| { | |
| "paper_id": "C88-1030", | |
| "header": { | |
| "generated_with": "S2ORC 1.0.0", | |
| "date_generated": "2023-01-19T12:15:55.047314Z" | |
| }, | |
| "title": "THE PARALLEL EXE~ERT PARSER (PEP): A THOROUGHLY %~IVISED DESCENDAN'~ OF THE WORD EXPERT PARSEI% (WEP)", | |
| "authors": [ | |
| { | |
| "first": "M", | |
| "middle": [], | |
| "last": "Devos", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "Katholieke Universiteit Leuven Campus Gasthuisberg Neurophysiology Lab", | |
| "institution": "", | |
| "location": { | |
| "addrLine": "Herestraat 49", | |
| "postCode": "B-3000", | |
| "settlement": "Leuven", | |
| "country": "Belgium" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "G", | |
| "middle": [], | |
| "last": "Adrxaens", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Universiteit Leuven", | |
| "location": { | |
| "addrLine": "M. Theresiastraat 21", | |
| "postCode": "B-3000", | |
| "settlement": "Leuven", | |
| "country": "Belgium" | |
| } | |
| }, | |
| "email": "" | |
| }, | |
| { | |
| "first": "Y", | |
| "middle": [ | |
| "D" | |
| ], | |
| "last": "Willems", | |
| "suffix": "", | |
| "affiliation": { | |
| "laboratory": "", | |
| "institution": "Katholieke Universiteit Leuven", | |
| "location": { | |
| "addrLine": "Celestijnenlaan 200A B", | |
| "postCode": "3030", | |
| "settlement": "Heverlee", | |
| "country": "Belgium" | |
| } | |
| }, | |
| "email": "" | |
| } | |
| ], | |
| "year": "", | |
| "venue": null, | |
| "identifiers": {}, | |
| "abstract": "", | |
| "pdf_parse": { | |
| "paper_id": "C88-1030", | |
| "_pdf_hash": "", | |
| "abstract": [], | |
| "body_text": [ | |
| { | |
| "text": "A prototype of PEP is implemented in Flat Concurrent Prolog ar{d runs in a Logix environment.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "of information is more complex tl~an a node in a connectionlst model (it may be a rule, for instance)~ but in which one attempts to keep the parallel, computation involving the items of information mere under control than can be done in a connectionist.model.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "(For examples of coarse-grain parallel NLU, see Hirakawa 1983 or Matsumoto 3.987).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "The research we present here is of the latter type of parallel NI.U. A potentially parallel NLU system (the Word Expert Parser, Small 1980) has been drastically revised so as to allow a truly parallel implementation (viz. in Flat ConCurrent Prolog, using the Logix environment (Silverman et al. 1986 )); we call the resulting system the Parallel Expert Parser (PEP, Devos 1987 ", | |
| "cite_spans": [ | |
| { | |
| "start": 216, | |
| "end": 221, | |
| "text": "(viz.", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 277, | |
| "end": 299, | |
| "text": "(Silverman et al. 1986", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 360, | |
| "end": 376, | |
| "text": "(PEP, Devos 1987", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "", | |
| "sec_num": null | |
| }, | |
| { | |
| "text": "Although WEP claimed to be \"potentially parallel\", it heavily (and implicitly) relied on sequentiality to make its principles work. (2) attribute-probing:", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "True ]~arallelism", | |
| "sec_num": "3.1" | |
| }, | |
| { | |
| "text": "Basing themselves on the attributes of the probed expert frames, experts decide which way to go in the analysis process. All attribute probing is in ~ the choose_alt command, that is described next. ( 2)The attribute-test suspends until the information to be tested is available.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "True ]~arallelism", | |
| "sec_num": "3.1" | |
| }, | |
| { | |
| "text": "There is one other command that may cause suspension of an expert, viz. begin_level (a_level).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Suspending", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "The execution of an expert that specifies begin_level(a_level), is only resumed after all attributes of incorporated expert frames are specified.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Suspending", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "This filling in of attributes takes place between different expert frames on the same level (intra-level communication).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Suspending", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "With rigid rules as to which expert fills in which frame, it is possible to prove that the expert code is deadlock free,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Suspending", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "We will further refer to these rules as the d~adlook avoidance rules.