Tryag File Manager
Home
-
Turbo Force
Current Path :
/
home
/
cluster1
/
data
/
bu01
/
1121861
/
html
/
mixtec
/
Programas
/
Shoebox
/
Doc
/
Upload File :
New :
File
Dir
/home/cluster1/data/bu01/1121861/html/mixtec/Programas/Shoebox/Doc/ShParse.doc
ࡱ > ' ( ! " # $ % & ) * + , - . / 0 1 2 3 4 5 6 7 8 9 : ; < = > ? R o o t E n t r y F M C o m p O b j n W o r d D o c u m e n t >s O b j e c t P o o l HM HM F Microsoft Word 6.0 Document MSWordDoc Word.Document.6 9q _ Oh +'0 $ H l D h * D:\MSOFFICE\WINWORD\TEMPLATE\ACPUBPNG.DOT Parsing with Shoebox David Bevan Mark R. Pedrotti @ ܥe = e A >s x> p V V g g g g g $ h h dm dm dm m m " h r d m m : m m m m : 0n $ Tn &p (p (p (p , Tp (q q ~r T r l r g hn # & m m hn hn r hn g g m m hn hn hn hn g m g m &p g 2 h g g g g hn &p hn hn Parsing with Shoebox David Bevan Introduction This document describes how to structure your lexical database for parsing with Shoebox for Windows and Macintosh. In Shoebox2 for MS DOS, there were two methods of providing parsing information: a parsing database and the conjoined affix parser. With Shoebox5, neither of these are necessary because the program has a morphological parser which can break off any number of affixes and handle morphophonemic changes to affixes and roots. If you desire, all the necessary information for parsing can be kept in your lexical database. This document uses English examples with the standard Shoebox interlinear settings and Multi-Dictionary Formatter (MDF) data field markers. Simple Parsing: Roots and Affixes With Shoebox, it is necessary in the lexicon to specify whether a morpheme is a root, prefix, suffix or infix by the correct use of morpheme break characters (normally hyphens) in the lexeme field. A prefix must have a final hyphen; a suffix must have an initial hyphen; a root must not have a hyphen; an infix must have an initial and a final hyphen. Note that this means that you cannot have hyphens on bound roots in the lexeme field. However, you can enter roots with hyphens in the \lc Lexical Citation data field. This is all thats necessary for simple parsing. For example, the lexical entries \lx un- \lx success \lx -ful \ps neg \ps n \ps nadjzr \ge OPPOS \ge achievement \ge ADJZR will cause unsuccessful to be interlinearized correctly: unsuccessful un- success -ful OPPOS- achievement -ADJZR neg- n -nadjzr Forms with hyphens between the morphemes are also parsed correctly. For example, with \lx non- \lx read \lx -er \ps neg \ps v \ps vnomzr \ge not \ge look_at_book \ge AGENT in the lexicon, non-reader is parsed as follows: non-reader non- read -er not- look_at_book -AGENT neg- v -vnomzr Note how great an improvement this is over Shoebox 2 where every word had to be parsed individually, or every combination of affixes listed in the conjoined affix parser. Infixes An infix is a morpheme embedded within a root or an affix. Some people use the term infix to refer to an affix which occurs only between a root and another affix; but for Shoebox that is just another prefix or suffix, not an infix. English doesnt have infixes, but if \lx -int- \ps intens \ge INTENS was an English intensive infix, then sintuccessful would be interlinearized correctly as sintuccessful success -int- -ful achievement -INTENS- -ADJZR n -intens- -nadjzr In the parse, infixes appear before or after the root or stem in which they are found. (Shoebox has an option which lets you choose.) These examples display the infix after. Infixes can be found anywhere in the word, so successfintul would be interlinearized as follows: successfintul success -ful -int- achievement -ADJZR -INTENS- n -nadjzr -intens- Compound Roots Words with compound roots will be parsed correctly if both roots are in the lexicon. For example, the following lexical entries \lx black \lx bird \lx -s \ps adj \ps n \ps ninfl \ge dark \ge flying_creature \ge PL will cause blackbirds to be interlinearized as follows: blackbirds black - bird -s dark - flying_creature -PL adj - n -ninfl Compounds with hyphens, such as sea-green and mother-in-law parse correctly. Shoebox has an option to allow or disallow compound roots. Alternate Forms If a morpheme has more than one surface form, this can be specified with an alternate form (\a) field: \lx a \lx telephone \a an \a phone \ps art \ps n \ge INDEF \ge transceiver The parsed form is taken from the \lx field: an enormous phone a enormous telephone INDEF very_big transceiver art adj n Here are lexical entries for some more morphemes with alternate forms: \lx in- \lx -s \lx not \a im- \a -es \a -nt \a il- \ps ninfl \ps neg \a ir- \ge PL \ge NEG \ps neg \ge OPPOS and some example interlinear text: impossible foxes faces haven't in- possible fox -s face -s have -not OPPOS- feasible animal_sp. -PL head -PL own NEG neg- adj n -ninfl n -ninfl v neg Note that foxes and faces are parsed correctly due to the presence of fox and face in the lexicon (but not foxe and fac). Note also that havent has been analyzed here as a compound. Alternative forms can have a different (affix) type from the lexeme (e.g. you can have a suffix as an alternative form of a root or prefix). Underlying Forms If you require that a variant form has its own lexical entry, you can use an underlying form (\u) field. Heres an alternative way of treating phone. The parsed output is the same as above. \lx phone \lx telephone \u telephone \ps n \ge transceiver You can enter the morphemic breakdown of a lexical entry in an underlying form field. (This is like the use of the parsing database in Shoebox 2.) For example, here is one way to handle the suppletive verb form went. \lx went \lx go \lx -ed \u go -ed \ps v \ps vinfl \ge proceed \ge PAST Heres how it parses: he went he go -ed 3SM proceed -PAST pron v -vinfl Rather than having a separate lexical entry for each irregular form, these forms may be included in the main entry by using an alternate form field and an underlying form field. Heres another way of handling went along with some other irregular verb forms: \lx go \lx find \lx hit \a went \a found \u hit \u go -ed \u find -ed \u hit -ed \ps v \ps v \ps v \ge proceed \ge locate \ge strike An underlying form (\u) field is associated with whichever \lx or \a field precedes it. (If no \u field follows a \lx or \a field, the contents of the \lx field are used as the underlying form.) Note: When you want to identify portions of underlying forms that are to be parsed as whole units, you must leave spaces between the morphemes in an underlying form field to distinguish roots, prefixes and suffixes. But if you want the operation to return a portion to the parser to add to the main block being processed, no spaces are allowed. Note that hit is ambiguous and so two underlying forms are given to make Shoebox display an Ambiguity Selection dialog box. Heres how the verbs parse (with the past tense form chosen for hit). went found hit go -ed find -ed hit -ed proceed -PAST locate -PAST strike -PAST v -vinfl v -vinfl v -vinfl Note that it doesnt matter to Shoebox whether a form that needs parsing information is included in a main lexical entry as an alternate form or has its own lexical