<?xml version="1.0" standalone="yes"?>
<Paper uid="P86-1022">
  <Title>THE CONTRIBUTION OF PARSING TO PROSODIC PHRASING IN AN EXPERIMENTAL TEXT-TO-SPEECH SYSTEM</Title>
  <Section position="6" start_page="149" end_page="151" type="evalu">
    <SectionTitle>
RESULTS AND FUTURE RESEARCH
</SectionTitle>
    <Paragraph position="0"> To date. our system has been tested systematically on a set of 39 sentences, and its performance has been observed less formally on a set of approximately 300 sentences. 9 The test corpus covers a repair manual for telephone switching systems and an introductory description of the Prose 2000 text-to-speech system.</Paragraph>
    <Paragraph position="1"> We added sentences cited in Umeda (1982) and sentences that we composed in order to extend the range of syntactic constructions represented in the test. In general, we have observed a significant improvement of prosodic quality in those test 9 The 39 sentences are listed in the appendix to this paper. sentences where the parser and the prosodic component have returned acceptable results.</Paragraph>
    <Paragraph position="2"> We have observed problems, however, especially in the formal test corpus, much of which we chose for its potential difficulty. Of the 39 test sentences, 38 parsed correctly. Of these, the prosodic component returned 26 sentences with a complete set of acceptable prosody markings. In terms of actual markings, the system marked 393 prosodic events, of which 21 markings were unacceptable. We can attribute errors in those sentences with unacceptable prosodic markings to three distinct problems discussed below.</Paragraph>
    <Section position="1" start_page="149" end_page="151" type="sub_section">
      <SectionTitle>
Complement Sentences.
</SectionTitle>
      <Paragraph position="0"> Five of the errors that arose from the prosody system's treatment of the test corpus result from the fact that the system sets off all subordinate sentences, including complement sentences, from the main sentence. Informal testing of the productions of four informants on the relevant data indicated that this approach works correctly for complement sentences such as (13)-(16). (Complement sentences are italicized):  (13) Health services cautioned Western residents -- that they should ask where their watermelons come from before buying.</Paragraph>
      <Paragraph position="1"> (14) We have to satisfy people -- that the crisis is past.</Paragraph>
      <Paragraph position="2"> (15) The vendors explained -- that this is the result of illness among 281 people who ate pesticidetainted watermelons.</Paragraph>
      <Paragraph position="3"> (16) Watermelon growers wonder -- whether this will continue throughout the rest of the season.</Paragraph>
      <Paragraph position="4"> However. the informant test consistently indicated that the complement sentences in (17)-(19)&amp;quot; are not set off by a comparable boundary: (17) They believe California sales are still off 75 percent.</Paragraph>
      <Paragraph position="5"> (18) They think the Southeast is shipping half its normal load.</Paragraph>
      <Paragraph position="6"> (19) Growers and retailers claimed the incident  hurt sales across the USA.</Paragraph>
      <Paragraph position="7"> Cases like (17)-(19). in which no break is perceived between the verb and its complement sentence, form a syntactically distinct class in Fidditch. This class is characterized by the fact that the verbal head in each case is one that does not require that its complement sentence begin with a complementizer (either that, for, or a wh- word). The class includes epistemic verbs, like those in (17)-(19), as well as a wide range of verbs that take either tensed sentences, or various types of non-tensed sentences as complements) deg The examples (20)-(26) demonstrate the range of this class (complement sentences are italicized): l0 Fidditch, in followin~ the outlines of Chomskv's (1981) Government and Binding theory, assumes that propositions, i.e., those elements that cBntain k\]oth a prkdicate and a perhaps null subject, are syntactically represented as sentences, regardless of tensing.</Paragraph>
      <Paragraph position="8">  (20) We had the ship's forces make temporary repairs.</Paragraph>
      <Paragraph position="9"> (21) We saw the crew repairing the unit.</Paragraph>
      <Paragraph position="10"> (22) He wants the units repaired by the ship's force. (23) The construction of the unit makes detailed investigation impractical.</Paragraph>
      <Paragraph position="11"> (24) Try to give the names of the characters in advance.</Paragraph>
      <Paragraph position="12"> (25) They will help finish the job.</Paragraph>
      <Paragraph position="13"> (26) The new equipment will facilitate making  repairs.</Paragraph>
      <Paragraph position="14"> Sentence-Final Constituents.</Paragraph>
      <Paragraph position="15"> Fifteen of the errors that arose from the system's treatment of the test corpus result from a high boundary value that sets final constituents off from the main sentence. The high value is due to the system's purely left-to-right attachment of syntactically unattached constituents (see rule iii.d above). The high boundary value is acceptable in sentences like (27)-(29). (The relevant final constituents in these examples are italicized).</Paragraph>
      <Paragraph position="16"> (27) In these instances it may be desirable to use phoneme characters instead of text characters to represent a word -- each time it appears in the input text.</Paragraph>
      <Paragraph position="17"> (28) Phonemic characters can also be used to handle syntactic data such as boundaries -which can improve speech quality.</Paragraph>
      <Paragraph position="18"> (29) We were unable to finish the work -- due to equipment failure.</Paragraph>
      <Paragraph position="19"> However. the high boundary value sets the final constituent off unnaturally from the main sentence in data such as (30)-(32).</Paragraph>
      <Paragraph position="20"> (30) The method by which you convert a word into phonemes is provided -- in Chapter 7.</Paragraph>
      <Paragraph position="21"> (31) The experimenters instructed the informant to speak -- naturally.</Paragraph>
      <Paragraph position="22"> (32) We discussed the techniques -- we had implemented.</Paragraph>
