An Experiment in Computational Parsing of the Navajo Verb Mans Hulden Shannon T. Bischoff University of Arizona 1 Introduction This paper presents preliminary research on a computational parser for core Navajo morphology where any inflected verb is automatically decomposed, together with the inflectional and derivational structure of the verb. The grammatical implementation largely follows Faltz (1998) and Young and Morgan (1987); Young et al. (1992). We also report some proposals for reducing the amount of allomorphy and phonological rules in the description of Navajo verbal morphology, and potential uses of such a parser. After an introduction to the elements of the Navajo verb andthe notation employed by lexicographers, pedagogists, and other scholars to express those elements, a brief discussion of subjects and objects is included. We also present some problems posed to the novice learner of Navajo attempting to navigate the primary reference material Young and Morgan (1987); Young et al. (1992). Finally, we discuss the parser and the advantages it presents to the Navajo learner. We conclude with remarks for future work on the proposed Navajo parser. 2 The Navajo Verb In this section we will provide a general overview of how the Navajo verb is presented and discussed in the primary reference material and in the (Faltz, 1998) primer. For notational simplicity, Faltz’ formalism will be adopted. However, while it does not follow the traditional sources Young and Morgan (1987); Young et al. (1992) in the letter, it does follow them in spirit. 101 Mans Hulden and Shannon T. Bischoff 2.1 General Assumptions: Templates and morphophono- logical complexity There are several general assumptions regarding the Navajo verb arrived at by Young and Morgan (henceforth YM) and adopted by Faltz (1998), among others, based on the phonological and morphological facts of the language (cf. esp. Young and Morgan (1987)). The Navajo verb: 1. is morphologically complex 2. it corresponds to phonological word 3. its stem is usually word final and monosyllabic 4. is commonly divided into tripartite structure based on morphotactic and phonological facts: disjunct, conjunct, and stem The tripartite structure presented in Table 1 illustrates each of these gen- eralizations. It should be noted, as mentioned, for notational simplicity we adopt a simplified structure more in line with Faltz (1998) as opposed to the more articulated structure of YM with 16 positions. Table 1: Navajo verb tripartite structure. outer prefixes plural object inner prefix subject classifier stem disjunct prefixes conjunct prefixes stem This filling of a template can be illustrated in Table 2 with a concrete in- stantiation of the verb meaning ‘They made a fire.’ Table 2: dideid´ı´ı lj´e´e’ ‘They made a fire’ (Faltz, 1998, p. 172) outer prefixes plural object inner prefix subject classifier stem di da y d ´ı´ı l j´e´e’ disjunct prefixes conjunct prefixes stem In Table 2 we see a phonological rule that changes da+y to dei. We also see each “slot” is filled in our template. In the following example we see that it is not necessary for each prefix position to be filled. Also, a prefix position is not strictly limited to one morpheme. This is the result of our simplified 102 An Experiment in Computational Parsing of the Navajo Verb Table 3: didideeshjah ‘I will make a fire’ (Faltz, 1998, p. 171) outer prefixes plural object inner prefix subject classifier stem di Ø Ø d+d eesh l jah disjunct prefixes conjunct prefixes stem template which departs from YM in that they assume a slot for each distinct morpheme. In Table 3, didideeshjah, we see that a phonological rule inserts i between d+d.1 It should be noted that the Navajo verb consists minimally of a subject, classifier (both of which may be null), and a stem. Next we turn to some interesting semantic relationships between the stem and obligatory prefixes. 2.2 Semantics of the Verb The generalmeaning ofa verb isarrived atvia thecombination ofthe various lexical morphemes and the stem—importantly the word-final stem does not provide the semantic content alone to derive necessary meaning. Meaning is arrived at via the combination the morphemes in the disjunct, conjunct, and stem. 2.2.1 Stem sets It appears to be standard practice in the Navajo literature to organize the stem (excluding the subject and the classifier) into stem sets. Following YM, Faltz (1998) also organizes stems into such sets. These are based on five elements, seen here in Table 4: From these “sets” a general semantics can be derived. That is, from the combination of each element listed in the stem set a semantics is arrived at. Next we will expand on each of these elements individually. 2.2.2 Stem sets: mode YM organize stems in sets that correspond to the seven modes. Faltz adopts this practice, often limiting the sets to the five primary modes, since the two missing modes can always be derived from the others. It will be seen below 1d+d is a portmanteau morpheme that is further divided in YM87, but not in Faltz. 