entry (perhaps in a separate database). The choice is yours. Sometimes, it may be necessary to use an underlying form field for affix sequences rather like using the conjoined affix parser in Shoebox 2. The affix -ability is an example in English: \lx -able \lx -ity \lx -ability \ps vadjzr \ps anomzr \u -able -ity \ge ABIL \ge NOMZR Heres how readability is parsed: readability read -able -ity look_at_book -ABIL -NOMZR v -vadjzr -anomzr Finally, here are some examples of underlying forms of compounds: \lx have \lx brunch \a Ive \u breakfast lunch \u I have \ps v \ge own which parse like this: I've brunch I have breakfast lunch 1S own morning_meal noon_meal pron v n n Forced Values Resolving Ambiguity Where a morpheme has more than one meaning, it is sometimes helpful in underlying forms to specify which one is required so that Shoebox doesnt display an Ambiguity Selection dialog box. For example, two English suffixes have the form -s, so in the underlying form of the irregular form men, it is worthwhile specifying which one is required. This is done by putting the gloss in curly braces after the morpheme: \lx -s \lx -s \lx man \a -es \a -es \a men \ps ninfl \ps vinfl \u man -s{PL} \ge PL \ge 3S \ps n \ge male_person Then men parses as follows without displaying an Ambiguity Selection dialog box. men man -s male_person -PL n -ninfl For forced glosses to work in this way it is necessary that in the parsed output the gloss line appears immediately after the parsed (morpheme breakdown) line. If the part of speech line was above the gloss line, then the forced value would have to be specified as -s{ninfl} or -s{ninfl}{PL}. Note that multiple forced values are possible in the order of the interlinear lines. Heres another example, with the ambiguity on the root: \lx bear \lx bear \ps n \a bore \ge animal_sp. \u bear{carry} -ed \ps v \ge carry With these lexical entries, bore parses like this without displaying an Ambiguity Selection dialog box. (Of course, bore is itself ambiguous, also meaning drill a hole.) bore bear -ed carry -PAST v -vinfl Forced glosses on prefixes must be placed after the hyphen, not before. Direct Parsing Preventing Incorrect Parses With the following lexical entries, \lx hop \lx hope \lx -s \ps v ; n \ps v \a -es \ge jump \ge expect \ps vinfl \ge 3S the word hopes will be parsed incorrectly: hopes *hop -s *jump -3S v -vinfl In parsing, Shoebox resolves much of the possible ambiguity by choosing the parse which cuts off the longest affix. This prevents the user being presented with an Ambiguity Selection dialog box with many possible parses mostly incorrect to choose from. In the case of hopes, this means that -es is cut off. In situations like this, to get the correct parse it is necessary add parsing information to over-ride the incorrect analysis (an alternative approach is outlined later): \lx hope \a hopes \u hope -s{3S} \ps v \ge expect This forces hopes to be parsed like this: hopes hope -s expect -3S v -vinfl Heres another example involving prefixes. To prevent the false parse of the word demisting in which the prefix demi- gets cut off, demisting *demi- sting *half- hurt *num- n it is necessary to provide an analysis of demist: \lx demist \lx mist \u de- mist{fog}{v} \ps n ; v \ge fog Then the word is correctly parsed: demisting de- mist -ing REVERS- fog -PTC neg- v -vinfl In some cases there is more than one correct parse of a word (or part of word) with morpheme breaks in different places. In this case, it is necessary to provide parsing information for both so that the ambiguity