { "text": "\u2022 the front end generator or 'morpho/syntactic component'.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The basic machinery", "sec_num": "3" }, { "text": "We will have a closer look at the operation of these modules now.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The basic machinery", "sec_num": "3" }, { "text": "The generator kernel starts from a semantic representation, i.e. a 'message' in the sense of [McDonald et al. 87] . Its task is to 'realize' the message, i.e. to decide how its content may be expressed in natural language:", "cite_spans": [ { "start": 93, "end": 113, "text": "[McDonald et al. 87]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "The generator kernel", "sec_num": "3.1" }, { "text": "\u2022 What is the adequate syntactic form for the utterance as a whole?", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The generator kernel", "sec_num": "3.1" }, { "text": "\u2022 How should the subparts of the conceptual representation be realized and integrated into the utterance?", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The generator kernel", "sec_num": "3.1" }, { "text": "\u2022 What are appropriate lexicalizations -as lexemes or whole phrasal structures of the target language -for the elements of the message?", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The generator kernel", "sec_num": "3.1" }, { "text": "The output of the generator kernel is a functional grammatical structure. This linguistic representation fully specifies the intended utterance:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The linguistic representation", "sec_num": "3.2" }, { "text": "\u2022 the syntactic category of the whole utterance and the grammatical functions and syntactic categories of all subparts,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The linguistic representation", "sec_num": "3.2" }, { "text": "\u2022 the syntactic features of the head of each syntactic entity,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The linguistic representation", "sec_num": "3.2" }, { "text": "\u2022 the lexemes or special lexical itemsmarked with category information like :*PN for proper names or :*NC for noun compounds -to be used.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The linguistic representation", "sec_num": "3.2" }, { "text": "The functional grammatical structures produced by the generator kernel are input to the front end generator. This module has to execute all syntactic and morphological processes that are necessary to produce the corresponding surface string. This involves:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The front end generator", "sec_num": "3.3" }, { "text": "\u2022 linearization, i.e. constituent ordering,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The front end generator", "sec_num": "3.3" }, { "text": "\u2022 agreement handling,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The front end generator", "sec_num": "3.3" }, { "text": "\u2022 inflection.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The front end generator", "sec_num": "3.3" }, { "text": "The need for an explicit linguistic representation of the intended utterance and a separate final processing step is especially obvious for highly inflectional languages with a rich repertoire of agreement phenomena (e.g. French, German).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "The front end generator", "sec_num": "3.3" }, { "text": "SEMSYN's generator kernel expects its input in a frame notation. Although there are minor variations between the different applications the basic format is fixed: frame structures consisting of a 'semantic symbol' as name and named roles or slots with -recursivelyframe structures as fillers.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Frame structures as semantic representation", "sec_num": "3.4.1" }, { "text": "An example of a case frame: Here the toplevel frame structure contains the semantic symbol 'GENERATE' and has two filled roles :AGENT and :OBJECT as further information.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Frame structures as semantic representation", "sec_num": "3.4.1" }, { "text": "When the generator kernel realizes this case frame as a clause in active voice this results in the following functional grammatical structure: This yields the following German sentence~:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A realization result", "sec_num": "3.4.2" }, { "text": "\"Das SEMSYN-Projekt generiert deutsche Sprache.\" (The SEMSYN project generates German language.)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A realization result", "sec_num": "3.4.2" }, { "text": "Object-oriented implementation of realization knowledge", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "The main features of the object-oriented paradigm that we exploited for the implemen-I English glosses added as convenience for the reader. tation of realization knowledge in the generator kernel are", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 hierarchy as organisation principle for the knowledge base and", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 message passing between objects as primary control structure.