      <Paragraph position="23"> In many cases it appears that the grammatical relation of the final constituent to the rest of the sentence determines the boundary value that sets off this constituent. In particular, sentence adjuncts, which bear no relation to any single item in a sentence, are set off by a minor phrase boundary. whereas final constituents that modify a particular item are less perceptibly set off. This is the distinction between the final constituents in (27)-(29), which are adjuncts, and those in (30)-(32), which are modifiers. However, while the distinction between the grammatical relations of the core sentence (complement and subject) and those of the periphery (adjunct and modifier) is fairly straightforward, and handled directly bv the mechanisms of the Fidditch parser, the distinctions between the peripheral elements of adjunct and modifier are complex and require the addition of costly mechanisms.</Paragraph>
      <Paragraph position="24"> The cost of adding adjunct/modifier distinctions is illustrated by the ambiguity that arises when both adjunct and modifier readings are possible. For example, on one reading of (31), naturally modifies the verb speak; i.e., the informants were to speak in a natural manner. On the other reading, naturally is an adjunct equivalent to of course. (To see this meaning more clearly, consider the rearrangement of this sentence with the adjunct at the beginning: Naturally, the3: instructed the informants to speak.) The context of speech analysis prefers the former reading. However, the net benefit of adding sophisticated contextual analysis to our system, if attainable, is, at best, unclear. The same may be said of adding selectional restrictions, or detailed information on logical form. In contrast, a finer treatment of local syntactic constraints on boundary values preceding final constituents is within reach. From the data we have examined, it appears that the character of the prosodic event before the final constituent can be locally determined to a great extent. For the most part. this determination depends on the category type of the final constituent and on the contents of the leading edge of the constituent. For example, interjections (however. moreover, therefore, alas, thus, of course, etc.) and sentence adverbs (apparently, generally, luckih' etc.) are uniformly set off by a high boundary value and should remain so. In contrast, the boundary value of final prepositional phrases, particularly those with a monosyllabic preposition (in, on. at, to. with, for) as 11 the left edge of the phrase, should be reduced. We are currently engaged in categorizing the constituent types and left-edge items that characterize final constituents with respect to the prosodic event that precedes them.</Paragraph>
      <Paragraph position="25"> Alternatively, we are considering the play-it-safe approach of reducing the high boundary values that set off final constituents to mid-boundary values. Currently these values are converted to a downstepping feature. This approach may also be useful in conjunction with our local determination approach for those constituents whose status is either undecidable or ambiguous under the latter approachJ ~ 11. In this view, expressions such as in principle, iJ~ eenerul, in particular, in consideration of, etc. must be treated like interjections.</Paragraph>
      <Paragraph position="26"> 12. Reducing the final boundary ~alue leaves ambiguities unresolved. For sentences such as (i! and (ii), below, we believe this lack of resolution is appropriate: (i) John saw a ~irl in the park with a telescope. park.liThe telesccTpe is witli John or the girl. or it's in the (ii) I need a woman to fix the sink.</Paragraph>
      <Paragraph position="27"> \[I need a woman so that I can fix the sink. I need a woman who can fix the sink.\] Our view, following. _Marcus. and.. Hinde (p.e.) is that in normal, spoken Enghsh, such ambl~ulnes are not processed unless the speaker or listener is directly questioned re~,arding the ambiguity, .... Likewise. the. _pr~osodic events . ~hat. mi g ht dlsamblguate are inappropriate unless such questioning occurs. Other cases are less clear. For example, it is difficult to imazine that, in (28) the difference between the readin~ of the whic~'h clause as a sentence adjunct and as a noun~phrase modifier on boundaries is not processed. We would hope that in such cases some local distinction, such as the presence or absence of the comma in (28), obtains.  k ! Sentence-Initial Constituents.</Paragraph>
      <Paragraph position="28"> When a sentence contains both sentence-initial and sentence-final adjuncts, the sentence-initial adjuncts will be less prominently set off than the sentence-final adjuncts due to the left-to-right attachment of adjuncts to the prosodic tree (see rule iii.b above). In data like (33), however, a more appropriate rendering would have the boundary after the adjunct 011 a clear day be strong relative to the boundary before the adjunct as it rises over the mountains.</Paragraph>
      <Paragraph position="29"> (33) On a clear day you can see the sun as it rises over the mountains.</Paragraph>
      <Paragraph position="30"> While it would be trivial to increase the value of the pertinent boundary, we are as yet unsure what the critical features are which require a more perceptible boundary. For example, while a higher boundary value after the prepositional phrase in (34) might b'e acceptable, it is not clear that it is necessary: (34) In the morning John left.</Paragraph>
      <Paragraph position="31"> Given the stylistically distinct nature of this data, we have not yet considered this question in detail.</Paragraph>
      <Paragraph position="32"> Summary.</Paragraph>
      <Paragraph position="33"> While we have systematically tested our system so far on a small set of examples, the number of prosodic events involved in those examples, 393. is high, due to the length of the sentences tested. We find the 5 percent error rate, representing 21 prosodic events, encouraging at this stage in the development of the system. In addition, we have delimited the problem areas of an approach that relies solely on information available in the syntax tree. Our initial investigation of these problems indicates that at least part of the necessary information about phrase-level prosody is conveyed in the lexicon per se. Additionally, due to the left-corner orientation of the Fidditch parser, which exists independently to optimize search strategies, the necessary lexical information is made easily available.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>