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Suspending", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "It suffices e.g. to prove that every frame that is at the lowest level that still contains unfilled frames, will eventually be filled in. It must then not be difficult to construct a deadlock analyser, that checks whether the deadlock avoidance rules are violated. This has not yet been further elaborated.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Suspending", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "However,, to ensure flexibility (especially from linguistic considerations) we are forced to allow inter-level communication, e.g. in sentences as \"the little girl loved her toy\", where \"her\" is level equivalent to \"little\", but anaphorically refers to \"the little girl\", which will probably be at a higher (hence, different) level than \"her\". ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Suspending", | |
| "sec_num": "3.4" | |
| }, | |
| { | |
| "text": "Below we present the code of some sample experts that allow the analysis of the sentence \"the little girl eats the apple\". The example is simplified, but illustrates well the crucial elements of PEp.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Example", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "First the appropriate levels and functions are declared.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Example", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "Then follows th~ code of the actual experts. Remember that expframe refers to the frame that is associated with the expezt and testframe refers to the frame that was referred to in the alternative of the preceding choose_air command. \"begin frame\" sets the appropriate level and \"refine function\" and \"refine_concept\" do the filling in of the attributes of the specified frame. the :begin frame(word level), refine_fuzkction (expframe, 'article'), reflneconcept (expframe, kled(\"defining\")), refine_concept (expframe, value(\"defined\")).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Example", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "little :begin frame(word_level), refine_funetion(expframe, 'adjective'), reflne_concept(expframe, kind(\"adjectival\")), reflne_concept(expframe, value(\"young, small\")).", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Example", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "girl :beglnframe(wordlevel), refine funotion(expframe, 'sestantive'), refine_concept(expframe, kind(\"person\"]), refine concept(expframe, valee(\"female, child or maiden\")), choose alt ((alt(frame(minus(1),function(equal(,article,))), invoke(articlelncorporation]), alt(frame(minus(1),functlon[equal('adjeetive,)) invoke(adjective_incorporation)), else(invoke(so_incorporatlon))]]. apple :-a0aloqoos to the code for girl. adJective lncorporation :incorporate(testframe), choose_alt ([alt(frame(minus(1) ,function(equal('article,))), Invoke(artlcleincorporation)), else(i~Ivoke(noincorporation)) ]).", | |
| "cite_spans": [ | |
| { | |
| "start": 480, | |
| "end": 500, | |
| "text": "([alt(frame(minus(1)", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Example", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "articleincorporatlon :incorporate(testframe), begin_frame(constltuentlevel [, reflneconcept(expframe, kind(\"unused\") ), refine coecept(expframe, value(\"unused\")).", | |
| "cite_spans": [ | |
| { | |
| "start": 75, | |
| "end": 116, | |
| "text": "[, reflneconcept(expframe, kind(\"unused\")", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Example", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "no_incorporation :begin frame(constituent level [, refine concept(expframe, klnd(\"unused\"] refine_concept(expfra~e, value(\"unused\" eats :-begin_frame(constituent level), refine_funetion(expframe, 'action'), reflneconcept(expframe, klnd(\"ingest\")), refine concept(expframe, value(\"ingest_food\"][, choose_alt ([alt(frame(plus(3),concept(view('eatable') someoneeatssomething :refine function(testframe, \"agent'), incorporate(testframe), begin frame(sentence_level), show soluhion.", | |
| "cite_spans": [ | |
| { | |
| "start": 48, | |
| "end": 90, | |
| "text": "[, refine concept(expframe, klnd(\"unused\"]", | |
| "ref_id": null | |
| }, | |
| { | |
| "start": 307, | |
| "end": 350, | |
| "text": "([alt(frame(plus(3),concept(view('eatable')", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "An Example", | |
| "sec_num": "3.5" | |
| }, | |
| { | |
| "text": "In et al. 1986 ).", | |
| "cite_spans": [ | |
| { | |
| "start": 3, | |
| "end": 14, | |
| "text": "et al. 1986", | |
| "ref_id": null | |
| } | |
| ], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "A Parallel Implementation", | |
| "sec_num": "4." | |
| }, | |
| { | |
| "text": "The prototype realization of OUr model allowing for correct analysis of very simple sentences (such as \"The man eats\", \"A man eats\", \" ", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "General Model Organization", | |