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "The specialization hierarchy used is rooted in a general class that defines the basic methods for realization (KBS-Schema). On the next level are general classes for", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 case frames (CASE-Schema)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 concepts (CONCEPT-Schema)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 relations (RELATION-Schema).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "These classes differ with respect to the possible realizations of their instances:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 concept-schemata allow only realizations as noun groups", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 case-schemata allow for various clausal forms (active, passive, topicalized) as well as nominalized forms", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "\u2022 subclasses of relation-schema incorporate knowledge about realization possibilities for (more complex) semantic relations like the relation between :MEANS and :PUR-POSE, :REASON and :RESULT etc.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "3.5", "sec_num": null }, { "text": "The semantic representation of a summary of Macbeth may contain the following frame structure: 4 Other applications of the system", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A relation-schema", "sec_num": "3.5.1" }, { "text": "(", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A relation-schema", "sec_num": "3.5.1" }, { "text": "In the meantime improved and extended versions of the SEMSYN generation system have been applied to quite a variety of input structures and generation tasks:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A relation-schema", "sec_num": "3.5.1" }, { "text": "\u2022 machine translation applications:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A relation-schema", "sec_num": "3.5.1" }, { "text": "- Although the basic design of the generator [R5sner 86b] proved to be flexible enough and could remain untouched each of these applications has led to additional features of the whole system.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A relation-schema", "sec_num": "3.5.1" }, { "text": "MT applications", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "4.1", "sec_num": null }, { "text": "In the first application of the system we started from semantic representations derived from titles of Japanese papers in the field of information technology. Titles are in most cases noun groups. In order to generate German equivalents we had to provide the prototype primarily with knowledge about German noun group structures. On the other hand, for many of these semantic structures clausal forms were possible as well. We therefore provided the system with \"stylistic\" switches that allowed the alternative generation of clauses from case frames as well.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Title translation", "sec_num": "4.1.1" }, { "text": "The sample of semantic structures in this experiment was taken from doctor/patient communication. The semantic structures produced by CMU's parsers for Japanese and English are basically case frames, but include syntactic information as well (e.g. about :MOOD or :TIME). The fragment of German generable by the SEMSYN system was extended by yes/no-questions and imperatives. This experiment was interesting under various aspects:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\u2022 The semantic representation used is based on systemic grammar; since the classes used are already hierarchically structured it was relatively easy to implement them as a FLAVOR hierarchy of realization classes.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\u2022 SEMTEX starts from mere labor market data, extract a list of semantic representations from them as \"text plan\" and then converts this list into texts like the following:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\"Die Zahl der Arbeitslosen in der Bundesrepublik Deutschland ist im Dezember spllrbar angestiegen. Sic hat yon 2210700 auf 2347100 zugenommen. Die Arbeitslosenquote betrug Ende Dezember 9.4 Prozent. Sic hatte sich Ende Dezember des letzten Jahres auf 9.3 Prozent belaufen. Der DGB hat erkllirt, er sehe in der Vergriil3erung der Arbeitslosenzahl ein negatives Zeichen.\"", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "The main concern in implementing SEM-TEX has been to provide the SEMSYN generator with mechanisms that keep track of previous generation decisions thus creating a representation of the textual context built up by the already uttered sentences. This context is used:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\u2022 to avoid repetition in wording,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\u2022 to deliberately elide information still valid (e.g. about the time period concerned),", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\u2022 to decide on pronominalisation and other types of reference.