a parser for real-time speech synthesis of conversational texts
Contents
1. incorporating a relatively small vocabulary and rule set the parser has proven unexpectedly successful in both laboratory and field tests REFERENCES Bachenko J A Taxonomy of Syntactic Variation in Written Language of the Deaf Unpublished data 1990 Bachenko J and E Fitzpatrick A Computational Gram mar of Discourse Neutral Prosodic Phrasing in English Computational Linguistics 16 155 17 1990 Charrow V Deaf English Technical Report 236 Insti tute for Mathematical Studies in the Social Sciences Stanford University 1974 Emorine O M and P M Martin The Multivoc Text to Speech System Proceedings of the Second Confer ence on Applied Natural Language Processing ACL 115 120 1988 Hindle D User Manual for Fidditch a Deterministic Parser NRL Technical Memorandum 7590 142 1983 Kukich K Spelling Correction for the Telecommunica tions Network for the Deaf Communications of the ACM 1992 Marcus M A Theory of Syntactic Recognition for Natural Language Cambridge MA MIT Press 1980 Olive J P and Liberman M Y Text to Speech An Overview Journal of the Acoustic Society of America Supplement 1 78 S6 1985 O Shaughnessy D D Parsing with a Small Dictionary for Applications such as Text to Speech Computational Linguistics 15 97 108 1988 Suri L Language Transfer A Foundation for Correcting the Written English of ASL Signers University of 32 Delaware Technical Report2. is short and at prepositional phrase boundaries unless the prepositional phrase is short A short noun is a single word noun phrase such as a pronoun or demonstrative this that a short prepo sitional phrase is one with a pronominal object with me about it etc Hence the noun verb rule will produce the phrasings below where double bars mark phrase boundaries this and the prepositional phrase rule are adaptations of the verb and length rules respectively given in Bachenko and Fitzpatrick 1990 MY CAR IS HAVING A TRANSMISSION PROBLEM IT IS HAVING A TRANSMISSION PROBLEM Our formulation of the phrasing rules assumes that in the absence of syntactic structure the subclass member ship past of speech and string position can provide sufficient information to infer structure in many cases For example we are assuming that the subclass nominative_pronoun which includes he she we etc acts consistently as the leading edge of a sentence so that the parser can compensate somewhat for the lack of punctuation by identifying and setting off some top level sentential constituents Similarly prepositions are assumed to act consistently as the leading edge of a prepositional phrase the parser guesses about preposi tional phrase length by checking the word class of the element following the preposition to see if the object is pronominal The phrase rules thus attempt to seek out major syn tactic constituents If ther
3. the listener This is a task that in itself is problematic because of the human error involved in the judgments We regard the field trial as the appropriate method of evaluation here and we use the results of the laboratory test to help us characterize the parser s failures rather than to evaluate the parser After the phrasing was assigned manually the TDD data were run through the parser and the parser s phras ing was compared with the human judgments Approxi mately 20 of the corpus had been used to extrapolate rules for the parser while the remainder of the corpus was saved for testing only there was no appreciable per formance difference between the two data subsets The corpus contained 8788 words and according to the human judgement 2922 phrases Since punctuation in these data is sparse very few of the actual phrase boun daries come for free Even so the parser performed well in the 2922 phrases it produced 573 misleading errors rendering 80 4 of the phrases acceptably There were 896 sites where the parser produced a phras ing different from that of the human judge but which we judged to be not misleading The parser s error rate reflects the constraints of its construction as a real time system in particular its three term buffer size its lack of hierarchical structure building rules and its pared down lexicon Figure 4 gives a char acterization of the most frequently encountered parsing errors alo
4. the third term of the buffer contains a word that is lexically marked as noun verb the word will be assigned the category verb when the second buffer element is the word to and the first buffer element is either a verb or adverb When applied to the word string expect to call this rule correctly analyzes call as a verb Other part of speech rules distinguish the preposition to from the use of to as an infinitive marker and distinguish the preposition vs verb uses of like Ambiguous abbreviations are items such as no which may signify either number or the negative particle Since TDD texts lack punctuation the only clue to usage in such cases is local context e g the presence of the words the or phone before no are used as disambiguating con text to identify no as number 3 1 3 Phrasing rules consider part of speech word subclass and length as measured by word count to TDD User TDD Parser User Figure 2 Block Diagram of TDD TTS System identify phrase boundary locations These mules are strictly ordered In general they instruct the synthesizer to set off interjections e g wow oh ok etc and to insert a phrase boundary before non lexical coordinate conjunctions e g and in I don t recall that and am not sure see Bachenko and Fitzpatrick 1990 163 before sentences and before subordinate conjunctions after during etc Boundaries are also inserted at noun verb junctures unless the noun