is recognized. Heres an example: \lx do \lx doe \a does \a does \u do -s{3S} \u doe -s{PL} \ps v \ps n \ge perform \ge female_deer Shoebox will display an Ambiguity Selection dialog box for you to choose which analysis you want: does does do -s doe -s perform -3S female_deer -PL v -vinfl n -ninfl Morphophonemics Simple morphophonemic alternations can be expressed by using alternate forms and underlying forms. Here is how the orthographic rules y ( i / __ +ed or y + ed ( ied y ( ie / __ +s or y + s ( ies are expressed: \lx -ed \x -s \a -d \a -es \a -ied \a -ies \u y+ed \u y+s \ps vinfl \ps vinfl \ge PAST \ge 3S The way this works is that the alternate form field contains the surface form including the whole of the suffix. The underlying form field contains the underlying form of the part of the root (or preceding suffix) that is modified followed by + and the underlying form of the suffix. Heres how forms of the verbs try and tie are parsed: tried tied tries ties try -ed tie -ed try -s tie -s attempt -PAST bind -PAST attempt -3S bind -3S v -vinfl v -vinfl v -vinfl v -vinfl Heres another example, for the doubling of consonants before the -ed suffix: \lx -ed \a -d \a -pped \u p+ed \ps vinfl \ge PAST and heres how hopped parses: hopped hop -ed jump -PAST v -vinfl At present, there is no way of specifying general rules, so this parsing information would have to be provided for each suffix and each consonant that gets doubled. The approach is equivalent for prefixes. There are very few processes involved in English prefixes, but if we suppose that the rule sp ( p/ dis+ __ or dis + sp ( disp as exemplified by the word dispirited was productive, then we would specify it like this: \lx dis- \a disp- \u dis+sp \ps neg \ge OPPOS and dispirited would be parsed as follows: dispirited dis- spirit -ed OPPOS- vitality -possessing neg- n -nadjzr Context Sensitivity Preventing Incorrect Parses Even when there is no morphophonemic process involved, it may still be beneficial to use the morphophonemic notation if an affixs occurrence is phonologically conditioned in order to reduce the number of false parses. For example, with the allomorphs of the prefix in- specified as follows, \lx in- \a im- \a il- \a ir- \ps neg \ge OPPOS words like image, imam, iris and iron will be parsed incorrectly if they dont already have their own lexical entry. For example: iron *in- on *OPPOS- upon *neg- prep If we use the morphophonemic notation to express the restriction on their occurrence as follows: \lx in- \a imb- \u in+b \a imm- \u in+m \a imp- \u in+p \a ill- \u in+l \a irr- \u in+r \ps neg \ge OPPOS then the false parses wont occur. Ensuring Ambiguity is Recognized The morphophonemic notation can also be used to achieve the opposite effect of increasing the number of parses. For example, the following specifies that there are two possible parses of word-final -es. Either it is simply -s (as in pushes) or (the additional information) it is e+s (with roots with final -e, as in likes): \lx -s \a -es \a -es \u e+s \ps vinfl \ge 3S In cases (such as hopes) where roots exist both with and without a final -e (hope and hop), as far as Shoebox is concerned the parse is ambiguous, but without specifying the above (apparently redundant) information Shoebox will always choose the one which cuts of the longer affix. Adding the extra parsing choice makes Shoebox display the Ambiguity Selection dialog box offering a choice (between hop -s and hope -s). The disadvantage of this approach is that the (false) ambiguity is always displayed unless for each form