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "In addition a representation of the temporal context is used", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\u2022 to dynamically determine grammatical tense and", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "\u2022 to produce appropriate natural language descriptions for the time units mentioned [R6sner 86b].", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "MT for doctor/patient communication", "sec_num": "4.1.2" }, { "text": "In the GEOTEX application the SEMTEX text generator is combined with a tool for interactively creating geometric constructions [Kehl 86 ]. The latter offers formal commands for manipulating (i.e. creating, naming anddeliberately -deleting) basic objects of Euclidean geometry. The generator is used to produce descriptive texts related to the geometric construction:", "cite_spans": [ { "start": 127, "end": 135, "text": "[Kehl 86", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "GEOTEX: Verbalizing objects and operations", "sec_num": "4.2.2" }, { "text": "\u2022 descriptions of the geometric objects involved,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "GEOTEX: Verbalizing objects and operations", "sec_num": "4.2.2" }, { "text": "\u2022 descriptions of the sequence of steps done during a construction.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "GEOTEX: Verbalizing objects and operations", "sec_num": "4.2.2" }, { "text": "When GEOTEX is describing the course of a construction in a concise and coherent text it starts from the sequence of commands of the geometry language. Let us look at an example:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "(PUN SA 15 10) (PUN $B 20 7) (KRE $K $8 $A)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "Each of these commands in turn causes GEOTEX", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "\u2022 to update the associated FLAVOR representation for the domain,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "\u2022 to display (if possible) the objects on the screen (in this case: point $A with coordinates (15, 10), point $B with coordinates (20, 7), circle $K with center $B and through $A),", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "\u2022 to create a message from the operation and give it as input to SEMTEX.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "SEMTEX renders this information in the order given. For the example this resulted in the following text: \"Ich zeichne den Punkt Sa (15/10) ein.\" (I draw point Sa (15/10).) \"Und den Punkt $b (20/7).\"", "cite_spans": [ { "start": 145, "end": 171, "text": "(I draw point Sa (15/10).)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "(And point Sb (20/7).)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "\"Urn ihn schlage ich den Kreis $k durch $a.\" (Around it I draw Sk through Sa.) To achieve this result SEMTEX' contexthandling mechanisms have been enriched:", "cite_spans": [ { "start": 45, "end": 78, "text": "(Around it I draw Sk through Sa.)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "Elision is no longer restricted to adjuncts. For repetitive operations verb and subject will be elided in subsequent sentences (cf. the sentences 1 and 2).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "The distinction between known information (i.e. known geometric objects) and new one (i.e. new objects created from known ones) is exploited to decide on constituent ordering: the constituent referring to the known object is \"topicalized\", i.e. put in front of the sentence (cf. sentence 3).", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "In addition the system allows for more ways to refer to objects introduced in the text: pronouns, textual deixis using demonstrative pronouns (\"dieser Punkt\",this point), names.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "The choice is done deliberately: Pronouns are avoided if their use might create an ambiguity; reference by name is used when an object has not constantly been in focus and therefore has to be re-introduced.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Verbalizing the course of a construction:", "sec_num": null }, { "text": "SEMSYN's Software Environment SEMSYN's generation system has been implemented on a SYMBOLICS lisp machine. During the implementation we aimed at utilizing as much of the functionality of this machine in order to get optimal support for our work. We have built up an environment of linguistic and software tools that, though designed for our projects purposes, may be -at least in part -of interest for other projects in MT and CL in general. 2", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "5", "sec_num": null }, { "text": "Interface tools:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "5.1", "sec_num": null }, { "text": "This comprises all software that provides easy and comfortable communication with the system (even for casual users). SEMSYN's user interface is centered around SEMNET-GRAPHICS, a tool for visualizing semantic nets -the starting point of the generation -as mouse-sensitive graphics [R6sner 86b ]. The graphical representation is embedded in an interface \"frame\" [Weinreb, Moon 81] whose \"panes\" are displaying various intermediate structures -depending on the users chosen \"frame configuration\" -and the generation result.", "cite_spans": [ { "start": 282, "end": 293, "text": "[R6sner 86b", "ref_id": null }, { "start": 362, "end": 371, "text": "[Weinreb,", "ref_id": null }, { "start": 372, "end": 380, "text": "Moon 81]", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "5.1", "sec_num": null }, { "text": "These tools extend the capabilities of the user interface and are intended to enable and support experiments with the system.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation tools:", "sec_num": "5.2" }, { "text": "SEMNET-EDIT is a tool for experimenting the generator by interactively editing semantic nets [Kehl 85 ]:", "cite_spans": [ { "start": 93, "end": 101, "text": "[Kehl 85", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "Experimentation tools:", "sec_num": "5.2" }, { "text": "\u2022 modification of given semantic nets", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation tools:", "sec_num": "5.2" }, { "text": "\u2022 creation of semantic nets from scratch", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation tools:", "sec_num": "5.2" }, { "text": "\u2022 generation of German from created or modified semantic nets and/or their subnets.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation tools:", "sec_num": "5.2" }, { "text": "Experimentation tools of this type are not only useful for purposes of debugging and system improvement but proved as well to be very helpful as comfortable means for introduction into the system's capabilities and limitations.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Experimentation tools:", "sec_num": "5.2" }, { "text": "In every realistic application dictionaries play an important role as body of linguistic knowledge; the need for support in maintaining and updating them seems obvious. SEMSYSTEM uses two types of dictionaries: A single German root form dictionary (with morpho/syntactic information) for the generator front end and so-called \"realization dictionaries\", that relate semantic symbols to German lexical items (root forms of verbs, nouns, adjectives .... ) and that may vary for different applications of the generator. For both types of lexica there are window-and menu-based tools for maintenance. -using the same representation for the different domains -the texts of SEMTEX and GEO-TEX may be produced in English as well.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Lexicon tools:", "sec_num": "5.3" }, { "text": "A system produced example text from the newspaper application:", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Lexicon tools:", "sec_num": "5.3" }, { "text": "Increase in the number of unemployed. NURNBERG/BONN (cpa) DECEM-BER 5,85. The number of unemployed in West Germany has increased slightly during November. It has increased from 2148800 by 61900 to 2210700.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Lexicon tools:", "sec_num": "5.3" }, { "text": "At the end of November the unemployment rate had a value of 8.8 percent. At the end of the year-ago period it had a value of 8.7 percent. Gerd Muhr, the speaker of the DGB, declares, it sees a bad sign in the increase in the number of unemployed.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Lexicon tools:", "sec_num": "5.3" }, { "text": "French will be the next target language; we have started to prepare the morphological and syntactic data for such an experiment.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Lexicon tools:", "sec_num": "5.3" }, { "text": "There is more recent work in \"multilingual generation\" from data: ", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Related work", "sec_num": "6.2" }, { "text": "RAREAS,", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Related work", "sec_num": "6.2" }, { "text": "Generation of natural language texts in different languages (and probably different styles) from the same knowledge base might be an interesting alternative to human or machine translation of these texts. Re-Generation (of e.g. software manuals or maintenance handbooks) in different languages might be much more economic than manually \"updating\" those texts when the underlying knowledge base changes.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Aspects of application:", "sec_num": "6.3.1" }, { "text": "Multilingual generation enforces the separation of generator knowledge into language dependent data and language independent machinery.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Aspects of implementation:", "sec_num": "6.3.2" }, { "text": "In order to keep a generator easily portable to other languages the implementor will have to allow as much declarativity as possible.