| "sec_num": "4.1" | |
| }, | |
| { | |
| "text": "The lexical-morphological analyzer schedules and invokes the experts corresponding to the elementary lexical units and outputs a blackboard, i.e. a matrix with slots whose columns correspond to those units and whose rows correspond to a level. Each expert has one expert frame associated with it; this expert frame fills one slot of the blackboard.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Data-structures: frame interconnection and blackboard information", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "In the beginning of the analysis process all frames and the blackboard contain uninstantiated slots. Experts gradually 146 instantiate the slots.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Data-structures: frame interconnection and blackboard information", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "Referring to another expert's expert frame requires walking to it over the blackboard,", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Data-structures: frame interconnection and blackboard information", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "The walk is defined in a unique way.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Data-structures: frame interconnection and blackboard information", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "All slots on the path should be instantiated, otherwise the walk suspends and waits for the instantiation. This is elegantly impiemented using the read-only unification of the parallel Prolog versions. Slots that will never be of any use any more, are instantiated to dummy constants in order not to indefinitely block suspended walks.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Data-structures: frame interconnection and blackboard information", | |
| "sec_num": "4.2" | |
| }, | |
| { | |
| "text": "In Besides polishing the implementation and writing/testing more complex experts, we also intend to look further into the necessary model of knowledge (concept) representation that has to con~lete our framework and how it can be tuned to PEP's needs. We hope that our attempt at realizing parallelism in the domain of NLU will enhance our overall understanding of the fascinating but as yet still poorly understood domain of parallel computing.", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "Conclusions and furthez res~aEch", | |
| "sec_num": "5." | |
| } | |
| ], | |
| "back_matter": [ | |
| { | |
| "text": "-", | |
| "cite_spans": [], | |
| "ref_spans": [], | |
| "eq_spans": [], | |
| "section": "ADRIAENS, G. (1986a)", | |
| "sec_num": null | |
| } | |
| ], | |
| "bib_entries": {}, | |
| "ref_entries": { | |
| "FIGREF0": { | |
| "uris": null, | |
| "num": null, | |
| "text": ")), invoke(eateomething)), else( ..................... )) ]). eatsomething :~ refine_function(testframe, 'object'), incorporate(testframe), choose alt ([alt(frame(minus(1),concept(view('person'))), invoke(someoneeatssomething)), else( . . . . . . . . . . . . . . . . . . . . . )) ]).", | |
| "type_str": "figure" | |
| }, | |
| "TABREF1": { | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table><tr><td colspan=\"2\">process</td><td colspan=\"4\">(broadcast</td><td>on</td><td>a</td><td colspan=\"4\">dedicated</td><td colspan=\"2\">signal</td></tr><tr><td colspan=\"2\">channel)</td><td>or</td><td colspan=\"6\">concepts .that</td><td colspan=\"4\">represent</td><td>the</td></tr><tr><td colspan=\"3\">meaning of</td><td colspan=\"2\">parts</td><td colspan=\"2\">of</td><td>the</td><td colspan=\"5\">linguistic</td><td>input</td></tr><tr><td colspan=\"5\">(broadcast on a</td><td colspan=\"6\">dedicated concept</td><td colspan=\"3\">channel).</td></tr><tr><td>The</td><td colspan=\"2\">experts</td><td colspan=\"5\">coordinate</td><td>the</td><td/><td colspan=\"4\">understanding</td></tr><tr><td colspan=\"2\">process</td><td>in</td><td colspan=\"3\">turn,</td><td colspan=\"5\">eventually</td><td colspan=\"3\">converging</td></tr><tr><td colspan=\"2\">towards</td><td>a</td><td/><td colspan=\"4\">conceptual</td><td/><td colspan=\"4\">structure</td><td>that</td></tr><tr><td colspan=\"14\">represents the meaning of a text fragment,</td></tr><tr><td colspan=\"2\">30 Fz'Om ~</td><td/><td colspan=\"3\">to PEP</td><td/><td/><td/><td/><td/><td/><td/></tr><tr><td/><td>In</td><td colspan=\"4\">general,</td><td>the</td><td colspan=\"2\">idea</td><td/><td colspan=\"4\">of interacting</td></tr><tr><td colspan=\"3\">processes is</td><td colspan=\"4\">a very</td><td colspan=\"4\">attractive</td><td colspan=\"3\">one if one</td></tr><tr><td colspan=\"13\">wants a flexible parser capable of using</td><td>any</td></tr><tr><td colspan=\"6\">type of information</td><td colspan=\"3\">at any</td><td colspan=\"4\">moment it</td><td>needs</td></tr><tr><td>it.</td><td>This</td><td colspan=\"3\">basic</td><td colspan=\"4\">principle</td><td>of</td><td colspan=\"4\">WEP has been</td></tr><tr><td colspan=\"5\">retained [or PEP.