5. 91 19 1991 Tsao Y C A Lexical Study of Sentences Typed by Hearing Impaired TDD Users Proceedings of 13th International Symposium Human Factors in Telecom munications Torino Italy 197 201 1990
6. A PARSER FOR REAL TIME SPEECH SYNTHESIS OF CONVERSATIONAL TEXTS Joan Bachenko Jeffrey Daugherty Eileen Fitzpatrick AT amp T Bell Laboratories Murray Hill NJ 07974 ABSTRACT In this paper we concern ourselves with an applica tion of text to speech for speech impaired deaf and hard of hearing people The application is unusual because it requires real time synthesis of unedited Spontaneously generated conversational texts transmitted via a Telecommunications Device for the Deaf TDD We describe a parser that we have implemented as a front end for a version of the Bell Laboratories text to speech synthesizer Olive and Liberman 1985 The parser prepares TDD texts for synthesis by a performing lexical regularization of abbreviations and some non standard forms and b identifying prosodic phrase boundaries Rules for identifying phrase boundaries are derived from the prosodic phrase grammar described in Bachenko and Fitzpatrick 1990 Following the parent analysis these rules use a mix of syntactic and phonological factors to identify phrase boundaries but unlike the parent system they forgo building any hierarchical structure in order to bypass the need for a stacking mechanism this permits the system to operate in near real time As a component of the text to speech system the parser has undergone rigorous testing during a successful three month field trial at an AT amp T telecommunications center in California In addi
7. as a b or c a so I feel there should be a better system to say bye b so I feel there should be a better system to say bye c so I feel there should be a better system to say bye If the parser assigns for example the phrasing in a while the human judge assigns b it must be counted a qualitatively different from the parser s assignment of misleading boundary where the hearer s understandin of the import of the utterance is altered because of th erroneous boundary placement An example of mislead ing boundary placement as assigned by the parser i given below where the hearer is incorrectly led to inte pret well as a modification of see rather than as discourse comment oh i see well so i call my boss In a similar vein giving equal weight in an evaluz tion to the locations where pauses do and do not occur i misleading The absence of a phrasal boundary betwee two words is much more common than the presence of boundary so that predicting the absence of a boundary i always safer and leads to inflated evaluation scores thi make comparison of systems difficult For example i the a sentence above there are 12 potential prosodi events one after each word If a given system assigr no breaks in this sentence and if non events are give equal weight with events then the system will get score for this sentence of 91 6 percent since it gets 11 the 12 judgments right Also if a system assigns o
8. d as a component in a version of the Bell Labs text to speech synthesizer Olive and Liberman 1985 The synthesizer forms the core of a telecommunications system that ran for three months as a feature of TRS in California Several thousand TDD texts were processed by the system Although restric tions on confidentiality prevented us from collecting actual TDD text data results of the field trial far sur passed expectations disconnect rates for text to speech calls averaged less than 20 and follow up surveys indi cated a high degree of interest in and acceptance of the technology A second type of testing that has enabled us to focus on the parser involves the collection of data from a ques tionnaire given to TDD users Phrasing for these data was assigned manually by a linguist unfamiliar with the rules of the parser to allow for comparison with the parser s output Several issues arise in the comparison of human judgements of phrasing with those of a phrase parser s output One of the more ubiquitous is that of phrasal 30 balancing Apparently acting under rhythmic constraints speakers tend to aim for equivalent numbers of stresse syllables on either side of a break However the incor poration of rhythm into phrasing varies from speaker t speaker as well as being partially dependent on semanti intent For example the sentence so I feel there shoul be a better system to say bye taken from our data coul be phrased either
9. e is evidence of constituency the parser may look for a short constituent or it will sim ply insert a prosodic boundary at a presumed syntactic boundary e g a verb phrase sentence or subordinate conjunction 28 3 2 PARSER IMPLEMENTATION 3 2 1 SYSTEM ARCHITECTURE The quickest way to incorporate a TDD parser into a ser vice using text to speech TTS synthesis is to implemen the parser in a separate front end module to the text to speech system The parser filters the input stream from TDD modem and sends the processed text to the text to speech system where it is synthesized for the voice tele phone user as shown in the block diagram in figure 2 This architecture minimizes the need to modify am existing equipment or system Also it allows us t maintain and change the parser module without introduc ing substantial or unpredictable changes elsewhere iu the system 3 2 2 IMPLEMENTATION Integrating the TDD parser into a near real time systen architecture is a difficult task To achieve it the parse must a filter the TDD input stream in real time in orde to identify tokens i e words abbreviations and expres sions that are suitable for processing by parser rules an b group these tokens into natural sounding phrases tha can be sent to the text to speech system as soon as the are formed In an ideal situation it is desirable to parse the entir TDD input before sending the processed text to the text to speech synt
