File size: 53,541 Bytes
6fa4bc9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 |
{
"paper_id": "W89-0115",
"header": {
"generated_with": "S2ORC 1.0.0",
"date_generated": "2023-01-19T03:44:52.558527Z"
},
"title": "Identifiering av diskursrefenter vid maskin\u00f6vers\u00e4ttning fr\u00e5n ryska till svenska",
"authors": [
{
"first": "Barbara",
"middle": [
"G"
],
"last": "Awronska-W Erngren",
"suffix": "",
"affiliation": {},
"email": ""
}
],
"year": "",
"venue": null,
"identifiers": {},
"abstract": "The problem o f tracking discourse referents in the process of machine translation which will be discussed in this paper is a part o f my work at SW E T R A (Swedish Computer Translation Research) at the Dept of Linguistics, Lund. The SWETRA-programs, implemented in Prolog and based on a GPSG-inspired formalism called Referent Grammau', abbrevi ated RG (Sigurd 1987, 1988), allow translation of a large repertoire of syntactic constructions between Russian, Swedish and English. Currently, the research is concentrated on translation from Russian into Swedish. One of the difRculties arising in the process o f translating from a Slavic language into a Germanic one is inserting correct definiteness values in the noun phrases, as e.g. Russian and Polish do not make use of definite ness as a regular grammatic category. Generating appropriate definiteness values and their morphological representations in the target language is a very complicated task, as the choice between definite and indefinite NPs depends not only on endophoric (textual), but also on many exophoric (external) factors. The most general rules for use of definite noun phrases in Swedish and in English axe, nevertheless, accessible for implementation in Prolog. Our approach to M T requires tracking discourse referents and the paper will also discuss some problems connected with this approach.",
"pdf_parse": {
"paper_id": "W89-0115",
"_pdf_hash": "",
"abstract": [
{
"text": "The problem o f tracking discourse referents in the process of machine translation which will be discussed in this paper is a part o f my work at SW E T R A (Swedish Computer Translation Research) at the Dept of Linguistics, Lund. The SWETRA-programs, implemented in Prolog and based on a GPSG-inspired formalism called Referent Grammau', abbrevi ated RG (Sigurd 1987, 1988), allow translation of a large repertoire of syntactic constructions between Russian, Swedish and English. Currently, the research is concentrated on translation from Russian into Swedish. One of the difRculties arising in the process o f translating from a Slavic language into a Germanic one is inserting correct definiteness values in the noun phrases, as e.g. Russian and Polish do not make use of definite ness as a regular grammatic category. Generating appropriate definiteness values and their morphological representations in the target language is a very complicated task, as the choice between definite and indefinite NPs depends not only on endophoric (textual), but also on many exophoric (external) factors. The most general rules for use of definite noun phrases in Swedish and in English axe, nevertheless, accessible for implementation in Prolog. Our approach to M T requires tracking discourse referents and the paper will also discuss some problems connected with this approach.",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Abstract",
"sec_num": null
}
],
"body_text": [
{
"text": "In n a n v i p rese n te ra r d en (p r e lim in \u00e4 r a ) a lg o ritm e n s o m a n v \u00e4 n d s i S W E T R A - (2 ). D e n p re lim in \u00e4 ra m o d e lle n in n eh \u00e5 ller fy r a n iv \u00e5 e r (s o m i sin tu r kan in d ela s i su b n iv \u00e5 er; n e d a n st\u00e5 en d e b e sk riv n in g \u00e4 r i viss m \u00e5 n f\u00f6 r e n k la d ): sa m m a k o g n itiv a e n h e t p \u00e5 en h \u00f6 g re n iv \u00e5 -en f\u00f6 re st\u00e4 lln in g a v t e x e tt o b je k t , ett faJctum, en eg en sk a p e tc . F \u00f6 r u ts \u00e4 ttn in g a r f\u00f6 r k oreferen s u p p s t\u00e5 r s\u00e5led es p \u00e5 n iv \u00e5 2",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "n i v",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "(b e g r e p p b a se ra d e p \u00e5 ren t lin g v istisk k u n sk a p ); i fo rts \u00e4 ttn in g e n k om m er v i d o c k a tt -f\u00f6 r a tt f\u00f6 re n k la p ro c e d u r b e s k r iv n in g e n -ib la n d a n v \u00e4 n d a form u lerin ga r s o m \" k o re fe re n ta o r d \" , \" k ore fe re n ta su b sta n tiv \" e t c -i st\u00e4 llet f\u00f6 r den m er e x a k ta , m en a lltf\u00f6 r l\u00e5 n g a fra sen : f\u00f6 r e v e n tu ell m e ta fo risk a n v \u00e4 n d n in g a v o r d e t i fr \u00e5 g a ). A tt v a ra n \u00e5 g on s g ran n e im p lice ra r norm aJlt a tt v a ra en m an n isk a, v ilk et i sin tu r im p lice ra r a tt vara a n tin g e n e n m a n eller en k v in n a -N 2 \u00e4 r a llts\u00e5 m in d re sp e cifik \u00e4n N I . D \u00e4 re m o t i en te x t so m :",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "(5 a ) J a g tr\u00e4 ffa d e en m a n .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "(5 b ) M in g ra n n e var b eru sa d .