103 Mans Hulden and Shannon T. Bischoff Table 4: Stem sets in Navajo are seen in the left-hand column. The right- hand column shows the information necessary for constructing a complete “lexical entry” of a Navajo verb—these include the classifier, knowledge about lexical prefixes in different positions, the verb’s transitivity, and a conjugation pattern. The classifier and some lexical prefixes also make it highly likely that a verb carries a certain telicity, or transitivity, but these two patterns are not completely regular. Stem set Classifier: I: P: Lexical prefixes: F: R: Transitivity: O: Perfective: s-P ... that the phonological realization of a given stem may vary in the different modes. Imperfective: I-Mode • – indicates ongoing action in the present, or sometimes in the im- mediate future Perfective: P-Mode • – indicates a completed event in the past - this mode is primarily divided into the y-P and s-P sub-modes Future: F-Mode • – indicates an event in the future Iterative: R-Mode • – indicates generic action, emphasizing the repetitive nature of the action 104 An Experiment in Computational Parsing of the Navajo Verb Usitative: U-Mode • – indicates generic action, emphasizing the habitual nature of the action Optative: O-Mode • – this mode is used in certain dependent structures whose meaning involves a notion of desire Progressive: Pg-Mode • – many Navajo verbs do not have a progressive mode, it is used rather than the imperfective to indicate ongoing action in the present It should be noted that not every stem set employs each of the modes. We turn now to obligatory morphology, that is, the obligatory prefixes. 2.2.3 Stem sets: Obligatory Prefixes Faltz (1998) identifies a number of obligatory prefixes—prefixes which must occur with the stem in all modes to arrive at the general semantics of the verb—many of these prefixes do not always avail themselves to a ready se- mantic reading in isolation. Some examples follow where outer corresponds with disjunct and inner conjunct and no gloss is given: outer prefixes inner prefixes ha d na j b’iki ’ A challenge posed by these obligatory morphemes will be discussed more at length below. In short, while the general semantics of a given verb in Navajo may be regular across dialects, the morphological parts of that verb may be analyzed in different ways by different speakers. That is, some speakers may analyzethemorphemeha, forexample, asrepresenting anunspecified object, while others may interpret it as a locative. However, both interpretations may contribute appropriate semantic content to arrive at the same semantics for a given complete verb-form. This will be clarified further below. 105 Mans Hulden and Shannon T. Bischoff 2.2.4 Stem sets: Classifier There arefour obligatoryclassifiers inNavajo—astem must appear withone: Navajo Classifiers barred-l l plain-l l d-classifier d zero Ø Theexactfunctionandmeaningofthesemorphemesisnotclear. Thebarred- l is generally associated with transitive constructions, while the plain-l is associated with intransitive. These two are collectively referred to as valence markers (McDonough, 2000a). A number of phonological rules often reduce the various classifiers to null. These rules will not be presented here, but it should be kept in mind that a given classifier may not surface due to phonological effects. 2.2.5 Stem sets: telicity Stem sets may be divided into pairs, with one occurring with the mandatory na prefix the other without: Table 5: Stem set pairs Stem sets with na: • – atelic in nature—referred to as continuative aspect stem-set Stem sets without na: • – telic in nature—referred to as momentanous aspect stem-set Not all stem sets come in such pairs. Such a generalization holds true much ofthetime, but like many otherphenomena inthelanguageisnot completely regular. 106 An Experiment in Computational Parsing of the Navajo Verb 2.2.6 Stem sets: examples Here are two examples of stem sets which form a pair. The first illustrates the continuative aspect and the second the momentanous aspect. Each is followed by an example of a fully inflected verb. Note that the stem set in each example has different phonological realizations for the various modes in some, but not all cases. In general, there is no apparent regularity between stems in the various modes of a given stem set. Some heuristics for guessing other stems based on knowledge of the stem in one mode is given by Faltz (1998), but such are not expected to work with any consistency. Table 6: Continuative aspect stem-set—atelic ‘carry him/her around’ Stem set Classifier: barred-l I: t´e P: t´˛ı Lexical prefixes: na (outer) F: tee l R: teeh Transitivity: transitive O: tee l Perfective: s-P Table 7: nan´ısht´e ‘I am carrying you around’ (Faltz, 1998, p. 118) outer prefixes plural object inner prefix subject cl stem mode na n sh l t´e I disjunct prefixes conjunct prefixes stem 2.3 Interim Summary The verbal word in Navajo comprises a stem and a variety of obligatory mor- phemes which in turn derive the semantics of the verbal ‘word.’ The verbal word in Navajo minimally consists of a stem, classifier and subject (the lat- ter two of which may be null). Further, it has been demonstrated that the templatic notion employed captures the facts of the Navajo verb composition 107 Mans Hulden and Shannon T. Bischoff Table 8: Continuative aspect stem-set—telic ‘carry him/her up out of some- thing’ Stem set Classifier: barred-l I: teeh P: t´˛ı Lexical prefixes: ha (outer) F: t´e´el R: t´e´eh Transitivity: transitive O: t´e´el Perfective: s-P Table 9: hash´ı lteeh ‘You are carrying me up’ (Faltz, 1998, p. 113) outer prefixes plural object inner prefix subject classifier stem ha Ø sh Ø ni l teeh disjunct prefixes conjunct prefixes stem rather straightforwardly. To illustrate two points—that obligatory morphol- ogy contributes necessary meaning and that speakers may interpret the verb word similiarly and its parts differently—the following example is provided. Faltz provides the following stem list for the verb word ‘sing.’ Faltz further provides the following paradigm for the Perfective mode. Note the I-mode forms in the below paradigm, in particular the first singular with the ha initial morpheme. We can illustrate the derivation of hashtaa l as follows if we assume Faltz’ analysis of hw as an obligatory object morpheme: Note here that a phonological rule in this particular form does change hw into ha. Navajo speakers indicate that you cannot simply say taa l on its own. That is, the prefix must accompany the stem. What is even more interesting is that some speakers analyze the prefix morpheme not as an obligatory unspecified or fourth person object, but rather as a locative meaning ‘up and out.’ Despite the different interpretations of the morpheme the semantics of theverbremainsthesame,addingstrengthtotheclaimthatitisnotthestem alone that derives meaning, but a composition of the lexical morphemes and 108 An Experiment in Computational Parsing of the Navajo Verb Table 10: ‘sing’ (lit. sing it) (Faltz, 1998, p. 132) Stem set Classifier: barred-l I: taa l P: t´a´a l Lexical prefixes: hw (4th P object) F: ta l R: ta l Transitivity: transitive O: taa l Perfective: s-P Table 11: P(y)-mode paradigm ‘sing (it)’ (Faltz, 1998, p. 133) sg dpl distr dpl 1 h´o´ot´a´al hwiit´a´al dahwiit´a´al 2 hw´ı´ın´ıt´a´al hoot´a´al dahoot´a´al 3 h´o´ot´a´al dah´o´ot´a´al 4 hoj´ı´ıt´a´al dahoj´ı´ıt´a´al the stem. We will return to this form and the potential ambiguity between the outer ha-prefix and the 4th person object prefix hw. Next we turn to a brief discussion regarding the phonological realization of subjects and objects. 2.4 Subjects Faltz(1998)identifies 28subject paradigms, adifferent sub-set foreachofthe seven possible modes. Some modes have several conjugation patterns. The P mode, for example, is divided into the P(s), P(y), and P(n) sub-modes, or si-perfective, yi-perfective, and ni-perfective, respectively (the latter being YM’s terminology). The subject allomorphy is usually determined by the combination of the classifier, mode, and whether it is preceded by a disjunct prefix, conjunct prefix, or no prefix. Each mode has a set of allomorphs for a variety of phonological realiza- tionsdepending onvariousphonologicalconditions. Apotentialcomplication 109 Mans Hulden and Shannon T. Bischoff Table 12: I-mode paradigm ‘sing (it)’ sg dpl distr dpl 1 hashtaa l hwiitaa l dahwiitaa l 2 h´otaa l hohtaa l dahohtaa l 3 hataa l dahohtaa l 4 hojitaal dahojitaal Table 13: hashtaa l ‘You are carrying me up’ (Faltz, 1998, p. 113) outer prefixes plural object inner prefix subject classifier stem Ø Ø hw Ø sh Ø taa l disjunct prefixes conjunct prefixes stem for the novice learner. In addition to the 1st through 3rd person subjects, there is also a 4th person, or “alternate third person” form which will be briefly discussed. Some allomorphs are not fully specified: for example, in the yi-perfec- tive, the hallmark third person subject form yi is sometimes represented as just two underspecified vowels vv—vowels which are copied over from earlier phonological material in the disjunct. In addition, the fourth person subject form is primarily marked both in the object and the subject positions—i.e. the template is not entirely rigid. 2.4.1 Fourth person or 3a “Alternate third person forms” The Navajo fourth person has the following properties. indicates an unspecified or indefinite subject (like “one” or “they”) • polite way of referring to somebody who is present in conversation • used for reference tracking, to keep one third person entity separate • from another when telling a story obligatorily appears with third subject • may appear adjacent to an object • 110