</td><td/><td colspan=\"6\">Yet, although the</td><td colspan=\"2\">design</td></tr><tr><td colspan=\"14\">of the sy:3tem seemed to lend itself easily to</td></tr><tr><td colspan=\"3\">a parallel</td><td colspan=\"6\">implementation,</td><td colspan=\"4\">linguistic</td><td>and</td></tr><tr><td colspan=\"6\">computational flaws</td><td>in</td><td colspan=\"2\">the</td><td colspan=\"5\">model have made</td></tr><tr><td colspan=\"14\">drastic revisions necessary before this could</td></tr><tr><td colspan=\"5\">actually ])e done.</td><td/><td/><td/><td/><td/><td/><td/><td/></tr><tr><td/><td colspan=\"5\">Io Introduction Work on understanding</td><td colspan=\"4\">parallel (NLU) is</td><td colspan=\"4\">natural only starting language to</td><td>2The 1980) is</td><td>a</td><td>Word natural Expert language Parser</td><td>(WEP, understanding Small</td></tr><tr><td/><td colspan=\"13\">emerge. (This even holds for work on any kind</td><td>program</td><td>in</td><td>the</td><td>AI</td><td>tradition</td><td>of</td><td>semantic</td></tr><tr><td/><td colspan=\"8\">of parallel AI (see e.g.</td><td/><td colspan=\"4\">Kowalik 1988)).</td><td>In</td><td>parsing (see</td><td>also</td><td>Hirst</td><td>1983v</td><td>Hahn</td><td>1986,</td></tr><tr><td/><td colspan=\"3\">general,</td><td colspan=\"2\">there</td><td colspan=\"2\">seem</td><td colspan=\"2\">to</td><td>be</td><td colspan=\"2\">two</td><td>kinds</td><td>of</td><td>Cottrell</td><td>1985,</td><td>Adriaens</td><td>1986a/b</td><td>for</td></tr><tr><td/><td colspan=\"10\">approaches to parallel NLU.</td><td colspan=\"3\">On the one hand,</td><td>WEP-inspired</td><td>or</td><td>-related</td><td>work).</td><td>The</td></tr><tr><td/><td colspan=\"2\">there</td><td>is</td><td/><td colspan=\"2\">what</td><td colspan=\"2\">we</td><td colspan=\"2\">call</td><td/><td colspan=\"2\">fine-grain</td><td>organization of</td><td>the</td><td>model</td><td>differs strongly</td></tr><tr><td/><td colspan=\"13\">parallelism; on the other hand, there is coarse-grain parallelism. With fine-grain parallel NLU we refer basically</td><td>from Rather than having a number of components that of a \"classical\" NLU system. of</td></tr><tr><td/><td>to</td><td>the</td><td/><td colspan=\"5\">connectionist</td><td/><td colspan=\"3\">approach</td><td>and</td><td>its</td><td>rules</td><td>that</td><td>are</td><td>applied</td><td>(serially)</td><td>to</td></tr><tr><td/><td colspan=\"4\">decendants.</td><td/><td colspan=\"5\">Connectionist</td><td colspan=\"3\">models</td><td>feature</td><td>linguistic input</td><td>by</td><td>a</td><td>general process, WEP</td></tr><tr><td/><td colspan=\"11\">huge networks of small nodes of</td><td colspan=\"2\">information;</td><td>considers</td><td>the</td><td>words</td><td>themselves</td><td>as</td><td>active</td></tr><tr><td/><td colspan=\"13\">computation is represented by fluctuations of</td><td>agents (word experts) that interact with each</td></tr><tr><td/><td colspan=\"13\">the (parallel) activation transmission levels inhibition along connections. of nodes of excitation and connectionism in general, see Feldman</td><td>by and (For &</td><td>other and order to find with text. Words are the meaning other knowledge of a implemented as i.e. processes that run for</td><td>sources in fraQT~ent of coroutines~ a while</td></tr><tr><td/><td colspan=\"2\">Ballard</td><td colspan=\"3\">1982,</td><td colspan=\"3\">VanLehn</td><td colspan=\"3\">1984,</td><td colspan=\"2\">Hillis</td><td>1986,</td><td>(broadcasting</td><td>information</td><td>or</td><td>performing</td></tr><tr><td/><td colspan=\"3\">McClelland</td><td/><td>&</td><td/><td colspan=\"4\">Rumelhart</td><td/><td colspan=\"2\">1986;</td><td>for</td><td>side-effect</td><td>operations</td><td>to</td><td>refine</td><td>the</td></tr><tr><td/><td colspan=\"12\">connectionist models of NLU, see</td><td>Cottrell &</td><td>representation</td><td>of</td><td>the</td><td>meaning</td><td>of</td><td>a</td><td>text</td></tr><tr><td/><td colspan=\"4\">Small 1983,</td><td colspan=\"3\">Cottrell</td><td colspan=\"3\">1985,</td><td colspan=\"3\">Pollack & Waltz</td><td>fra~nent), and suspend when they have to wait</td></tr><tr><td/><td colspan=\"10\">1985, McClelland & ~melhart</td><td colspan=\"3\">1986) o</td><td>With</td><td>for information</td><td>from</td><td>other</td><td>experts.</td><td>The</td></tr><tr><td/><td colspan=\"9\">coarse-grain parallel NL~</td><td colspan=\"4\">we refer</td><td>to a</td><td>information they send or wait for are</td><td>either</td></tr><tr><td/><td colspan=\"13\">more modest kind, in which the smallest</td><td>item</td><td>signals relating to the status of the parsing</td></tr></table>", | |
| "text": ").", | |
| "num": null | |
| }, | |