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "u p p le v s n om in a lfra se r e n m a n o c h m tn g ra n n e s o m ick e-k oreferen ta , e ftersom k o n c e p te t \" m in g ra n n e \" \u00e4 r m e ra sp e cifik t \u00e4n \" en m a n \" .",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "D is tin k tio n e n m e r /m in d r e sp e cifik \u00e4 r n a tu rlig tv is in te a lltid lika klar; e x em p e l (6 ) -e tt fra g m e n t a v S W E T R A -\u00f6 v e r s \u00e4 t t n in g av en ry sk tid n in g ste x t - Ovanst\u00e5ende regel kan l\u00e4sas p\u00e5 f\u00f6ljande s\u00e4tt: ett substantiv med betydelsekod m (A ,p l) kan associeras med en m\u00e4ngd best\u00e5ende av \u00e5tminstone tv\u00e5 tidigare in troducerade referenter, om \u00e5tminstone tv\u00e5 tidigare n\u00e4mnda substantiv kan tolkas som hyponymer till det aktuella substantivets singularform (symboliserad som m(A, s g ) ). Detta \u00e4r en av de enklaste varianterna av coref-predikatet (i praktiken anv\u00e4nds en del ytterligare restriktioner; deras utformning kr\u00e4ver fortsatt arbete).",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "I ex 9 Regeln s\u00e4ger, att begreppet associerat med en lexikal enhet med betydelseko den A \u00e4r mera specifik \u00e4n begreppet associerat med B, om A har de seman tiska drag (Features2) som anses vara mest karakteristiska f\u00f6r B, och dessu tom \u00e5tminstone ett drag, som \u00e4r mera specifikt. I ex (9) inneh\u00e5ller de lexikala enheterna f\u00f6r \" pojke\" och \" flicka\" specifika drag \" male\" resp \" female\" och dessutom edla drag som till\u00e5ter anv\u00e4ndningen av enheten \" barn\" och, f\u00f6ljaktigen, tolkas de som hyponymer till samma ord. Formatet f\u00f6r lexikala enheter \u00e4r i det h\u00e4r fallet f\u00f6ljande (f\u00f6renklad notation): ",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "Begreppen diskursreferent och koreferens",
"sec_num": "1"
},
{
"text": "Proceedings of NODALIDA 1989",
"cite_spans": [],
"ref_spans": [],
"eq_spans": [],
"section": "",
"sec_num": null
}
],
"back_matter": [],
"bib_entries": {
"BIBREF0": {
"ref_id": "b0",
"title": "The introduction och maintenance o f discourse referents. Papers from the Ninth Scandinavian Conference o f Linguistics",
"authors": [
{
"first": "Kaxi",
"middle": [],
"last": "Fraxud",
"suffix": ""
}
],
"year": 1986,
"venue": "",
"volume": "",
"issue": "",
"pages": "",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Fraxud, Kaxi. 1986. The introduction och maintenance o f discourse referents. Papers from the Ninth Scandinavian Conference o f Linguistics. Stockholm.",
"links": null
},
"BIBREF1": {
"ref_id": "b1",
"title": "A Referent Grammatical Analysis of Relative Clauses in Polish",
"authors": [
{
"first": "Barbara",
"middle": [],
"last": "Gawronska^werngren",
"suffix": ""
}
],
"year": 1988,
"venue": "Stadia linguistica",
"volume": "42",
"issue": "",
"pages": "18--48",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Gawronska^Werngren, Barbara. 1988. A Referent Grammatical Analysis of Relative Clauses in Polish. Stadia linguistica 42(l):18-48.",
"links": null
},
"BIBREF3": {
"ref_id": "b3",
"title": "Focusing in the Comprehension o f Definite Anaphora",
"authors": [
{
"first": "Candace",
"middle": [
"L"
],
"last": "Sidner",
"suffix": ""
}
],
"year": 1983,
"venue": "Computational Model o f Discourse",
"volume": "",
"issue": "",
"pages": "267--330",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sidner, Candace L. 1983. Focusing in the Comprehension o f Definite Anaphora. Brady, M. &c Berwick, R. C. [eds.] Computational Model o f Discourse:267-330. MIT Press, Cambridge, Mass.",
"links": null
},
"BIBREF4": {
"ref_id": "b4",
"title": "Referent Grammar (R G ). A generalized phrase structure grammar with built-in referents",
"authors": [
{
"first": "Bengt",
"middle": [],
"last": "Sigurd",
"suffix": ""
}
],
"year": 1987,
"venue": "Studia linguistica",
"volume": "41",
"issue": "2",
"pages": "115--135",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sigurd, Bengt. 1987. Referent Grammar (R G ). A generalized phrase structure grammar with built-in referents. Studia linguistica 41(2):115-135.",
"links": null
},
"BIBREF5": {
"ref_id": "b5",
"title": "Using Referent Grammar (R G ) in computer analysis, generation and translation o f sentences",
"authors": [
{
"first": "Bengt",
"middle": [],
"last": "Sigurd",
"suffix": ""
}
],
"year": 1988,
"venue": "Nordic Journal o f Linguistics",
"volume": "11",
"issue": "",
"pages": "129--150",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sigurd, Bengt. 1988. Using Referent Grammar (R G ) in computer analysis, generation and translation o f sentences. Nordic Journal o f Linguistics 11:129-150.",
"links": null
},
"BIBREF6": {
"ref_id": "b6",
"title": "The potentials o f SW ETRA, a multilanguage MT-system",
"authors": [
{
"first": "Bengt",
"middle": [],
"last": "Sigurd",
"suffix": ""
},
{
"first": "Sc",
"middle": [],
"last": "Gawrodska^werngren",
"suffix": ""
}
],
"year": 1988,
"venue": "Computers and Translation",
"volume": "3",
"issue": "",
"pages": "238--250",
"other_ids": {},
"num": null,
"urls": [],
"raw_text": "Sigurd, Bengt, Sc Gawrodska^Werngren, Barbara. 1988. The potentials o f SW ETRA, a multilanguage MT-system. Computers and Translation 3:238-250.",
"links": null
}
},
"ref_entries": {
"FIGREF0": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "p r o g r a m f\u00f6 r a tt id e n tifie ra d isk u rsreferen ter o c h v \u00e4 lja n o m in a lfra sern a s b e st\u00e4 m dh e ts v \u00e4 r d e , v ill v i k o r tfa tta t b e sk riv a d en d isk u rsm o d e ll, so m p ro ce d u re n b y g g er p \u00e5 o c h f\u00f6 r s \u00f6 k a d e fin ie ra sj\u00e4 lv a b e g r e p p e t \" disk u rsreferen t\" . In fo rm ellt brukar d isk u rsre feren te r k arak teriseras so m \" saker o c h fa k ta s o m m a n ta la r o m \" , vilket im p lic e r a r en e x o fo r is k (ic k e -te x tu e ll) re la tio n -m a n refererar till o b je k t i d e n ick e -sp r\u00e5 k lig a v erk lig h e te n . D e fin itio n e n a v b e g r e p p e t d isk u rsreferen t m \u00e5ste d o c k s a m tid ig t v a ra r e la tera d till te x te n (d e t g\u00e4 ller b l a a tt \u00e4 g n a u p p m \u00e4 rk sa m h et 170 Proceedings of NODALIDA 1989, pages 170-182 \u00e5t s\u00e4 ttet a tt in tr o d u c e r a n y a disk u rsreferen ter i te x te n o c h till d e lin g v istisk a fa k to re r so m m \u00f6 jlig g \u00f6 r a s s o cie rin g a v t v \u00e5 fraser m e d sa m m a d isk u rsre fe re n t). D en p re lim in \u00e4 ra m o d e lle n f\u00f6 r id en tifie rin g av d isk u rsreferen ter s o m k o m m e r a tt presen teras n ed a n in n e h \u00e5 lle r d \u00e4 rf\u00f6 r b \u00e5 d e e n d o fo r is k o c h e x o fo r is k k o