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Aspects of implementation:", "sec_num": "6.3.2" }, { "text": "Work in multilingual generation from semantic representations may be seen as an exercise in contrastive linguistics: A central issue for any generator starting from semantic structures is the choice of an appropriate syntactic structure for the expression of a given meaning structure. What are the differences and correspondencies between the different target languages with respect to this choice? (E. g. Focus may be expressed by constituent order in German, in English you may have to choose a passive.) Similarly. What are the differences between the target languages with respect to the semantic features that are obligatory in order to be able to produce surface text? (E.g. In the Japanese/German MT application we were confronted with the fact that the semantic structures derived from Japanese did not contain semantic information about in/definiteness or multiplicity.)", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "Aspects of linguistic theory:", "sec_num": "6.3.3" }, { "text": "The SEMSYN project is funded by the West German Ministry for Research and Technology (BMFT). The project is currently cooperating with partners from Japan (University of Kyoto; NTT Basic Research Laboratories) and USA (International Center for MT at CMU). We have to thank all partners for their support.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "ACKNOWLEDGEMENTS:", "sec_num": null }, { "text": "The SEMSYN system is the joint effort of a variety of people. Special thanks to M. Emele (Stuttgart) for his work on the front end generator, to W. Kehl (Stuttgart) for his implementation of GEOTEX and the editor for semantic nets and to O. Rambow (Ithaca, N.Y.) for our joint experiment to teach English to the system.", "cite_spans": [ { "start": 148, "end": 164, "text": "Kehl (Stuttgart)", "ref_id": null } ], "ref_spans": [], "eq_spans": [], "section": "ACKNOWLEDGEMENTS:", "sec_num": null }, { "text": "The SEMSYN generator and the applications as described in this paper are fully implemented and run in ZetaLISP and FLAVORS on SYMBOLICS lisp machines. We would like to demonstrate the system at the conference.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "A note on implementation and demonstration", "sec_num": null }, { "text": "These tools are best illustrated by an interactive demo.", "cite_spans": [], "ref_spans": [], "eq_spans": [], "section": "", "sec_num": null } ], "back_matter": [], "bib_entries": { "BIBREF0": { "ref_id": "b0", "title": "FREGE -Entwicklung und lmplementierung eines objektorientierten FRont-End-GEnerators fffr das Deutsche", "authors": [ { "first": "C", "middle": [], "last": "Contant", "suffix": "" }, { "first": "M", "middle": [], "last": "Emele", "suffix": "" }, { "first": "U", "middle": [], "last": "Heid", "suffix": "" }, { "first": "D", "middle": [], "last": "Rssner", "suffix": "" }, { "first": "B", "middle": [], "last": "Weck", "suffix": "" }, { "first": "G", "middle": [], "last": "Helbig", "suffix": "" }, { "first": "J", "middle": [], "last": "Buscha", "suffix": "" }, { "first": "W", "middle": [], "last": "Kehl", "suffix": "" } ], "year": 1985, "venue": "Generation automatique de texte: application au souslanguage boursier", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Contant, C. \"Generation automa- tique de texte: application au sous- language boursier\", M.A. thesis, Dept. de Linguistique, Univ. de Montreal, 1986. [Emele 86] Emele, M. \"FREGE -Entwick- lung und lmplementierung eines objek- torientierten FRont-End-GEnerators fffr das Deutsche\", Diplomarbeit, Institut ffr Informatik, Uni Stuttgart, 1986. [Heid, R6sner, Weck 87] Heid, U.; RSsner, D.; Weck, B.: 'Das EUROTRA- D/SEMSYN-Experiment: Generierung deutscher S5tze aus semantischen Repr~i- sentationen', in: Tillmann, H.G.; Will6e (Hrg.):\"Analyse und Synthese gesproch- ener Sprache\", (Hildesheim: Olms), 1987. [Helbig, Buscha 86] Helbig, G.; Buscha, J.: \"Deutsche Grarnmatik. Ein Handbuch fh'r den Ausl~nderunterricht\", Leipzig, 1986. [Kehl 85] Kehl, W.: \"Erweiterung der graphischen Schnittstelle des SEMSYN- Projekts\", Studienarbeit, Insitut ffir In- formatik, Univ. Stuttgart, 1985.", "links": null }, "BIBREF1": { "ref_id": "b1", "title": "Natural language generation: New results in Artificial Intelligence, Psychology and Linguistics", "authors": [ { "first": "W", "middle": [], "last": "Kehl", "suffix": "" }, { "first": "R", "middle": [], "last": "Kittredge", "suffix": "" }, { "first": "A", "middle": [], "last": "Polguere", "suffix": "" }, { "first": "E", "middle": [], "last": "Goldberg", "suffix": "" } ], "year": 1986, "venue": "", "volume": "87", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Kehl, W.:\"GEOTEX -Ein System zur Verbalisierung geometrischer Kon- struktionen', Diplomarbeit, Institut fiJr Informatik, Univ. Stuttgart, 1986. [Kempen 87] Kempen, G. (Ed.): \"Natural language generation: New results in Ar- tificial Intelligence, Psychology and Lin- guistics\", Kluwer Academic Publishers, Dordrecht/Boston, 1987. [Kittredge et al. 86] Kittredge, R., A. Polguere, E. Goldberg: \" Syn- thesizing Weather Forecasts from format- ted Data\", in: COLING-86, Proceedings, Bonn, August 