| "TABREF2": { | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table><tr><td/><td colspan=\"3\">experts</td><td colspan=\"4\">relatively</td><td colspan=\"3\">small</td><td colspan=\"2\">(hence,</td><td>more</td><td>computational</td><td>viewpoint,</td><td>squeezing</td><td>all</td></tr><tr><td/><td colspan=\"12\">readable) and to figure out more easily where</td><td>interactions</td><td>into</td><td>words</td><td>makes</td><td>it</td><td>almost</td></tr><tr><td/><td colspan=\"12\">things could go wrong in the parsing process.</td><td>impossible to figure out what is going on</td><td>in</td></tr><tr><td/><td colspan=\"4\">A possible</td><td colspan=\"3\">hierarchy</td><td colspan=\"2\">of</td><td colspan=\"2\">levels</td><td>might</td><td>be:</td><td>the overall</td><td>parsing</td><td>process.</td><td>Words have to</td></tr><tr><td/><td colspan=\"12\">morpheme, word, constituent, clause, sentence</td><td>decide</td><td>on</td><td>everything,</td><td>from</td><td>morphological</td></tr><tr><td/><td colspan=\"2\">(each</td><td colspan=\"3\">level</td><td colspan=\"3\">having</td><td colspan=\"2\">its</td><td colspan=\"2\">own</td><td>function</td><td>issues</td><td>to</td><td>pra~natic</td><td>issues,</td><td>with</td><td>jammed</td></tr><tr><td/><td colspan=\"4\">attributes).</td><td colspan=\"8\">In the somewhat oversimplified</td><td>communication channels as a result.</td></tr><tr><td/><td colspan=\"7\">example below we will</td><td colspan=\"5\">be using three</td><td>levels</td><td>In</td><td>PEP,</td><td>experts</td><td>are</td><td>associated</td><td>with</td></tr><tr><td/><td colspan=\"6\">(between brackets:</td><td colspan=\"2\">the</td><td colspan=\"4\">respective</td><td>function</td><td>concepts rather than with words,</td><td>it is very</td></tr><tr><td/><td colspan=\"4\">attributes),</td><td/><td/><td/><td colspan=\"2\">viz.</td><td/><td/><td>word level</td><td>natural to</td><td>do</td><td>so:</td><td>words</td><td>are only used to</td></tr><tr><td/><td colspan=\"11\">[article;adjective;substantive],</td><td>evoke</td><td>the concepts</td><td>that</td><td>conshitute</td><td>the</td></tr><tr><td/><td colspan=\"12\">constituent_level [action;agent;object],</td><td>and</td><td>meaning of</td><td>a</td><td>fra~lent</td><td>of</td><td>text.</td><td>St:i.l i ~</td></tr><tr><td/><td colspan=\"5\">sentence level.</td><td/><td/><td/><td/><td/><td/><td>concepts have a</td><td>concrete link</td><td>to words</td><td>and</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>can be regarded as being associated with</td><td>hhe</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>group of words</td><td>that evokes</td><td>them.</td><td>E.g.</td><td>in</td></tr><tr><td/><td colspan=\"7\">3.3 Broadcasting vs.</td><td/><td/><td/><td/><td>\"the</td><td>young</td><td>girl\"</td><td>three</td><td>concepts</td><td>can</td><td>be</td></tr><tr><td/><td/><td colspan=\"8\">explicit communication</td><td/><td/></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>discovered,</td><td>associated</td><td>with</td><td>the</td><td>b~sic</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>word-groups \"the\", \"young\" and \"girl ~'. At: a</td></tr><tr><td/><td colspan=\"5\">Experts are</td><td colspan=\"4\">the active</td><td colspan=\"3\">components of</td><td>the</td><td>higher level a compound concept</td><td>constituting</td></tr><tr><td/><td colspan=\"6\">analysis system.</td><td/><td>New</td><td colspan=\"4\">concepts</td><td>come into</td><td>the meaning</td><td>of</td><td>the</td><td>entire</td><td>construc~</td><td>\"the</td></tr><tr><td/><td colspan=\"3\">existence</td><td colspan=\"2\">only</td><td colspan=\"3\">through</td><td colspan=\"3\">their</td><td>interaction.</td><td>young glrl\" is invoked.</td></tr><tr><td/><td colspan=\"7\">Since parallelism was</td><td colspan=\"3\">a major</td><td colspan=\"2\">goal we</td><td>have</td><td>Concretely,</td><td>in</td><td>PEP</td><td>a</td><td>specific</td><td>data</td></tr><tr><td/><td colspan=\"12\">based our communication protocols on explicit</td><td>structure (the expert ~rame) is</td><td>associ~ted</td></tr><tr><td/><td colspan=\"12\">identification of the expert frames</td><td>involved</td><td>with</td><td>every</td><td>expert.</td><td>The</td><td>hierarchy</td><td>that</td></tr><tr><td/><td colspan=\"12\">in some interaction, which allows us to</td><td>keep</td><td>originates from the concepts is reflected</td><td>by</td></tr><tr><td/><td colspan=\"10\">communication under control.</td><td colspan=\"2\">Two kinds</td><td>of</td><td>the interconnection</td><td>of</td><td>the</td><td>expert</td><td>fraracs.</td></tr><tr><td/><td colspan=\"9\">communication take place:</td><td/><td/><td>These</td><td>are</td><td>vertically</td><td>related</td><td>by</td><td>].eve].</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>interdependencies, and</td><td>horizonta]ly</td><td>by</td><td>the</td></tr><tr><td/><td colspan=\"8\">(I) attribute-refining:</td><td/><td/><td/><td>relative role the concepts of the frames play in the frame that is being built out of them</td></tr><tr><td/><td/><td/><td colspan=\"3\">Experts</td><td colspan=\"2\">are</td><td colspan=\"3\">allowed</td><td>to</td><td>refine</td><td>the</td><td>one level</td><td>higher.