m p o n e n t; i den sen are sp e la r b e g r e p p e t \" m en ta l v e rk lig h e t\" (d v s en v \u00e4 rld a v k o g n itiv a en h eter o c h rela tio n e r m ella n d e m ) en p rim \u00e4 r roll. In n an v i \u00f6 v e r g \u00e5 r till den re feren t-g ra m m a tisk a d isk u rsm o d e lle n , v ill v i k o r tfa tta d k o m m e n te ra e n k lassisk d efin ition av s\u00e4 tte t a tt in tr o d u c e r a disk u rsreferen ter fo rm u le ra d a v K a r ttu n e n (K a x ttu n en 1 9 7 6 ): th e a p p e a r a n c e o f a n in d efin ite n o u n p h ra se esta b lish es a d isco u rse referen t ju s t in ca s e it ju stifie s th e o c c u r e n c e o f a co re feren tia l p r o n o u n o r a d e fin ite n o u n p h ra se la te r in th e te x t. O v a n st\u00e5 e n d e d e fin itio n kan in te till\u00e4 m p a s i S W E T R A r s \u00f6 v e r s \u00e4 ttn in g s p r o c e d u r, efte rso m d istin k tio n e n m ellan in d e fin ita o c h in d e fin ita n om in a lfra ser \u00e4r (v id \u00f6 v e r s \u00e4 ttn in g fr\u00e5n ry sk a eller p o ls k a ) o m a rk e ra d i in p u tte x te n . M a n m \u00e5 ste d e ssu to m ta h \u00e4 n syn till d e t falctu m , a tt en n y disk u rsreferen t kan in tro d u ce ra s in te en b a rt a v en N P , u tan o c k s \u00e5 a v t e x e tt v e rb eller en h el m e n in g -s o m i ( 1): ( l a ) Id a g f\u00f6 r f\u00f6 ljd e en o id e n tifie ra d u b \u00e5 t en sv en sk tr\u00e5 la re . ( l b ) J a k ten p \u00e5 g ick i u n g ef\u00e4 r en tim m e . eller av e tt a d je k tiv -s o m i (2 ): (2 a ) K \u00f6 p in te g u la b lo m m o r till h en n e. (2 b ) D e n f\u00e4 rg en ty ck er h on in te o m . D en d isk u rsm od ell, so m v i vill f\u00f6 re sl\u00e5 h \u00e4 r, ta r h \u00e4 n sy n till olik a s\u00e4 tt a tt in tr o d u c e r a n y a d isk u rsreferen ter -b l a till d e m s o m illu stre ra d es i (1 ) o c h"
},
"FIGREF1": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "\u00e5 1 : T e x te n . n i v \u00e5 2 : M e n ta la k o n c e p t so m u p p s t\u00e5 r p \u00e5 b a sis a v lin g v istisk k u n sk ap (d v s k u n skap o m o rd e n s in ten sion e r o c h e x te n s io n e r o c h f\u00f6 r m \u00e5 g a n a tt to lk a sy n ta k tiska s tru k tu re r); d essa k o n ce p t v ill v i p relim in \u00e4 rt in d e la i tre h u v u d g ru p p er -s k n o m in a lisera re (n o m in a liz e rs ), eg e n sk a p e r o c h rela tio n e r; skill n aden m ellan d e ssa b e g r e p p kan (f\u00f6 r e n k la t) f\u00f6rk la ra s i p red ik a tsk a lk y len s term er: \" r e la tio n e r\" o c h \" eg en sk a p er\" kan j\u00e4 m f\u00f6 r a s m e d p re d ik a t, o c h \" n om in a lisera re\" m e d a rg u m en t. V a rje en h et p \u00e5 d e n n a n iv \u00e5 kan a sso cie ra s m ed flera o r d eller ord se k v en ser i te x te n . n i v \u00e5 3 : D isk u rsreferen ter, d v s k o g n itiv a e n h eter s o m k o n stitu e ra s p \u00e5 b a sis av d e slu tsa tser so m kan d ras m e d u tg \u00e5 n g sp u n k t fr\u00e5 n d e e g en sk a p er h os n o m in a lisera re o c h d e re la tion er m ella n n om in a lisera re v ilk a rep resen tera s p \u00e5 n iv \u00e5 2. S lu tsa tsern a kan v a ra b a s e ra d e a n tin g en p \u00e5 lin g v istisk a eller ick e-lin gvistisk a k u n sk aper. E n disk u rsreferen t kan a ssociera s m e d flera e n h eter p \u00e5 n iv \u00e5 2 (k orefe ren s). n i v \u00e5 4 : O b je k t o c h fa k ta i d e n \" v e rk lig a v \u00e4 rld e n \" ; d e n n a n iv \u00e5 b e h \u00f6 v e r inte v a ra rep resen tera d i a lla d isk u rsty p e r, e fte rso m o b je k te t s ex isten s respek tiv e ick e -e x iste n s i d en fy sisk a verk lig h eten in te p\u00e5verk ar m \u00f6jlig h e te n a tt k o n s titu e ra d isk u rsreferen ter s o m i e tt a v e x e m p le n i F raru d (P raru d 1986o u ld o n ly fig h t w ith th e ca t. I e n lig h et m e d d en o v a n sk issera d e m o d e lle n u p p s t\u00e5 r k oreferen s o m tv \u00e5 (eller fler) m e n ta la k o n c e p t fra m k a lla d e a v o r d eller o rd se k v en s e r kan associera s m ed"
},
"FIGREF2": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "\" o r d a sso cie ra d e m e d k o re feren ta b e g r e p p \" . E tt a v v illk o ren f\u00f6 r p o te n tie ll k oreferen s (s o m s to r a d ela r a v v \u00e5 r k om p uta tio n e lla m o d e ll b y g g e r p \u00e5 ) kan form u lera s p \u00e5 f\u00f6 lja n d e s\u00e4 tt: t v \u00e5 k o n ce p t ( C l o c h C 2 ) fra m k a lla d e a v o r d eller ord sek v en se r i te x te n ( T I o c h T 2 ) kan i regel a sso cie ra s m e d sa m m a referen t R , o m d e t sen are in tr o d u c e r a d e k o n ce p te t (C 2 ) \u00e4r in te o j\u00e4 m f\u00f6 r b a r m e d C l o c h o m C 2 in te \u00e4 r m e ra sp ecifik t \u00e4n C l . M ed \" m e r a s p e c ifik t\" m e n a r v i f\u00f6 lja n d e rela tion : C l \u00e4 r m era sp ecifik t \u00e4n C 2 , om f\u00f6 re st\u00e4 lln in g e n \" a tt v a ra C l \" im p lice ra r (d irek t eller in d ire k t) \" a tt o c k s \u00e5 vara C 2 \" . R e s o n e m a n g e t kan illu strera s m ed e x e m p e l (4 ) o c h (n e n v a r b e ru sa d . N o m in a lfra s e rn a m in g ra n n e o c h m a n n en (e ller, m e ra e x a k t, d e \" nom inaJise ra re\" s o m a ssocie ra s m e d d e m -l\u00e5 t o ss k alla d e m f\u00f6 r N I o c h N 2 ) u p p fa tta s som k o re fe re n ta e fte r s o m d e t s te r e o t y p a k o n c e p te t a v \" g ra n n e \" kan represen teras som \" en m \u00e4 n n isk a s o m b o r n \u00e4 ra t a la r e n /d e n tid ig a re n \u00e4 m n d a p e rso n e n \" (v i b ortse r"
},