10. e the adverb occasionally modifies in other than that the services is great occa sionally some operators are pretty slow depends on knowing that one does not give expansive praise to something that happens only occasionally In general it appears that the parser s accuracy in pbrasing the incoming messages cannot be improved without a severe loss in real time efficiency that the storage of hierarchical structure would involve Given this situation it is worthwhile to consider that despite what is probably about a 20 error rate in the system consumers used it successfully and willingly It may be that the system did no worse than the trs operators who unlike our laboratory linguist do not have the luxury of stopping to consider the felicity of a particular phrasing This may be compounded with the possibility that users may be able to compensate more easily for machine degradation of an utterance than for an operator s error since their expectations of the latter s performance are greater 5 CONCLUSION We have described a text to speech parser for conver sational texts generated by users of TDD s The parser s main tasks are to provide some regularization of non standard items and to determine prosodic phrasing of the text Phrasing allows processing to take place in near real time because the synthesizer can generate speech while the TDD message is being typed instead of waiting for the finished text Finally although
11. ent of the text to speech system the parser has undergone rigorous testing during a successful three month field trial at an AT amp T telecommunications center in Califor nia In addition laboratory evaluations indicate that the parser s performance compares favorably with human judgments about phrasing In Section 2 of the paper we describe the application and the texts Section 3 provides a technical description of the parser and Section 4 discusses evaluation of the parser s performance 2 THE APPLICATION Users of Telecommunications Devices for the Deaf TDD s can communicate with voice telephone users via services such as AT amp T s Telecommunications Relay Ser vice TRS During a TRS call special operators read incoming TDD text to the voice telephone user and then type that person s spoken responses back to the TDD user this makes for a three way interaction in which the special operator is performing both text to speech and speech to text conversion Text to speech synthesis Spelling Punctuation Case Syntax Expected texts 1 errors standard upper and lower st English e g AP newswire case conventions dialect TDD texts 5 errors little or single case only written none language of the deaf Figure 1 TDD vs Expected Text Input makes it possible to automate part of this arrangement by reading the TDD text over the telephone to the voice user The synthesizer thus replaces an operator on the TDD half of the conv
12. ersation providing increased privacy and control to the TDD user and presumably cost savings to the provider of the service TDD texts present unusual challenges for text to speech Except in laboratory experiments large scale applications of text to speech have tended to focus on name pronunciation and canned text such as catalogue orders To the best of our knowledge the TRS text to speech field trial in Califomia represents the first large scale attempt to use speech synthesis on spontaneously generated conversational texts and also the first to use this technology on texts that are orthographically and linguistically non standard Unlike the written material that most text to speech systems are tested on e g the AP newswire TDD texts observe few of the wnting con ventions of English All text is in upper case and punc tuation even at major sentence boundaries rarely occurs spelling and typographical errors complicate the picture even further Tsao 1990 Kukich 1992 In addition nearly all texts employ special abbreviations and lingo e g CU stands for see you GA is the message termina tor go ahead The following example illustrates a typical TDD text OH SURE PLS CALL ME ANYTIME AFTER SAT MORNING AND I WILL GIVE U THE NAMES AND PHONE NOS OK QGA Oh sure please call me anytime after Saturday morning and I will give you the names and phone numbers OK Go ahead Finally many texts are written in a variety of E