</td><td>Besides</td><td>its</td><td>level~ a~</td></tr><tr><td colspan=\"13\">Especially for attributes of ~xpert frames. the restarting experts, a last-in first-out regime of suspended The attributes are considered to be information that is (stack) took care of contention for messages: the accessible by all experts.</td><td>expert frame function attribute (stating what the role is has three attribute slots: level), a concept attribute (representing the expert concept plays at a spec.i.fJc a</td></tr><tr><td colspan=\"2\">expert</td><td colspan=\"2\">that</td><td colspan=\"2\">placed</td><td colspan=\"2\">an</td><td colspan=\"4\">expectation</td><td>for</td><td>a</td><td>the contents of</td><td>the expert)</td><td>and a</td><td>lexical</td></tr><tr><td colspan=\"8\">message 16.st~ mostly got</td><td colspan=\"5\">it first. Also,</td><td>to</td><td>attribute (simply correspondin~ to the</td><td>gr<~u~></td></tr><tr><td colspan=\"9\">avoid com~lications in expert</td><td colspan=\"4\">communication,</td><td>of</td><td>words</td><td>associated</td><td>with</td><td>Lhe</td><td>conc<~pt) .</td></tr><tr><td colspan=\"2\">no new</td><td colspan=\"3\">e~perts</td><td colspan=\"2\">were</td><td colspan=\"6\">initialized before the</td><td>Below, we will see that this definition of e:~</td></tr><tr><td colspan=\"13\">queue of leady-to-run experts was empty.</td><td>The</td><td>expert frame</td><td>is</td><td>crucial</td><td>for the rest~:/<:!t~:!<l</td></tr><tr><td colspan=\"5\">adherence to this</td><td colspan=\"8\">sequentialization, not</td><td>to</td><td>cormmunication protocol among experts.</td></tr><tr><td colspan=\"11\">mention the side-effects involved,</td><td colspan=\"2\">obviously</td><td>The</td><td>\"analysis</td><td>process\"</td><td>consists</td><td>of the</td></tr><tr><td>made</td><td colspan=\"2\">WEP's</td><td colspan=\"3\">claim</td><td>of</td><td colspan=\"3\">being</td><td colspan=\"3\">\"potentially</td><td>collection of currently</td><td>active experts</td><td>tha[+</td></tr><tr><td colspan=\"6\">parallel\" invalid.</td><td/><td/><td/><td/><td/><td/><td>try to establish</td><td>new concepts.</td><td>~f a</td><td>ne~</td></tr><tr><td/><td/><td colspan=\"2\">Ir~ a</td><td colspan=\"2\">truly</td><td colspan=\"4\">parallel</td><td colspan=\"3\">environment,</td><td>concept</td><td>can</td><td>successfully</td><td>be</td><td>formed,</td><td>the</td></tr><tr><td colspan=\"5\">sequentiality can</td><td colspan=\"2\">no</td><td colspan=\"2\">longer</td><td colspan=\"2\">be</td><td colspan=\"2\">relied on.</td><td>corresponding expert is added to the analysis</td></tr><tr><td colspan=\"13\">PEP uses [parallelism whenever possible:</td><td>for</td><td>process, while the combined concepl r s expe~t~</td></tr><tr><td>the</td><td colspan=\"3\">execution</td><td/><td>of</td><td colspan=\"3\">expert</td><td colspan=\"2\">code</td><td colspan=\"2\">AND</td><td>for</td><td>may die.</td><td>They pass their expert frames~</td><td>a~0</td></tr><tr><td colspan=\"13\">initializing new experts (initializing all of</td><td>so the</td><td>contained</td><td>information,</td><td>to</td><td>the</td><td>new</td></tr><tr><td>them</td><td>as</td><td/><td colspan=\"2\">soon</td><td>as</td><td/><td colspan=\"2\">they</td><td colspan=\"2\">are</td><td colspan=\"2\">read</td><td>and</td><td>expert, which will</td><td>usually incorporate</td><td>them</td></tr><tr><td colspan=\"5\">morphologically</td><td colspan=\"4\">analyzed).</td><td colspan=\"4\">In order</td><td>to</td><td>in its own expert</td><td>frame.</td><td>Notice that</td><td>hh~</td></tr><tr><td colspan=\"4\">realize this,</td><td colspan=\"2\">the</td><td colspan=\"2\">most</td><td colspan=\"5\">important departure</td><td>view</td><td>has</td><td>interesting</td><td>software</td><td>engineering</td></tr><tr><td colspan=\"7\">from the original model</td><td colspan=\"6\">is that experts</td><td>are</td><td>aspects not</td><td>present</td><td>in</td><td>WEP:</td><td>by having a</td></tr><tr><td colspan=\"3\">no longer</td><td colspan=\"2\">only</td><td colspan=\"4\">associated</td><td colspan=\"2\">with</td><td colspan=\"2\">words (the</td><td>leveled approach</td><td>expert</td><td>code</td><td>becomes</td><td>more</td></tr><tr><td>only</td><td colspan=\"4\">linguistic</td><td colspan=\"3\">entities</td><td colspan=\"5\">acknowledged</td><td>by</td><td>local, modular</td><td>and</td><td>adaptable.