"FIGREF3": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "isa r e tt m e r a p r o b le m a tis k t fall: (6a) Israeliska flygplan utf\u00f6rde idag tre bombattacker \u00f6ver libanesiskt territorium. (6b) Femton personer d\u00f6dades som resultat av luftpiraternas barbariska aktion. D e t fa k tu m , a tt b e g r e p p e n \" israeliska fly g p la n \" o c h \" lu ftp ira te rn a \" in ter p reteras s o m k ore fe ren ta , kan m \u00f6 jlig a n f\u00f6 rk la ra s p \u00e5 f\u00f6 lja n d e s\u00e4 tt: C l -so m kan rep resen teras s o m i s r a e l i ( N l ) , d \u00e4 r N I m o ts v a r a r b e g r e p p e t \" fly g p la n \" (b e g r e p p e t in k lu d era r in te b a r a fly g p la n so m m a sk in er m en o c k s \u00e5 d e ra s \" em im ata d ela r\" -p ilo te r ) im p lice ra r -i en ligh et m e d d e v \u00e4 rd e rin g a r s o m \u00e4 r a k tu e lla i s\u00e4n d aren s k u ltu r v is s a n e g a tiv a eg e n sk a p er (a t t v a ra en israelisk p ilo t kan in n eb \u00e4 ra -m ed en h \u00f6 g sa n n olik h etsg ra d -a tt v a ra o c k s \u00e5 en lu ftp ir a t, en b a n d it o s v ). V id en s\u00e5 d a n in te rp re ta tio n sk u lle d e t o v a n fo r m u le r a d e v illk o re t f\u00f6 r m \u00f6 jlig k oreferen s v a ra u p p fy llt. E tt f\u00f6 rs \u00f6 k a tt m ed h j\u00e4 lp a v en k o m p u ta tio n e ll p ro c e d u r id en tifie ra k o referen s i lik n a n d e fa ll k o m m e r a tt p resen tera s i a v s n itt 2. 1. En prelimin\u00e4r modell for identifiering av diskursreferenter S W E T R A -p r o g r a m f\u00f6 r \u00f6 v e r s \u00e4 ttn in g m ella n v is s a slavisk a o c h g erm a n sk a sp r\u00e5 k \u00e4r -so m d e t h ar p \u00e5 p ek a ts tid ig a r e -b a se ra d e p \u00e5 en G P S G -in s p ir e r a d fo r m a l ism -refere n tg ra m m a tik (R e fe re n t G r a m m a r , R G ; S ig u rd 1987, 1 9 8 8 ). M \u00f6 jlig h eter a tt a n v \u00e4 n d a R G f\u00f6 r p a rsin g o c h \u00f6 v e r s \u00e4 ttn in g a v v issa p o lsk a o c h rysk a syn ta k tisk a k o n stru k tio n e r h a r b e sk riv its i G a w ro n sk a rW ern g ren (1 9 8 8 ) o c h i S igu rd & G a w ro n sk a -W ern g ren (1 9 8 8 ). P r o c e d u r e n s o m f\u00f6 r n \u00e4 rv a ra n d e a n v \u00e4 n d s v id \u00f6 v e r s \u00e4 ttn in g fr\u00e5 n ry sk a till sven ska kan in d ela s i f\u00f6 lja n d e tre h u v u d sta d ie r: 1. p a rsin g a v in p u t-m e n in g e n o c h fo rm u le rin g a v en so rts in terlin g u a rep rese n ta tio n , s k fu n k tio n e ll rep rese n ta tion (f-r e p r e s e n ta tio n ), s o m in n eh \u00e5 lller in fo r m a tio n o m m en in g en s sy n ta k tisk a stru k tu r (u tt r y c k t m e d h j\u00e4 lp a v s \u00e5 d a n a tra d itio n e lla te rm er s o m s u b je k t, o b je k t , p re d ik a t, a d v e rb ia l o c h s a ts a d v e rb ia l), o rd e n s b e ty d e ls e k o d e r (fo rm u le ra d e i \" m a sk in en g elsk a \" ) sa m t k o d e r f\u00f6 r v issa g ra m m a tisk a d r a g , s o m n u m eru s, g en u s, v \u00e4 r d e t + /a n im a t, te m p u s o s v . T e x en en k el ry sk m e n in g a n a ly sen f\u00f6 lja n d e f-re p re s e n ta tio n (f\u00f6 r e n k la t): s ( s u b j ( n p ( r ( _ , m ( b o y , s g ) , D , s g , a n i , m a , _ , _ ) , H , R e l ) ) , p r e d ( r u n . p a s t ) , s a d v l ( [ ] ) , s a d v l ( [ ] ) , a d v l (h o m e) , a d v l ( [ ] ) , a d v l ( [ ] ) ) . S y m b o le n [ ] (t o m m a m \u00e4 n g d e n ) a v sp e g la r d e t fa k tu m a tt m en in g e n in te in n eh \u00e5ller n \u00e5 g ra sa tsa d v e rb ia l o c h in g a fler a d v e rb ia l \u00e4n d o m o j (h o m e ). P r o lo g -a t o m e n m e d fu n k to rn r b ru k a d es i n \u00e5 g ra tid ig a re sk rifter o m R G kallas f\u00f6 r \" re feren tb esk riv n in g \" (refere n t d e s k r ip tio n ), v ilk et n a tu rlig tv is m \u00e5 ste b e tra k ta s s o m en a p p r o x im a tio n ; d e n n a en h et in n eh \u00e5 lle r sn<irare en b e s k r iv n in g a v n o m in a lfra se n s h u v u d o rd , so m u n d er \u00f6v e rs\u00e4 ttn in g sp ro ce ssen s n \u00e4 s ta e ta p p m \u00f6 jlig g \u00f6 r id e n tifie rin g av d isk u rsreferen ter. I fo r ts \u00e4 tt n in g e n k o m m e r d en d elen a v n o m in a lfra se n s rep rese n ta tion (o ft a b eteck n a d m e d s y m b o le n R) kallas f\u00f6 r \" referen t n u cleu s\" . V a ria b eln D -b e st\u00e4 m d h e t -f\u00f6 r b lir o in s ta n tie ra d u n d er d en f\u00f6 r s ta fasen a v \u00f6v e rs\u00e4 ttn in g sp ro ce d u re n . V a ria b le rn a H o c h Rel a n v \u00e4 n d s f\u00d6r a tt la g ra in fo r m a tio n o m ev en tu ella a tt r ib u t: Rel kan in n e h \u00e5 lla en fu n k tio n e ll rep re sen ta tion a v en rela tivsa ts, m e d a n k o d e r f\u00f6 r \u00f6 v r ig a p re -o c h p o s tn o m in a la a ttr ib u t p la cera s i en heten H. 2. b e ty d e ls e k o d e r f\u00f6 r sy n ta k tis k a k o n stitu en ter la gra s i listo r; d \u00e4 re fter b \u00f6 rja r s\u00f6 k n in g e fte r e v e n tu e lla k orefere n ta fraser o c h a n a ly s a v d e n tid ig a re te xtu e lla in fo r m a tio n e n ; p r o c e d u r e n s m \u00e5 l \u00e4r a tt in sta n tiera va ria b eln D som \" d e f ' (-l-d e fin it) i d e fa ll, d \u00e5 te x tu e lla fa k to re r im p lice ra r b e st\u00e4 m d h e t. M en in g en s fu n k tio n e lla rep resen ta tion la gra s o c k s \u00e5 i d a ta b a se n . 3. g e n ere rin g a v in p u tm e n in g e n s ek v iv a len t i m \u00e5 lsp r\u00e5 k et. P \u00e5 sta d iet B sker en o m fo r m u le r in g a v m en in g en s f-re p re s e n ta tio n till en P R O L O G -lis t aen en k el tra n sitiv sa ts f\u00e5 r d \u00e5 f\u00f6 lja n d e fo rm : [subj(X),pred(Y),obj(Z),sadvl(Sl),sadvl(S2),advl(Al), advl(A2),advl(A3)] D \u00e4 r e fte r j\u00e4 m f\u00f6 r s v a rje n o m in a l k on stitu en t m e d tid ig a re \u00f6 v e r s a tta su b s ta n tiv , v e r b o c h a tt r ib u t. D e tid ig a re \u00f6 v e r s a tta o rd e n s b e ty d e ls e k o d e r \u00e4r sa m la d e i t v \u00e5 se p a r a ta lis to r -lista n s o m in n eh \u00e5 ller su b sta n tiv en s o c h v e rb e n s b e ty d e ls e k o d e r k o m m e r i fo rts \u00e4 ttn in g e n a tt kallas f\u00f6 r R -lista n , lista n i v ilk en a tt r ib u tiv a b e s t\u00e4 m n in g a r p la cera s -f\u00f6 r A -lis ta n . S k\u00e4let f\u00f6 r d e n so rte n s in d e ln in g \u00e4 r ick e -lin g v istisk t -s\u00f6 k n in g sp ro ce d u re n verkar helt en k elt fu n g e ra m e ra e ffe k tiv t, o m h u v u d o rd e n s k o d e r sa m la s i en sep 2wat lista . S \u00f6 k n in g e fte r k oreferen ta k on stitu e n ter o c h ev en tu e lla b e st\u00e4 m d h e tsin d ic e r g e n o m f\u00f6 r s m e d h j\u00e4 lp a v en rek u rsiv p r o c e d u r , s o m term in erar n \u00e4 r d en fu n k tio n e lla re p resen ta tion en in te in n eh \u00e5 ller flera n o m in e lla kon stitu e n te r. D e t rysk a le x ik o n e t s o m in k lu d era r en v iss (m y c k e t f\u00f6ren k la d ) in fo r m a tio n o m o r d e n s sem a n tisk a k a te g o ritillh \u00f6 rig h e t \u00e4r tillg \u00e4 n g lig t u nder d e n n a d el a v p ro c e d u r e n . I b \u00f6 r ja n a v \u00f6 v e r s \u00e4 ttn in g s p r o c e s s e n \u00e4r b \u00e5 d e R -o c h A -lista n to m m a . P r o c e d u ren s o m a n v \u00e4 n d s f\u00f6 r a tt la g ra d e n in fo r m a tio n , so m sen are kan u tn y ttja s f\u00f6 r id en tifierin g a v d isk u rsreferen ter, kan besk riv a s p \u00e5 f\u00f6 lja n d e s\u00e4 tt: 1 A r s u b je k ts p la ts e n i d en fu n k tio n e lla re p re sen ta tion en to m ? (su b je k te t rep resen tera s s o m en to m m \u00e4 n g d t e x v id cinalys a v ryska o p e r s o n lig a k o n stru k tio n e r a v ty p e n : u hito p j a t ' \u00e9 e lo v e k -'fem m \u00e4 n n isk or d \u00f6 d a d e s ' , d \u00f6 d a fe m m \u00e4 n n isk or -l-u npers O m ja , g \u00e5 \u00f6 v e r till f-re p re s e n ta tio n e n s n \u00e4 sta k o n stitu en t (p re d ik a te t), la g r a d ess b e ty d e ls e k o d i R -lis ta n o c h fo r ts \u00e4 tt till n \u00e4 sta k on stitu en t -tills en k on stitu en t so m in n e h \u00e5 lle i en N P p \u00e5 tr\u00e4 ffa s. E fte r a tt h a h itta t d en f\u00f6 r s ta N P , g \u00e5 \u00f6 v e r till 2. O m s u b je k ts p la ts e n in te \u00e4x to m , g \u00e4 lle r d et o c k s \u00e5 a tt g \u00e5 \u00f6 v e r till 2. 2 K o n tr o lle r a f\u00f6 r s t, o m e n h e te n H in n eh \u00e5 ller n \u00e5 g ra a ttr ib u tk o d e r . O m a lla p la tse r i H \u00e4 r to m m a , p la c e r a h u v u d o r d e ts b e ty d e ls e k o d i R -lis ta n o c h f\u00f6rse d e n m e d n u m ret N I = N + 1, d \u00e4 r N = a n ta le t elem en t i R -lis ta n . G \u00e5 \u00f6 v e r till 5. O m n \u00e5 g ra a ttr ib u t f\u00f6 r e k o m m e r , g \u00e5 \u00f6 v e r till 3. 3 K o n tr o lle r a o m p la tsen a v se d d f\u00f6 r r\u00e4 k n eord in n eh \u00e5 ller en k o n sta n t, t e x 2. O m in te , g \u00e5 \u00f6 v e r till 4. O m ja , \u00e4 n d r a h u v u d o r d e ts b e ty d e ls e k o d R \u00e5 n m ( X , p l ) till m ( X ,2 ) o c h p la c e r a d en i R -lis ta n (m e d l\u00e4 m p lig t in d e x n u m m e r ). D e n n a d el a v p r o c e d u r e n a n v \u00e4 n d s f\u00f6 r a tt m \u00f6 jlig g \u00f6 r a g e n ererin g a v b e s t\u00e4 m d fo r m i te x te r a v ty p e n : T v\u00e5 p o jk a r spran g. D e n e n e v a r liten . D en andre . . . G \u00e5 \u00f6 v e r till 4. 4 In n eh \u00e5ller den sista p la tsen i e n h e ten H en k o d a v ty p e n m ( X , p r o p ) , d v s e tt e g e n n a m n ? (a p p o s it io n s fa ll). O m j a -p la c e r a h u v u d o r d e ts b e ty d e ls e k o d o c h eg e n n a m n e ts rep re se n ta tio n i R -lis ta n o c h f\u00f6 rse b \u00e4 g g e m ed sa m m a n u m m er (f\u00f6 r a tt h a n tera fa ll s o m t e x p r o fe s s o r A n d e r s s o n , d \u00e4 r sa m m a referen t kan i fo r ts \u00e4 ttn in g e n \u00e5 b e r o p a s a n tin gen m ed h j\u00e4 lp a v e n b a rt e g e n n a m n e t eller e n b a rt tite ln ). In sta n tie ra va ria b eln D (b e s t \u00e4 m d h e t ) s o m d e f (-| -best\u00e4m d). O m n o m in a lfra sen in n eh \u00e5ller a n d r a a ttr ib u tk o d e r , p la /:era d e m i lista n A . G \u00e5 till 5. O m en h eten H in te in n eh \u00e5 ller n \u00e5 g o t e g e n n a m n i a p p o s itio n , m en a n d ra a ttr ib u t, p la ce ra o c k s \u00e5 d era s k o d e r i A -lis ta n , p la c e r a h u v u d o r d e ts k o d i R -lis ta n o c h g \u00e5 \u00f6 v e r till 5. 5 O m d en fu n k tio n e lla re p re sen ta tio n e n in n eh \u00e5 ller fler k o n stitu en ter, s \u00f6 k efte r n \u00e4 sta N P o c h -e fte r a tt h a fu n n it den -u p p r e p a p r o ce d u re n f r o m p u n k t 2. O m d en fu n k tio n e lla re p re se n ta tio n e r in te in n eh \u00e5 ller flera nom inaJfraser, o m fo r m u le r a d en fu n k tio n e lla rep resen ta tio n e n till d ess u rsp ru n g lig a fo r m (en P R O L O G -a t o m m e d fu n k to rn s) o c h g \u00e5 \u00f6 v e r till e ta p p C (g e n e re rin g a v m e n in g e n i m \u00e5 lsp r\u00e5 k e t). D en o v a n b e s k riv n a p r o c e d u r e n in n eh \u00e5 ller i p ra k tik e n flera sta d ie r: b l a en d e lp r o c e d u r s o m m \u00f6 jlig g \u00f6 r in s\u00e4 ttn in g a v sven ska p osse ssiv a p r o n o m e n fr a m f\u00f6 r b e te ck n in g a r f\u00f6 r sl\u00e4 k tska p o c h lik n a n d e rela tio n e r (t y p m o rfa r, g ra n n e e t c ) o c h en d el a n d ra su b s ta n tiv s o m k r\u00e4ver f\u00f6 re k o m ste n a v e tt p o sse ssiv t a ttr ib u t i svenskan, m en in te i ryskan o c h p olsk a n (e n in p u tm e n in g a v ty p e n sp otk a lem sgsiad a sk u lle a llts\u00e5 \u00f6 v e rs \u00e4 tta s till sven sk a so m ja g tr\u00e4 ffa d e m in g ra n n e, \u00e4ven o m d en p o lsk a n om in a lfra sen sg.siada (g r a n n e ) in te in n eh \u00e5 ller n \u00e5 g o t p o sse ssiv t p r o n o m e n ). D ela r a v k o re fere n ss\u00f6k n in g sp roced u re n k o m m e r a tt illu strera s m e d n \u00e5 g ra te st-e x e m p e l (d e m o s ). tax, a tt \u00e5 tm in s to n e en m e n in g h ar b liv it a n a ly sera d o c h \u00f6 v e rsa tt. R -lista n (s u b s ta n tiv -o c h v