13. hesizer But the practical situatio demands that the voice user hear TDD text synthesize as soon as it is reasonably possible so that long period of silence can be prevented Figure 3 below shows th basic architecture chosen to implement the parse described in this paper 3 2 2 1 The canonical input filter process has to dea with the TDD input characters as they are being typed The output of the canonical filters consists of TDD wor tokens i e groups of characters separated by whit spaces Input characters arrive at irregular speeds wit nondeterministic periods of pauses due to uneven typin by the TDD user Also incidences of spelling error typographical mistakes and attempts to amend previousl typed text occur at very irregular rates Even the TDI modem can contribute text to the input stream that i seen by the canonical input filter For instance the TDD modem might periodically insert a carriage retum charac ter to prevent text wraparounds on the special operator s terminal Unfortunately these carriage retum characters could split words typed by the TDD user into incoherent parts e g advantage might become adva lt CR gt nuage Since the voice telephone user needs to hear TDD text synthesized after some hopefully short interval of time the input filter cannot wait indefinitely for TDD characters that are delayed in arriving as might occur when the TDD user pauses to consider what to type next Hence the filter includes an
14. in either words or structures once a lexical or prosodic structure is built it cannot be undone As TDD text is typed incoming words are collected in the buffer where they are formed into structures by mules described below Phrasing mies then scan buffer structures If a phrasing rule applies all text up to the element that trig gered the rule is sent to the synthesizer while during synthesis the buffer is reset and the rules restart anew Once a structure has moved out of the buffer it cannot be recovered for examination by later phrasing rules Our approach differs from other recent efforts to build small parsers for text to speech e g O Shaughnessy 1988 and Emorine and Martin 1988 where savings are sought in the lexicon rather than in processing O Shaughnessy 1988 henceforth O describes a non deterministic parser that builds sentence level structure using a dictionary of 300 entries and a medium sized grammar which we guess to be slightly under 100 rules The lexicon is augmented by a morpho logical component of 60 word suffixes used principally to derive part of speech for example ship and ness are considered good indicators that a word of two or more syllables has the category noun O gives a thorough account of his parser Much of his exposition focusses on technical details of the syntactic analysis and support ing linguistic data are plentiful However evaluation of O s proposals for speech synthe
15. input character timeout mechanism The timeout interval is set to an appropn ately short duration to ensure the timely synthesis of available TDD text but still long enough to prevent the exclusion of forthcoming input characters 3 2 2 2 Lexigraphical analysis examines the TDD word tokens to identify contiguous words that should be grouped together as individual units The multi word expressions include contractions e g it s which becomes it s and commonly used short phrases that can be viewed as single lexical units e g my goodness as long as and mother in law A simple stacking mechanism is used to save tokens that are identified as potential elements of multi word expressions The tokens are stacked until the longest potential multi word expression has been identified with three words being the maximum After which the stack is popped and the corresponding structures described below are con structed 3 2 2 3 The lexical lookup process builds a tdd term structure record from these tokenized words and multi word expressions in preparation for invoking the phrasal segmentation rules Fields in the structure include the tokenized input text the original orthographic representation the output orthography lexical category Noun Verb Adverb NIL etc word subclass and other fields used internally by the phrasal segmentation process At this point in the processing only the input text field ha
16. n break in this utterance but puts it in a clearly inappropr ate place say before the word bye it will get a score 83 percent since it gets 10 of the 12 judgements righ While 83 percent sounds like a decent score for a syste that must capture some subjective performance th method of evaluation has completely failed to capture tk fact that assigning an inappropriate prosodic break in th instance has completely misled the listener Therefor we need to evaluate a phrasing system on the basis positive occurrences of phrase boundaries only Assigning phrases to TDD output is not a clear ci task The output is not intended to be spoken an because of the device it has telegraphic characteristic In addition many TDD users do not have standard sp ken English at their command Nevertheless an effort CATEGORY ERROR EXAMPLE Ambiguous pronoun Verbal Complement Non Standard Syntax Copular verb Nominal modification a m o i Tase PB Adverbial modification why not surely i think need interview 59 who i long to talk to it will be great Ambiguous Interjection sorry no no other than that 3 9 is aime Tray TT phone 13 i think they the crs Figure 4 Distribution of TDD Production Errors was made to approximate the performance of TRS opera tors who speak the TDD output to voice users Care was also taken to mark as parser errors those prosodic events that would mislead