</td><td>The dynamic</td></tr><tr><td colspan=\"2\">WEP) \u00b0</td><td colspan=\"11\">We will now discuss what experts</td><td>are</td><td>process hierarchy enables the linguist/expert</td></tr><tr><td colspan=\"5\">associated with,</td><td colspan=\"2\">and</td><td>how</td><td colspan=\"2\">the</td><td colspan=\"2\">new</td><td>view of</td><td>writer to write generic</td><td>experts that can</td><td>be</td></tr><tr><td colspan=\"4\">experts leads</td><td>to</td><td colspan=\"3\">clearer</td><td colspan=\"5\">and more explicit</td><td>parameterized with the</td><td>value of</td><td>the concept</td></tr><tr><td colspan=\"3\">concepts o~</td><td colspan=\"3\">waiting</td><td colspan=\"2\">and</td><td colspan=\"5\">communicating in a</td><td>they</td><td>represent</td><td>(cp.</td><td>object-oriented</td></tr><tr><td colspan=\"7\">parallel environment.</td><td/><td/><td/><td/><td/><td>programming).</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>A final note about the levels.</td><td>Each level</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>is intended</td><td>to</td><td>deal</td><td>with</td><td>a more</td><td>or le~;s</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>independent</td><td>part</td><td>in</td><td>the</td><td>derivation</td><td>and</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>composition of meaning.</td><td>However, we leave i[:</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>up</td><td>to</td><td>the</td><td>linguist</td><td>writing</td><td>the</td><td>expe~-L</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td/><td>processes to declare (i) what levels he wants</td></tr><tr><td colspan=\"4\">A major item</td><td colspan=\"8\">of criticism uttered</td><td>against</td><td>to consider</td><td>and</td><td>(2)</td><td>what</td><td>the</td><td>appropriate</td></tr><tr><td colspan=\"4\">WEP has been</td><td colspan=\"3\">that it</td><td colspan=\"5\">considers the</td><td>word as</td><td>functions are</td><td>that-he</td><td>wants</td><td>to use at the</td></tr><tr><td colspan=\"6\">the only entity to</td><td colspan=\"5\">be turned into</td><td colspan=\"2\">an expert</td><td>respective</td><td>levels.</td><td>By</td><td>combining</td><td>this</td></tr><tr><td colspan=\"13\">process. existence undeniable and must Linguistically of larger be taken into account, speaking, the constituents is whatever ]aodel one advocates. From the</td><td>flexible filling in model, we force specify the experts and help him to keep of a rigorously defined the linguist to clearly the</td></tr></table>", | |
| "text": "", | |
| "num": null | |
| }, | |
| "TABREF3": { | |
| "type_str": "table", | |
| "html": null, | |
| "content": "<table><tr><td/><td/><td/><td/><td/><td/><td/><td>alternatives</td><td>succeed,</td><td>an</td><td>arbitrary</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>corresponding expert is invoked, whereas</td><td>the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>others are not further considered (don't-care</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>committed choice;</td><td>see</td><td>also</td><td>below and Devos</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>1987, however,</td><td>for</td><td>a</td><td>suggestion</td><td>of how to</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>realize non-determinism in view of</td><td>possible</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>ambiguity).</td><td>Only</td><td>after</td><td>failure</td><td>of</td><td>all</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>tests</td><td>is</td><td>the</td><td>elsative-expert</td><td>executed.</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>Tests consist of</td><td>a frame-</td><td>specification and</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>an</td><td>attribute-condltion.</td><td>The</td><td>latter</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>constitutes the actual test on the</td><td>attribute</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>of</td><td>the</td><td>frame</td><td>selected</td><td>by</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>\"frame-specification\".</td><td>This</td><td>frame</td><td>can</td><td>be</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>referred</td><td>to</td><td>with</td><td>testframe</td><td>in</td><td>the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>corresponding invoked expert.</td><td>One</td><td>will</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>already</td><td>have</td><td>noticed</td><td>tha~ the</td><td>choose alt</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>predicate</td><td>does</td><td>not</td><td>contain</td><td>any</td><td>expl~cit</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>scheduling commands.</td><td>Indeed, we intend</td><td>to</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>entirely</td><td>mask</td><td>the</td><td>program</td><td>flow</td><td>by</td><td>a</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>declarative reading.