e r b k o d e r ) o c h m \u00f6 jlig e n o c k s \u00e5 A -lis ta n (a ttr ib u tk o d e r ) inne h \u00e5 ller s\u00e5 led es n \u00e5 g ra elem en t. D e n f\u00f6 r s t a fr\u00e5gan so m st\u00e4 lls n \u00e4r en N P p\u00e5tr\u00e4ffas i d e n a k tu e lla f-re p re s e n tra tio n e n \u00e4r: kan n \u00e5 g o t a v d e tid ig a re \u00f6 v e rsa tta ord en a sso cie ra s m e d sa m m a referen t so m d e n a k tu ella n om in a lfra sen ? F \u00f6 r s ta steget i s v a rs p ro ce d u re n \u00e4r a tt u n d ers\u00f6k a , o m den lexik ala en h eten so m m otsv a ra r d en a k tu e lla N P :s h u v u d o r d \u00e4r su b k la ssificera t so m \" re la tio n \" eller \" egen sk ap\" . O m d en lex ik a la e n h e ten (r le x ) in n eh \u00e5 ller k on sta n ten \" p r o p e r ty \" m ed even tu ell v id a r e su b k la ssifice rin g ( e x e m p e lv is \" c o lo u r \" ), b e s t\u00e5 r p ro c e d u r e n s n \u00e4 sta steg i a tt u n d e rs\u00f6k a , o m A -lis ta n in n eh \u00e5 ller e tt a d je k tiv s o m tillh \u00f6 r sa m m a kategori; o m s \u00e5 \u00e4 r fa lle t, kan k on sta n ten \" d e f\" p la cera s i d et a k tu e lla su b sta n tiv ets karak te ristik . D e t t a \u00e4r n a tu rlig tv is en a p p r o x im a tio n ; a lg o ritm e n b o r d e berikas m ed en d e l r e strik tio n e r, m en d en g e r i reg el k o rrek ta re su lta t v id \u00f6 v e rs\u00e4 ttn in g av e n k la te x tfr a g m e n t. E n a n a lo g p r in c ip till\u00e4 m p a s o m d e t a k tu ella su b sta n tiv et \u00e4 r i le x ik o n e t su b k la ssificera t s o m \" re la tio n \" eller \" a k tio n \" -i d et h \u00e4r fallet s\u00f6 k er p ro g r a m m e t i f\u00f6 r s t a h a n d ig e n o m R -lis t2 in o c h le ta r e fte r e tt v e rb vars sem a n tisk a b e s k riv n in g i le x ik o n e t sku lle m \u00f6 jlig g \u00f6 r a k oreferen s m e d den a k tu el la n om in a lfra se n . D e n d elen a v p ro c e d u r e n m \u00f6 jlig g \u00f6 r k orrek t \u00f6 v e rs \u00e4 ttn in g av sek v en ser s o m (1 ). D e n rysk a v ersion en a v e x e m p e l (1 ) \u00e4 r f\u00f6 lja n d e : sk a le x ik o n e n h e te r so m m o ts v a ra r d e t tra n sitiv a v e rb e t o c h d et v e rb a la su b s ta n tiv e t i te x te n h a r f\u00f6 lja n d e fo r m (f\u00f6 r e n k la t): r l e x C p r e s l e d o v a t ' ,m(chase,_),v,vt,inf [follow,Chase,hunt]. rlex(presledovanie,m(hunt,sg ),n, sg, ina,ne,nom, argument-relation O m p r o g r a m m e t in te h itta r n \u00e5 g o t k oreferen t a d je k tiv eller v e rb , b \u00f6 r ja r d et s\u00f6k a e fte r en tid ig a r e \u00f6 v e r s a tt k oreferen t N P . D e t en k last fa llet \u00e4r n a tu rligtv is k o referen s m ella n n o m in a l& a ser m e d id en tisk a b e ty d e ls e k o d e r. O m R -lista n in n eh \u00e5 ller en k o d s o m in te sk iljer sig fr\u00e5 n d en a k tu e lla , \u00e5 terst\u00e5 r b a r a a tt unders\u00f6ka ev en tu ella m o tin d ic ie r (p r o g r a m m e t k o n tro lle ra r t e x o m d e t a k tu e lla o r d e t in te f\u00f6 reg \u00e5 s a v e tt a ttr ib u t a v ty p e n annari) o c h -ifall in g a s \u00e5 d a n a finn sin sta n tiera v a ria b eln D s o m d ef. D en d elen a v a lg o ritm e n m \u00f6 jlig g \u00f6 r g e n e re rin g a v b e s t \u00e4 m d /o b e s t \u00e4 m d fo r m i e x e m p e l s o m (7 ) o c h (8 Ett israeliskt flygplan d\u00f6k upp s\u00f6der om Saida. (7d) Flygplanet f\u00f6rflyttar sig v\u00e4sterut."
},
"FIGREF4": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "upp ett annat flygplan. O m R -listein in te in n eh \u00e5 ller n \u00e5 g o n k o d s o m \u00e4 r id en tisk m e d d e t a k tu e lla o r d e ts b e ty d e ls e k o d , fo r ts \u00e4 tte r s \u00f6 k n in g s p ro ce d u re n ; d e t g \u00e4 ller b l a a tt g e svar p \u00e5 f\u00f6 lja n d e fr\u00e5 g or: \u2022 o m d en a k tu ella n om in a lfra sen h ar v \u00e4 r d e t + p lu r a l: in n eh \u00e5 ller R -lis ta n \u00e5 tm in s to n e t v \u00e5 e n h eter m ed sa m m a b e ty d e ls e s y m b o l (d e n f\u00f6 r s t a d elen av m e n in g s k o d e n ), m en m ed v \u00e4 rd e t \" sg \" ? E ller h a r m a n tid ig a re p \u00e5 tr\u00e4 ffa t \u00e5 tm in s to n e t v \u00e5 en h e ter so m b ild a r en m \u00e4 n g d s o m kan v a ra k orefe ren t m e d d en a k tu e lla fra sen ? \u2022 in n eh \u00e5 ller R -lis ta n en k o d , s o m kan a sso cie ra s m e d e tt k o n c e p t ( \" n o m in alizer\" ) s o m in te \u00e4 r o j\u00e4 m f\u00f6 r b a r t o c h in te m er sp e cifik t \u00e4 n d e t a h tu ella o r d e ts \" n o m in a liz e r\" ? D essa d ela r a v p ro c e d u r e n id en tifiera r k oreferen s l'\u00e9ik (pojke) och d ev o \u00e9k a (flicka) \u00e4r i lexikonet speciflerade som subkategorier av \" barn\"; det PROLOG-predikatet som identifierar korefer ens i fall som (9) \u00e4r formulerat som: coref(m (A ,pl).R list)h ypon ym sC m C A pSg), [H IT ].R lis t), T / -[ ] ."
},
"FIGREF5": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "identifieras allts\u00e5 substantiven m a l'\u00e9ik och d ev o \u00e9k a som hyponymer till den lexikala enheten med betydelsekod m(c h i l d , s g ) och koreferensen upp t\u00e4ckts med hj\u00e4lp av f\u00f6ljande predikat, som rekursivt letar efter m\u00f6jliga hy ponymer till det aktuella ordets {h \u00e4i-.rebjata -b a m ) singularform: hyponyms(m(A,N), [m(B,N) I R est], [r(_,m (B,N)) I R e s t l ] ) :-more_restricted(in(B,N) ,m(A,N)), hyponyms(m(A,N),Rest,Restl). hyponyms(iii(A,N) .Hyponymlist, [H lT ]):hypoiiyiiis(iii(A,N) .Hyponymlist,T). hyponyms(m(A,N), [ ] , [ ] ) : -! . Variabeln H yponym list (symboliserad i huvudpredikatet \" coreP som [H lT]) betecknar naturligtvis en lista, i vilken eventuella hyponymer lagras under s\u00f6kningsproceduren. Predikatet m o r e _ r e s t r ic t e d \u00e4r -i dess enklaste versionformulerat som: m ore_restricted(A ,B ): -rlez(_,A , F e a t u r e s l , Features2 rlex(_,B,_,_,_,_,_,Features2,."