you add specific parsing information. On the other hand, for forms like does which really is ambiguous, it makes it unnecessary to provide specific parsing information for either parse. Reduplication Simple reduplicative processes can be represented in Shoebox. The following example specifies a reduplication of from one to three consonants followed by a vowel at the beginning of the reduplicated root, stem or word. For Shoebox to recognize that this is an entry for reduplication, the \lx field must contain the letters dup somewhere. (It may be best for these to occur at the beginning of the field so that all entries for reduplication sort together.) The \a fields are used for specifying the pattern to match using variables defined in the language encoding properties. \lx dupCV- \a [cons][vowel]- \a [cons][cons][vowel]- \a [cons][cons][cons][vowel]- \ps intens \ge very English doesnt make regular use of reduplication, but if the process above was English, then the following would be correct parses: bibig strostrong blablack dupCV- big dupCV- strong dupCV- black very- large very- powerful very- dark intens- adj intens- adj intens- adj Reduplicative suffixation can also be specified, as can reduplication with fixed letters. For example, the following specifies a reduplication of the final consonant cluster with an intermediate i. \lx -dupiC \a -i[cons] \a -i[cons][cons] \a -i[cons][cons][cons] \ps dimin \ge a_bit Heres what the parses would look like: bigig stronging blackick big -dupiC strong -dupiC black -dupiC large -a_bit powerful -a_bit dark -a_bit adj -dimin adj -dimin adj -dimin It is possible to specify the type of reduplication in English which copies a syllable, replacing the vowel usually with i as in tip-top, tick-tock, criss-cross, flip-flop and wishy-washy. The entry would be as follows: \lx dupCiC \a -[cons]i[cons] \a -[cons]i[cons][cons] \a -[cons][cons]i[cons] \ps redup \ge intens Reduplication of a whole unit can be specified as follows: \lx dup \a [...] \ps redup \ge informal By adding hyphens, you can also specify prefix full reduplication or suffix full reduplication. Heres an English example of full reduplication, which also demonstrates its interaction with suffixation and the fact that reduplicative forms with hyphens are parsed correctly. goody-goody dup - good -y informal - nice -FAMIL redup - adj -familiar PNG Computing Handbook Appendix STYLEREF Title \* MERGEFORMAT Parsing with Shoebox PAGE3 page1 ; ]]=/&&}} .)))()()()()() }}=/&& .)))()()()()() S u m m a r y I n f o r m a t i o n ( @ V @ D# 2 @ [U Microsoft Word 6.0 65 + / F M R K L O X * v # ' @ D I L ^ f w # & ` d * , R U Y [ v x A E = @ a C b m Pj PA P? U PC P> ` g" p" t" " # # # # /% 4% Q% R% X% Y% `% a% j% k% t% u& & & & & ' ' &( 0( D( J( X( Y( c( d( p( q( |( }( ( ( + + + + + + , , , , , , - - - - - - . . . . */ / V0 W0 c0 e0 o0 p0 0 0 1 1 2 2 2 3 3 3 3 3 3 3 v3 w3 |3 }3 3 3 3 3 V J U ] PC `3 q5 t5 5 5 5 5 5 5 5 5 5 5 .6 36 e6 g6 i6 m6 r6 u6 7 7 7 7 p8 t8 8 8 : : 5: 8: : : < "< '< +< 2< 5< n= o= = = = = > > xA A A A A A A A A A A A A A A A A A A A A E u uD Pj PC F ! " / F N ( ~ F & 0 ? M * c c = c = ) c Js c 4 # U = O * % G ] 3 a C W y J a p c J c ` E c Y # c J c , uH 4 c !p :! ^! " # f# $ 5$ }$ $ $ &% Q% t% u& & V' ' ' ' X( ( ( ( ) L) C* * + M+ ]+ + c u c 6% x c c c c c = 4 + , /, , - - h. . . / */ / S0 v0 0 0 )1 o1 1 |2 2 2 v3 3 3 g4 4 4 5 6 7 N8 8 8 c c 4 8 HF c m c c9 6 4hl 4 4 8 HF !8 4; ; < < q= = = y> Y? ? ? @ ~@ 1A xA A A A A A A A A A A A A i i i i i i i{ i{ (# (# (# (# c p@ c m/ u c c Ft 4 k K @ Normal 3 ] a ` @ ` Heading 1 C ,< & 4 . c V @ V Heading 2 : 4 c Z @ Z Heading 3 = 4 . c Z @ Z Heading 4 = 4 . c Z @ Z Heading 5 = 4 . c R @ R Heading 6 : 4 . R @ R Heading 7 : 4 . R @ R Heading 8 : 4 . R @ R Heading 9 : 4 . " A@ " Default Paragraph Font * O * Heading Base U] O Paragraph " B@ " Body Text x 3 *@ ! Endnote Reference h &@ 1 Footnote Reference h @QB TOC 4 c . O R. TOC Base 0 E# ] @Qb TOC 3 c @Qr TOC 2 c " @Q" TOC 1 x x Uc @ Footer (# O Header Base & @& Header (# U] & @& Footnote Text p * O * Footnote Base < c $ J@$ Subtitle < Vc $ >@ $ Title c$ k $ O $ Author Uc * O * Institution ! x x Uc , O ", Abstract " h V* OA2* Num Ex Base # 8 ` $ OB$ Example Base $ < 4 OaR4 Bibliog Author (AAA style) % O b Bibliog Base & < Oar< Bibliog Citation (AAA style) ' < ( Oa( Bibliog Entry ( 0x , O , Chapter Author ) x Uc * O * Chapter Number * Uc * O * Chapter Title + c & O& Example in Footnote , * O* Footnote Continuation - ( O( Example in Paragraph . * O* Paragraph Continuation / $ O$ Figure Body 0 O Figure Base 1 4 O 4 Front/Back Matter Heading 2 c ( O( Figure Source 3 < ] < OAB< Plain Example 4 x 8 H ] a * Oa* Quotation Continuation 5 2 O2 Quotation Paragraph 6 h c & @ & Index 1 7 8 E# O1 Text Example 8 2 O12 Phono Ex Lettered 9 [ +@ Endnote Text : $ O$ Acronym (all caps) [c O Bold U$ O$ Book or Journal Title V" O" Emphasized Word(s) UV O Defined Word(s) U O Foreign Word(s) V$ O$ Metalinguistic Word(s) V" O!" Paragraph Title U] $ O1$ Vernacular Word(s) ] a OA Italic V OQ Comment \ Oa Index Entry \ Oq Subscript h O Superscript h @Q TOC 8 I V 0@V List Bullet < J 4 h & O& List Base K 8< Z 6@Z List Bullet 2 = L 4 h /@ List M 2@ List 2 N V 1@V List Number < O 4 h . Z :@Z List Number 2 = P 4 h ) " "@" Caption Q x < ] . O". List Labelled R $ D@2$ List Continue S & O A& Phonetic Characters ] a O Q StyleName ] c $ E@b$ List Continue 2 V )@ q Page Number " O " Figure Number UZ] $ O$ Table Heading Y U O Table Entry Z 2 ,@Q 2 Table of Authorities [ 8 E# . #@Q . Table of Figures \ p E# !@ r Index Heading ] O " Date ^ x U* O1* Phono Ex _ ` [ $ O12$ Interlin Example ` & O1& Free Translation a * O1* Interlin Heading b x @ O12@ Interlin Continuation c x & ' ( ) ] a " OB" StyleTable d p @QR TOC 5 e @Qb TOC 6 f @Qr TOC 7 g @Q TOC 9 h , O2, Emphasized Vernacular Word(s) U " > B D # " ( % f $ ' h+ o. 1 48 y; > _ * ! B ! " / F N ( ~ F & 0 ? M * U = O * % G ] 3 a C W y J a p : ^ f ! 5! }! ! ! &" Q" t" u# # V$ $ $ $ X% % % % & L& C' ' ( M( ]( ( ) /) ) * * h+ + + , *, , S- v- - - ). o. . |/ / / v0 0 0 g1 1 1 2 3 4 N5 5 5 48 8 9 9 q: : : y; Y< < < = ~= 1> x> y> > 4 c 4 c 4 c c 4 c 4 c 4 c 4 4 c 4 c 4 c 4 c 4 c 4 c 4 c 4 c c 4 c 4 c 4 4 c 4 c 4 4 c 4 c 4 4 4 c 4 c 4 4 c ! Y c e o r 3 E # $ % p + 8 A & ' ( ) * > ! A V Y ^ ` e j l r !!: _Toc360527954 _Toc360615144 _Toc361623456 _Toc361626072! ! ! ! > / / / / > David BevanC:\SHOEBOX\DOC\SHPARSE.DOC David BevanC:\SHOEBOX\DOC\SHPARSE.DOC David BevanC:\SHOEBOX\DOC\SHPARSE.DOC David BevanC:\SHOEBOX\DOC\SHPARSE.DOC David BevanC:\SHOEBOX\DOC\SHPARSE.DOCDallas Computer ServicesD:\SHW\README\SHPARSE.DOC Bryan Wussow0\\SWDEVSERVER\SOURCE\SHW\RELEASE\DOC\SHPARSE.DOCAuthorized Gateway CustomerC:\SHOEBOX\DOC\SHPARSE.DOCMark R. PedrottiC:\SHW\DOC\COPYOF~1.DOCMark R. PedrottiC:\SHW\DOC\SHPARSE.DOC@HP LaserJet 6P \\JARNET-A\CCSswdev PCL5EMS2 HP LaserJet 6P @ g X X @ MSUDHP LaserJet 6P d HP LaserJet 6P @ g X X @ MSUDHP LaserJet 6P d k k k k d Times New Roman Symbol & Arial 5 Courier New 1 Courier SILDoulosIPA " V h h &b ҋCCA c 2 l 5 !l )D:\MSOFFICE\WINWORD\TEMPLATE\ACPUBPNG.DOTParsing with Shoebox David BevanMark R. Pedrotti