</td><td>However, flow control</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>remains</td><td>necessary</td><td>and</td><td>St</td><td>is</td><td>realized</td><td>by</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>suspending a n expert rou\u00a2ine (or a branch</td><td>in</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>the</td><td>choose alt</td><td>command,</td><td>since</td><td>the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>alternatives</td><td>in</td><td>the</td><td>choese_alt</td><td>may</td><td>be</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>executed</td><td>in</td><td>parallel),</td><td>if</td><td>it</td><td>requires</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>information that is not yet available.</td><td>Only</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>after this required information is filled in,</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>does the</td><td>expert-routine</td><td>resume.</td><td>This can</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>cheaply</td><td>be</td><td>implemented</td><td>using</td><td>read-only</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>unification</td><td>(Shapiro</td><td>1986).</td><td>Intuitively,</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>predicates</td><td>that</td><td>probe</td><td>for</td><td>information</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>suspend, if the</td><td>variable that</td><td>supplies this</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>information is not</td><td>yet instantiated.</td><td>This</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>suspension takes place during unification</td><td>of</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>the Flat</td><td>Concurrent</td><td>Prolog</td><td>(FCP) predicate</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>(see below), into</td><td>which expert</td><td>routines are</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>compiled.</td><td>Resumption occurs whenever</td><td>the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>required</td><td>variable</td><td>gets</td><td>instantiated.</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>Suspension of</td><td>a</td><td>choose_alt</td><td>branch may take</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>place in the following cases:</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>(i) If the search for the testframe requires</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>information that</td><td>is</td><td>not</td><td>yet</td><td>available, it</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>simply</td><td>suspends.</td><td>As</td><td>a</td><td>result</td><td>the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>frame-specification</td><td>always</td><td>leads</td><td>to</td><td>the</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>selection of a frame in a deterministic</td><td>way.</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>Hence,</td><td>explicit</td><td>communication</td><td>becomes</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>possible.</td></tr><tr><td/><td>Let</td><td>us</td><td>now</td><td colspan=\"2\">turn</td><td colspan=\"2\">our attention to the</td></tr><tr><td colspan=\"8\">command that allows\u2022 experts to</td><td>decide which</td></tr><tr><td colspan=\"2\">way to go</td><td colspan=\"3\">in their</td><td colspan=\"3\">analysis process</td><td>on the</td></tr><tr><td colspan=\"5\">basis of information</td><td colspan=\"3\">they expect</td><td>from other</td></tr><tr><td colspan=\"2\">experts.</td><td>We</td><td colspan=\"2\">have</td><td>in</td><td colspan=\"2\">fact</td><td>localized</td><td>all</td></tr><tr><td colspan=\"8\">possible choice-points in one command:</td></tr><tr><td colspan=\"3\">choose_alt([</td><td/><td/><td/><td/></tr><tr><td/><td/><td colspan=\"6\">alt(frame(frame-specification,</td></tr><tr><td/><td/><td/><td/><td/><td colspan=\"3\">attrlbuteconditlon),</td></tr><tr><td/><td/><td/><td/><td/><td/><td/><td>invoke(expert)),</td></tr><tr><td/><td/><td colspan=\"6\">alt(frame(frame-specification,</td></tr><tr><td/><td/><td/><td/><td/><td colspan=\"3\">attribute condition),</td></tr><tr><td/><td/><td/><td/><td/><td colspan=\"3\">invoke(expert)),</td></tr><tr><td/><td/><td colspan=\"6\">alse(invoke(e~q~azt))</td></tr><tr><td/><td/><td>]).</td><td/><td/><td/><td/></tr><tr><td colspan=\"4\">It consists of a</td><td colspan=\"4\">number of alternatives</td><td>and</td></tr><tr><td>an</td><td colspan=\"2\">optional</td><td colspan=\"3\">elsative.</td><td/><td>The</td><td>alternatives</td></tr><tr><td colspan=\"4\">contain a test,</td><td colspan=\"3\">which may</td><td>fail, suspend</td><td>or</td></tr><tr><td colspan=\"2\">succeed.</td><td colspan=\"6\">In the last case the corresponding</td></tr><tr><td colspan=\"6\">expert may be invoked.</td><td colspan=\"2\">If tests from several</td></tr><tr><td>14&</td><td/><td/><td/><td/><td/><td/></tr></table>", | |
| "text": "", | |
| "num": null | |
| } | |
| } | |
| } | |
| } |