},
"FIGREF6": {
"uris": null,
"type_str": "figure",
"num": null,
"text": "rlez(m al*eik ,m (boy.sg),n ,sg,an i,m a,n om ,[m ale], [ch ild ] rle z (d e v o C k a ,m (g irl,s g ),n ,s g ,a n i,fe ,n o m ,[fem ale], [ch ild ] rlez(reben ok ,m (ch ild,sg),n ,B g,an i,m a,nom ,[ c h il d ] , [human].D e t ov a n v isa d e fa llet a v k o referen sid en tifierin g \u00e4r n a tu rlig tv is m y ck e t en k elt; a tt b y g g a u p p en h ierark i a v sem a n tisk a d r a g s o m sk u lle fu n g e r a e ffe k tiv t v id s\u00f6k n in g efter h y p o n y m e r i m era k o m p lic e r a d e te x te r \u00e4r in g e n en k el u p p g ift; u tfo rm n in g a v lexik ala e n h e ter i S W E T R A \u00e4r i d e t t a a v se e n d e \u00e4 n s \u00e5 l\u00e4 n g e in te fu llst\u00e4 n d ig . V id \u00f6 v e r s \u00e4 ttn in g a v k o r ta te x tfra g m e n t \u00e4r d e t d o c k m \u00f6 jlig t a tt id en tifiera en g em e n sa m d isk u rsreferen t i fa ll s o m \u00e4 r m in d re \" s j\u00e4 lv k la ra \" \u00e4n e x (9 ) -so m t e x d en tid ig a r e d isk u te ra d e koreferen sen m ella n \" israelisk a fly g p la n \" o c h \" lu ftp ira te r\" . I d e n sorten s fall Jinv\u00e4nds f\u00f6 lja n d e P R O L O G -p r e d ik a t: p o B s i b le _ c o r e f ( A , B ) r l e x ( _ , A , i ( _ , B,_ , _ , _ , _ , _ , F 2 , F 3 , e v a lu a t i o n _ i n ( F 2 ), c o _ e l t ( F l , F 2 ) . P re d ik a te t e v a l u a t i o n . i n in n eb \u00e4 r, a tt lista n F 2 (se m a n tisk k a ra k teristik ) in n eh \u00e5ller e tt d r a g so m \u00e4r k lassificerat s o m \" v \u00e4 rd e rin g \" ; d e t en k la ste s \u00e4 tte t a tt i en k o m p u ta tio n e ll m o d e ll id en tifiera k oreferen s m ella n e tt su b s ta n tiv s o m \u00e4r m ark erat i fr\u00e5 g a o m v \u00e4 rd e rin g s k o m p o n e n te n o c h e tt o m a rk e ra t s\u00e5 d a n t \u00e4r a tt b e tra k ta v \u00e4 rd e rin g sk o m p o n e n te n e n b a rt so m e tt u ttr y c k f\u00f6 r s\u00e4 n d a ren s a tt ity d , o c h in te so m en fa k to r s o m p \u00e5 v erk a r referen srela tion en . E fte rso m d e lex ik a la en h ete rn a i v \u00e5 rt e x e m p e l in n eh \u00e5 ller g e m e n s a m m a d r a g (i lis to r n a F l o c h F 2 ) o c h p ro g ra m m e t in te h itta r n \u00e5 g ra in d ice r m o t k oreferen s, a sso ciera s b \u00e5 d e \" israeliska fly g p la n \" o c h \" lu ftp ira te r\" m ed seim m a referen t. N e d a n f\u00f6 r visa s -i en f\u00f6 re n k la d n o ta tio n -d e lexik a la e n h ete r so m i d e t d isk u te ra d e fa llet m \u00f6 jlig g \u00f6 r id en tifie rin g av koreferen srela tion en : r l e z ( s a m o le ty , 0( a i r p l a n e , p l ) , n , p l , i n a , ma, nom, [airplane,pilot], [machina,h u m a n ] . r le z ([v o z d u B n 7j e , p i r a t 7] , m ( a i r _ p i r a t e , p l ) , n , p l , a n i , ma,nom,[neg,pilot], [ h u m a n ] . F r\u00e5 g a n , vilk a o c h hur m \u00e5 n g a g e m e n sa m m a d r a g so m b e h \u00f6 v s f\u00f6 r a tt t v \u00e5 \" n o min alizers\" ska u p p fa tta s so m k oreferen ta , \u00e4r n a tu rlig tv is k o m p lice ra d , o c h d e n ak tu ella v e rsion en a v p r o g r a m m e t u teslu ter in te v issa fel o c h \u00f6 v erg en era liserin g a x. F \u00f6 r u t o m d e n \u00e4 m n d a pred ik a ten in n eh \u00e5ller p r o g r a m m e t o c k s \u00e5 en p re lim in \u00e4 r p ro c e d u r f\u00f6 r id e n tifierin g disk u rsreferen ter \u00e4 v en n \u00e4r n o m in a lfra sen s h u v u d o r d \u00e4r ellid erat. P r o c e d u r e n k r\u00e4 v e r v id a re e la b o re rin g , m en dess n u v a ra n d e u tfo r m n in g g er m \u00f6jlig h e t a tt g e n e re ra r\u00e4 tt artikel o c h k on tro lle ra k on g ru en s i sek v en ser Tv\u00e5 flygplan d\u00f6k upp s\u00f6der om Saida (lOd) Det ena f\u00f6rflyttar sig mot v\u00e4ster. juznee Sajdy. s\u00f6der om saida E fte r refe ren tid en tifierin g f\u00f6 lje r \u00f6 v e rs\u00e4 ttn in g sp ro ce sse n s sista e ta p p -g en e rerin g a v o u tp u tm e n in g e n . O m v a ria b eln D fo rtfa ra n d e \u00e4r o in sta n tie ra d (so m t e x v id \u00f6 v e r s \u00e4 ttn in g a v te x te n s f\u00f6 r s t a n om in a lfra s eller i d e fall, d \u00e5 in gen koreferent fra s h a r u p p t \u00e4 c k ts ), g en erera s d en sven sk a N P ;s m o rfo lo g isk a fo rm i en ligh et m ed sp r\u00e5 k sp e cifik a regler i p r o g r a m m e ts sven sk a m o d u l. D e n o b e s t\u00e4 m d a artikeln ge n e re ra s a llts\u00e5 in te a u to m a tis k t n \u00e4 r va ria b eln D in te heir f\u00e5 tt v \u00e4 rd e t d e f u n d er e ta p p B -p r o g r a m m e t u n d e rs\u00f6k er f\u00f6 rs t, o m n om in a lfra sen s h u v u d o rd \u00e4r klas sifice ra d s o m \" u n ik t f\u00f6 r e m \u00e5 l\" (s o m so len , jo r d e n e t c ) eller s o m \u00e4 m n esn a m n , o m d e t f\u00f6 re g \u00e5 s a v in d efin ita , d eik tisk a eller p ossessiva p ro n o m e n o s v . O g ra m m a tisk a k o n s tru k tio n e r s o m * en m in bil eller * en in g en bil kan s\u00e5 led es in te gen ereras. O m v a ria b e ln D h a r b liv it in sta n tie ra d , \u00e4 r g e n e re rin g sp ro ce d u re n b e ty d h g t kortare -m a n b e h \u00f6 v e r b l a in te u n d ers\u00f6k a , o m su b sta n tiv et i fr \u00e5 g a \u00e4r en b en \u00e4 m n in g p \u00e5 e tt u n ik t f\u00f6 r e m \u00e5 l, eller o m d e t f\u00f6 re g \u00e5 s a v in d efin ita p r o n o m e n . E n h e te n R ( \" referen t n u cleu s\" ) a n v \u00e4 n d s i g e n e re rin g sp ro ce d u re n f\u00f6 r a tt kon tr o lle r a k on g ru en s in o m n om in a l& a sen . E fter \u00f6v e rs\u00e4 ttn in g sp ro ce sse n f\u00f6 r s ta fas (e t a p p A ) in n eh \u00e5 ller e n h eten R f\u00f6 lja n d e kon sta n ter: \u2022 b e ty d e ls e k o d -t e x m ( b o y , s g ) \u2022 n u m e ru s v \u00e4 rd e -h \u00e4 r, sg \u2022 d e t g e n u sv \u00e4 rd e , s o m su b sta n tiv e t h a r i k \u00e4llspr\u00e5ket \u2022 v \u00e4 r d e t + / -a n im a t \u00c5 te r s t\u00e5 e n d e p la tser i en h eten R kan sen are fy lla s m e d sy m b o le r f\u00f6 r d e g ra mm a tisk a d ra g , s o m \u00e4 r n \u00f6 d v \u00e4 n d ig a