1986.", "links": null }, "BIBREF2": { "ref_id": "b2", "title": "Design and Implementation of a Knowledge-Based Report Generator", "authors": [ { "first": "K", "middle": [], "last": "Kukich", "suffix": "" } ], "year": 1983, "venue": "ACL Annual Meeting", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "[Kukich 83] Kukich, K. \"Design and Imple- mentation of a Knowledge-Based Report Generator\", ACL Annual Meeting, Pro- ceedings, 1983.", "links": null }, "BIBREF3": { "ref_id": "b3", "title": "Natural Language Generation as a Computational Problem: an introduction", "authors": [ { "first": "D", "middle": [ "D" ], "last": "Mcdonald", "suffix": "" }, { "first": "D", "middle": [ "D" ], "last": "Mcdonald", "suffix": "" }, { "first": "J", "middle": [ "D" ], "last": "Pustejovsky", "suffix": "" }, { "first": "M", "middle": [ "M" ], "last": "Vaughan", "suffix": "" } ], "year": 1983, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "McDonald, D.D. \"Natural Language Generation as a Com- putational Problem: an introduction\", in: Brady & Berwick (eds.) \"Computational Models of Discourse\", MIT Press, 1983. [McDonald et al. 87] McDonald,D.D., J.D. Pustejovsky, M.M. Vaughan \"Fac- tors contributing to efficiency in natural language generation\", in: [Kempen 87]", "links": null }, "BIBREF4": { "ref_id": "b4", "title": "~Vhen Mariko talks to Siegfried -Experiences from a Japanese~German Machine Translation Project", "authors": [ { "first": "D", "middle": [], "last": "Rssner", "suffix": "" } ], "year": 1986, "venue": "Proceedings", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "[R6sner 86a] RSsner, D.:~Vhen Mariko talks to Siegfried - Experiences from a Japanese~German Machine Translation Project\", in: COLING-86, Proceedings, Bonn, August 1986.", "links": null }, "BIBREF5": { "ref_id": "b5", "title": "Ein System zur Generierung yon deutschen Texten aus semantischen Reprdsentationen", "authors": [ { "first": "D", "middle": [], "last": "Rssner", "suffix": "" } ], "year": 1986, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "[RSsner 86b] RSsner, D.: \"Ein System zur Generierung yon deutschen Texten aus semantischen Reprdsentationen\", Disser- tation, Institut f. Informatik, Univ. Stuttgart, 1986.", "links": null }, "BIBREF6": { "ref_id": "b6", "title": "The automated news agency: the SEMTE,Y text generator for German", "authors": [ { "first": "D", "middle": [], "last": "R6sner", "suffix": "" } ], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "R6sner, D.: \"The automated news agency: the SEMTE,Y text generator for German\", in: [Kempen 87]", "links": null }, "BIBREF7": { "ref_id": "b7", "title": "Generating Se", "authors": [ { "first": "E", "middle": [], "last": "Steiner", "suffix": "" } ], "year": null, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "Steiner, E.: \"Generating Se-", "links": null }, "BIBREF8": { "ref_id": "b8", "title": "Another Stride Towards Knowledge-Based Machine Translation", "authors": [ { "first": "M", "middle": [ "& J" ], "last": "Tomita", "suffix": "" }, { "first": "", "middle": [], "last": "Carbonell", "suffix": "" } ], "year": 1986, "venue": "mantic Structures in EUROTRA-D', in: COLING-86, Proceedings", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "mantic Structures in EUROTRA-D', in: COLING-86, Proceedings, Bonn, August 1986. [Tomita, Carbonell 86] Tomita, M. & J. Carbonell: \"Another Stride Towards Knowledge-Based Machine Translation\", in: COLING-86, Proceedings, Bonn, August 1986.", "links": null }, "BIBREF9": { "ref_id": "b9", "title": "LISP machine manual, MIT", "authors": [ { "first": "D", "middle": [], "last": "Weinreb", "suffix": "" }, { "first": "D", "middle": [], "last": "Moon", "suffix": "" } ], "year": 1981, "venue": "", "volume": "", "issue": "", "pages": "", "other_ids": {}, "num": null, "urls": [], "raw_text": "[Weinreb, Moon 81 ] Weinreb, D. & Moon, D.: \"LISP machine manual, MIT, 1981.", "links": null } }, "ref_entries": { "FIGREF2": { "uris": null, "text": "One possible way to express this relation is to realize the fillers of :REASON and :RE-SULT as clauses and add the clause from :REASON as a subordinate to that of :RE-SULT: \"Macbeth ermordete Duncan, da die ehrgeizige Lady Macbeth Macbeth fiberredete\" (Macbeth murdered Duncan because the ambitious Lady Macbeth persuaded Macbeth)", "num": null, "type_str": "figure" }, "FIGREF3": { "uris": null, "text": "An example: English input to CMU's parser: \"i have a pain in the throat\" Semantic structure as input to SEMSYN: 3 EUROTRA-D/SEMSYN experimentIn order to support the EUROTRA-D group, we ran this experiment:A sample of semantic structures as proposed for use within ECIROTRA[Steiner 86] should serve as input to our generator.", "num": null, "type_str": "figure" }, "TABREF1": { "text": "In order to be able to generate all of these surface forms we had to further enrich the generable fragments with e.g.", "content": "
-infinitival complements
-genitive objects
-subject and object clauses.
4.2Text generation
4.2.1 SEMTEX: Generation of news stories
", "type_str": "table", "num": null, "html": null } } } }