17. ng with their frequency of occurrence and an example for each characterization In the examples represents a prosodic pause and indicates that the parser performed incorrectly at this site 31 Most of the parsing errors given in Figure 4 would be resolvable if the parser were to incorporate non local structural information For example the pronouns it and you function as both subject and object In a three ele ment buffer then the status of it in to it will is undecid able since it can be the object of to or the subject of will In the context of the sentence i have friends who i long to talk to it will be great where an element corresponding to who functions as the object of to the function of it as subject of will be great and the con comitant prosodic break before it are apparent but only when the structure of the who relative clause is available The errors involving non standard syntax would require sublanguage rules that indicate the possibility of non overt subjects oh i see understand and copulas there a pause among other things but again given the limitation to local information and the lack of punctua tion this is not straightforward For example oh i see understand could continue as that i don t speak well A smaller set of errors is undecidable even given non local structural information and require further prag matic knowledge of the discourse For example the decision as to which claus
18. nglish that departs from expected lexical and syntactic patterns of the standard dialect Charrow 1974 For example WHEN DO I WILL CALL BACK U Q GA is a short TDD text that we believe most native speakers of English would recognize as When should I call you back Go ahead The attested example below is less clear but interpretable I WISH THAT DAY I COULD LIKE TO 26 MEETING DIFFERENT PEOPLE WHO DOES THIS JOB AND THE WAY I WANT TO SEE HOW THEY DO IT LIKE THAT BUT THIS PLACES WAS FROM SAN FRANCISCO I GUESS Syntactic variation in such texts is systematic and con sistent Bachenko 1989 Charrow 1974 Although a complete account has yet to be formulated Suri 1991 reports that aspects of the variation may be explained by the influence of a native language ASL on a second language English Figure above summarizes the points about TDD texts Spelling error estimates come from Kukich 1992 and Tsao 1990 Timing creates an additional obstacle since we expect TRS text to speech to synthesize the text while it is being typed much as an operator would read it at the TRS center How to chunk the incoming text now becomes a critical question Word by word synthesis where the listener hears a pause after each word is the easiest approach but one that many people find nerve wracking N word synthesis where the listener hears a pause after some arbitrary number of words is nearly as simple but runs the risk of creating unaccep
19. s any non null information The output orthography lexical category and word subclass fields are filled via lexical lookup The lexicon is organized into the four fields men tioned above The tdd term input text field is compared with the corresponding field in the lexicon until a match is found and the three remaining fields in the matched entry are then copied into the tdd term structure If no match is found then the input text field is copied into the output text field and the other two lexicon fields are set to NIL As an illustration if the single letter u is identified as our TDD token the lexical lookup process might retum with a tdd term structure that looks like 29 tote input text u output text you lexical category NOUN subclasses DESTRESS_PRONOUN SHORT_SUBJECT other fields NIL For the input text oic the structure might look like input text oic output text oh I see lexical category INTJ subclasses INTERJECTION _ other fields NIL 3 2 2 4 The phrasal segmentation process applies a modest set of disambiguation and phrasing rules to a sliding window containing three contiguous tdd term structures In the start condition the sliding window will not have any tdd term structures within it Each new tdd term structure generated by lexical lookup enters the first term position in the window bumping existing terms forward one position with the last third term discarded after i
20. s to make additional lexical informa tion available to the phrasing rules Let us briefly describe each of the three components that make up the grammar lexicon disambiguation rules and phrasing rules 3 1 1 The lexicon contains 1029 entries consisting of words abbreviations and two to three word phrases Each entry has four fields the input word e g u the output orthography you lexical category Noun and a list of word subclasses destress_pronoun short_subject Word subclasses reflect co occurrence pattems and may or may not have any relationship to lexical categories For example Interjection_1 includes the phrase byebye for now the adverb however the noun phrase my good ness and the verb smile as in I APPRECIATE THE HELP SMILE THANK YOU SO MUCH Both the lexi cal category and subclass fields are optional either may be marked as NIL Abbreviations and acronyms are usu ally nouns and make up 20 of the lexical entries Nouns and verbs together make up about 50 We expect that additions to the lexicon will consist mostly of new abbreviations and short phrases 3 1 2 Lexical disambiguation rules identify part of speech and expand ambiguous abbreviations Currently part of speech disambiguation is performed by ten mules Most apply to words lexically marked for both noun and verb e g act call need assigning a single category either noun or verb when a rule s contextual tests are satisfied For example if