f\u00f6 r a tt \u00e5 s ta d k o m m a kon gru en s i m \u00e5lspr\u00e5ket. U n d e r g e n e re rin g sp ro ce sse n s\u00f6k er p r o g r a m m e t e fte r en sven sk lexik al en h et m ed sa m m a b e ty d e ls e k o d o c h -o m v a ria b eln D (b e s t\u00e4 m d h e t) har f\u00e5 tt ett k on stan t v \u00e4 r d e u n d e r e ta p p B -m e d sa m m a b e s t\u00e4 m d h e ts k o d s o m d en a k tu ella n om i n a lfra sen i f-re p re s e n ta tio n e n . O m t e x en n o m in a lfra s m ed h u v u d o rd e ts b e ty d e ls e k o d m ( b o y , s g ) h a r f\u00e5 tt v \u00e4 r d e t d e f e fte r referen tid en tifierin g sp ro ced u ren , v \u00e4 ljs f\u00f6 lja n d e lex ik a la en h et (o m frasen in te in n eh \u00e5 ller a ttrib u t so m uteslu ter a n v \u00e4 n d n in g a v e fte rst\u00e4 lld a rtik el): l e x (p o j k e n ,m (b o y ,s g ) .n ,d e f ,s g ,r e ,m a ,. r e --realis o c h m \u00e5 lsp r\u00e5 k ets g en u s-o c h s e x u sv \u00e4 rd e n \u00f6 v e rf\u00f6 rs till \" referen t n u cleu s\" , so m f\u00e5r f\u00f6 lja n d e fo r m : r(_,m(boy,sg),def,sg,ani,ma,re,ma). O m frasen in n eh \u00e5 ller e tt eller flera a d je k tiv a ttr ib u t, v \u00e4 ljs d en b e s t\u00e4 m d a axtikeln s r \u00e4 tt a n u m eru s-o c h g e n u sfo rm e n lig t f\u00f6 lja n d e regel: art(R,D) -> [X],R-r(_,_,_,D,N,_,_,G,_), lex(I,D,art,M.G, R e ge ln in n eb \u00e4 r, a tt k a teg orin \" a rtik el\" (a r t ) m ed b e s t\u00e4 m d h e ts v \u00e4 r d e D kan realiseras so m en o r d fo r m X o m X h a r s a m m a b e s t\u00e4 m d h e ts -, n u m eru s-(N) o c h gen u sv\u00e4 rd en (G) s o m n om in a lfra sen s h u v u d o r d . A n a lo g a re g ler a n v \u00e4 n d s fo r a tt v \u00e4 lja d e \u00f6 v r ig a k o n g ru era n d e a ttrib u te n s m o r fo lo g is k a fo rm e r. K o n g r u e n s k o n troll m e d h j\u00e4 lp a v \" referen t n u cleu s\" fu n g e ra r m y ck e t e ffe k tiv t, o c h n \u00e5 g ra fel h ar i d e tta a v see n d e in te o b se rv era ts. 3 Sammanfattning Id en tifierin g a v d isk u rsreferen ter i S W E T R A :s p ro g ra m b y g g e r p \u00e5 en fy r a -n iv \u00e5 d isk u rsm od e ll ( l : o r d o c h o rd se k v en s er, 2 :b e g r e p p b a se ra d e p \u00e5 ren t lin g v istisk kunskap, 3 :d isk u rsreferen ter -k o g n itiv a e n h e ter so m kan a ssociera s m e d flera b e g re p p p \u00e5 n iv \u00e5 2 p \u00e5 basis a v a n a ly tisk a slu tsa tser o c h ku nsk ap o m d e n ick elin gvistisk a verk lig h eten , 4: o b je k t o c h faJcta i d en v erk lig a v \u00e4 r ld e n ). D e n ak tu ella a lg o ritm e n f\u00f6 r id en tifierin g a v d isk u rsre fe n ter m e d h j\u00e4 lp a v d e n te x tu e lla in fo rm a tio n e n b efin n er sig p \u00e5 e tt e x p e rim e n te llt sta d iu m o c h b y g g e r d e ls p \u00e5 d en te x tu e lla in fo rm a tio n e n (b e ty d e ls e k o d e r o c h m e n in g a rn a s fu n k tio n e lla rep resen ta tion er la g ra d e i olik a P R O L O G -lis t o r ), dels p \u00e5 p re lim in \u00e4 ra f\u00f6 r s \u00f6 k a tt rep resen tera s te r e o ty p a b e g r e p p i le x ik o n e t (lis to r \u00f6 v e r sem a n tisk a d ra g , se m a n tisk su b k a teg oriserin g ). P r o g r a m m e t m \u00f6 jlig g \u00f6 r f\u00f6 r n \u00e4 rv a ra n d e id en tifierin g av koreferen s i k o r ta te x tfra g m e n t ( 5 -6 m e n in g a r) v id rela tio n e r a v ty p e n : k orefer en s m ellan id en tisk a b e g r e p p , k oreferen s m ella n e tt m e ra g en ere llt b e g r e p p o c h dess m era sp ecifik a a n tece d en t (t y p g ra n n e -m \u00e4 n n isk a n ), k oreferen s m ellan re feren tm \u00e4 n g d er {p o jk e o c h flic k a -b a rn en ) sa m t m ella n d era s elem en t {tv \u00e5 flyg pla n -d et en a fly g p la n e t). R efe re n tid e n tifie rin g \u00e4 r d e s s u to m m \u00f6 jlig i v issa ellip sfa ll {tv \u00e5 flyg pla n -d et e n a ), v id k oreferen s m ella n ett e m o tio n e llt la d d a t b e g re p p o c h d ess n eu tra la m o tsv a rig h e t {isra elisk a flyg p la n -lu ftp ir a te m a ), sa m t i d e fa ll, d \u00e5 k oreferen ta b e g r e p p in sta n tiera s i te x te n m ed h j\u00e4 lp a v kateg o rie llt o lik a fra ser (t e x o m disk u rsreferen ten in tr o d u c e r a s a v e tt a d je k tiv eller en hel m e n in g ). G e n e re rin g a v b e s t\u00e4 m d h e ts v \u00e4 r d e i m \u00e5 lsp r\u00e5 k et sker p rim \u00e4 rt p \u00e5 basis av k oreferen srela tion er, d \u00e4 re fte r till\u00e4 m p a s sp r\u00e5 k sp ecifik a reg ler (n om in a lfra sen s fo rm v \u00e4 ljs m ed h \u00e4 n sy n till a ttr ib u tty p o c h su b sta n tiv en s su b k a teg oriserin g i d e t svenska le x ik o n e t). P r o c e d u r e n k r\u00e4ver v id a r e e la b o r e r in g -d e sem a n tisk a rep re sen ta tion e rn a i le x ik o n e t m \u00e5 ste u tv id g a s, d e s s u to m g \u00e4 lle r d et a tt \u00e5 s ta d k o m m a e tt m era gen erellt sy stem f\u00f6 r u tfo rm n in g av lis to r m ed sem a n tisk a d ra g . D a ta b a s e n m \u00e5ste d e ssu to m berikas m ed en f\u00f6 re n k la d re p resen ta tio n a v s te r e o t y p a k u n sk aper o m d en ick e -lin g v istisk a v \u00e4 rld en (d e t t a \u00e4r en u p p g ift s o m kan f\u00f6 rv e rk lig a s en b a rt i b e g r\u00e4 n sa d u tstr\u00e4 ck n in g o c h le d e r ur le x ik o n e t in i e n c y k lo p e d ie n ). D e t finns o c k s \u00e5 e tt b e h o v a v v id a re stu d ier k rin g d e fa k to r e r so m im p lice ra r va let av b e st\u00e4 m d fo r m i svenskan i d e fall d \u00e5 in gen k oreferen srela tion f\u00f6 re lig g e r (i m \u00e5 n g a em pirisk a sven ska te x te r h ar t e x d en f\u00f6 r s ta n o m in a lfra sen b e s t\u00e4 m d fo r m ). D e n k o m p u ta tio n e lla m o d e lle n so m p resen tera ts h \u00e4r h ar a llts\u00e5 en ra d b e g r\u00e4 n sn in g a r, m en v id \u00f6 v e rs\u00e4 ttn in g a v k o r ta te x te r m ed ett b e g r\u00e4 n sa t a n ta l d isk u rsreferen ter fu n g era r d en t\u00e4 m lig en e ffe k tiv t. P r o c e d u r e n s u tv e ck lin g \u00e4 r f\u00f6 re m \u00e5 l f\u00f6 r fo r ts a tt a rb e te in o m S W E T R A ."
}
}
}
} |