21. sis is difficult since he gives us only a vague indication of how the parsed sen tences would be prosodically phrased in a text to speech system Without an explicit description of the syntax prosody relation we cannot be sure how to assess the suitability of O s analysis for speech applications The system described by Emorine and Martin 1988 henceforth E amp M incorporates a 300 entry dictionary and approximately 50 rules for identifying syntactic con Stituents and marking prosodic phrase boundaries The rules in this system build sentence level structures that are syntactically simpler than those given in O but more geared to the requirements of phrasing in that prosodic events e g pause are explicitly mentioned in the rules Unfortunately E amp M share few technical details about their system and like O provide no examples of the prosodic phrasing produced by their system making evaluation an elusive task 27 Applications such as TRS which requires near real time processing make systems based on sentence level analyses infeasible In our parser decisions about phras ing are necessarily local they depend on lexical informa tion and word adjacency but not upon relations among non contiguous elements This combined with the need for lexical regularization in TDD texts motivates a much stronger lexicon than that of O or E amp M In addition our parser incorporates a small number of part of speech disambiguation rule
22. tably high levels of ambiguity and for long texts may be as irritat ing as single word synthesis Our solution was to build a TDD parser that uses linguistic rules to break up the speech into short natural sounding phrases With partial buffering of incoming text the parser is able to work in near real time as well as to perform lexical regularization of abbreviations and a small number of non standard forms 3 A TEXT TO SPEECH PARSER 3 1 PARSER STRUCTURE AND RULES In constructing the parser our goal was to come up with a system that a substitutes non standard and abbreviated items with standard pronounceable words and b pro duces the most plausible phrasing with the simplest pos sible mechanism Extensive data collection has been the key to success in regularizing lexical material e g the conversion of fwy pronounced fwee to freeway Phrasing is accomplished by a collection of rules derived from the prosodic phrase grammar of Bachenko and Fitzpatrick 1990 with some important modifications The most radical of these is that the TDD phrasing rules build no hierarchical structure Instead they rely on string adjacency part of speech word subclass and length to make inferences about possible syntactic consti tuency and to create enough prosodic cohesion to deter mine the location of phrase boundaries The parser works deterministically Marcus 1980 Hindle 1983 It uses a small three element buffer that can conta
23. tion laboratory evaluations indicate that the parser s performance compares favorably with human judgments about phrasing 1 INTRODUCTION Text to speech researchers and developers tend to assume that applications of their technology will focus on edited text either canned material such as name and address lists or free text like the AP newswire There has been much effort aimed at preparing text to speech for appli cations such as caller identification and newsreading 1 AT amp T Bell Laboratories Naperville Ilinois 25 services in which texts are generally proofed and the pri mary challenges come from issues of name pronuncia tion intonation contouring etc In this paper we con cem ourselves with an application of text to speech for speech impaired deaf and hard of hearing people The application is unusual because it requires text to speech synthesis of unedited spontaneously generated conversa tional text Moreover the synthesis must occur in near real time as the user is typing We will describe a parser that prepares conversational texts for synthesis by first performing lexical regulariza tion of nonstandard forms and then identifying prosodic phrase boundaries The parser is derived from the pro sodic phrase system presented in Bachenko and Fitzpa trick 1990 and has been implemented as the front end of a version of the Bell Laboratories text to speech syn thesizer Olive and Liberman 1985 As a compon
24. ts output orthography is copied into a text buffer awaiting transmission to the text to speech synthesizer The various rules described in Section 3 1 above are then applied to the available tdd term structures After a pro nounceable phrase is identified the output orthography of all active tdd terms is then copied to the TTS text buffer which is subsequently sent to the synthesizer for play back to the voice telephone user Also the invocation of a timeout alarm due to tardy TDD input text causes flushing of the sliding window and text buffer into the synthesizer The sliding window and TTS text buffer are cleared and the rules restarted anew TDD Text Input Canonical Input Filter Lexigraphical Analysis Lexical Lookup Phrasal Segmentation TTS Figure 3 TDD Parser Architecture Listed below are a few examples of TDD text pro cessed by the parser TDD I DONT THINK SO I WILL THINK ABOUT IT GA TTS I don t think so I will think about it Go ahead TDD HELLO HOW ARE U Q GA TTS hello how are you Go ahead TDD OK YES I AM WILLING TO GIVE INFO GA TTS okay yes I am willing to give information Go ahead TDD MY GOODNESS UR MOTHER IN LAW IS HERE GA TTS my goodness your mother in law is here Go ahead 4 EVALUATION OF PERFORMANCE Evaluation of the parser has involved two quite different forms of testing a field trial and laboratory evaluation First the parser was implemente
Download Pdf Manuals
Related Search