Downloads & Free Reading Options - Results

The Hearsay II speech understanding system being developed at Carnegie-Mellon University has an independent knowledge source module for each type of speech knowledge. Modules communicate by reading, writing, and modifying hypotheses about various constituents of the spoken utterance in a global data structure. The syntax and semantics module uses rules (productions) of four types: (1) recognition rules for generating a phrase hypothesis when its needed constituents have already been hypothesized; (2) prediction rules for inferring the likely presence of a word or phrase from previously recognized portions of the utterance; (3) respelling rules for hypothesizing the constituents of a predicted phrase; and (4) postdiction rules for supporting an existing hypothesis on the basis of additional confirming evidence. The rules are automatically generated from a declarative (i.e., non-procedural) description of the grammar and semantics, and are embedded in a parallel recognition network for efficient retrieval of applicable rules. The current grammar uses a 450- word vocabulary and accepts simple English queries for an information retrieval system. (Author)

Contents: Grammar; Semantics.

Contents: Grammar; Semantics.

Contents: Grammar; Semantics.

Contents: Grammar; Semantics.

Contents: Grammar; Semantics.

Contents: Parsers; Grammars; and Prosodics.

Contents: Parsers; Grammars; and Prosodics.

This document and overview of the SDC/NYU natural language system. Contents: grammar; noun phrase resolution; clause semantics', facilities.

This paper describes several experiments combining natural language and acoustic constraints to improve overall performance of the MIT VOYAGER spoken language system. This system couples the SUMMIT speech recognition system with the TINA language understanding system to answer spoken queries about navigational assistance in the Cambridge, MA, area. The overall goal of our research is to combine acoustic, syntactic and semantic knowledge sources. Our first experiment showed improvement by combining acoustic score and parse probability normalized for number of terminals. Results were further improved by the use of an explicit rejection criterion based on normalized parse probabilities. The use of the combined parse/acoustic score, together with the rejection criterion, gave an improvement in overall score of more than 33% on both training and test data, where score is defined as percent correct minus percent incorrect. Experiments on a fully integrated system which uses the parser to predict possible next words to the recognizer are now underway.

A merged lexicon for all domains has been created. The merged lexicon contains about 2400 distinct root words. A tool for extracting lexical items from a dictionary, given a word list, was used to obtain items from the merged lexicon for the resource management domain. Approximately 47% of the vocabulary (or 448 words) in this domain is new. However a large number of these words are proper names of ships and locations, which are relatively trivial to enter. There are about 225 other words, including about 50 verbs. Because there are so few verbs, we expect adding the lexical entries to take only a few days. Keywords: Syntax, Semantics, Natural language, Information retrieval.

Click here to view the University of Florida catalog record

Click here to view the University of Florida catalog record

In an effort to document demographic changes in the academic linguistics community, the Linguistic Society of America (LSA) has commissioned an annual report detailing information about job titles, students, degrees awarded, average salaries, and more. The report offers information about trends in student enrollment and employment by industry. Data for the report was compiled primarily from self-reported information to the LSA's online Directory of Linguistics Programs and Departments. It also includes information based on U.S. federal surveys of PhD linguists and institutions, along with data gleaned from the LSA membership database. Trends in linguistics for this year include: (1) The most common career outcome for linguistics PhDs is a position in higher education; (2) Within higher education, departments report that most members of their faculty are full professors, but the non-professorial category is growing, particularly for women in other full-time and part-time positions; (3) The field of linguistics is growing most rapidly for undergraduates, with an increase of approximately 120 more students awarded BA degrees annually for the past 13 years; (4) Most linguistics undergraduate degrees are awarded to women. Although women represent just over half of graduate students in linguistics, a number which has been increasing over the last 50 years, women comprise almost 65 percent of the undergraduate population; and (5) The top five specializations for graduate students in linguistics are: Syntax, Applied Linguistics, Phonology, Semantics, and Historical Linguistics. [This second edition is based on the original report prepared by Lauren Friedman and Alyson Reed. The data found in this edition were compiled with assistance from Dan Foley, Robert Townsend, and Michal Temkin Martinez. For the 2013 report, see ED607296.]

We provide the syntax and semantics of the cat language, a domain specific language to describe consistency properties of parallel/distributed programs. The language is implemented in the herd7 too (http://diy.inria.fr/doc/herd.html)l.

We provide the syntax and semantics of the cat language, a domain specific language to describe consistency properties of parallel/distributed programs. The language is implemented in the herd7 too (http://diy.inria.fr/doc/herd.html)l.

Instruments constitute a classic member of the thematic role inventory, yet they are usually analyzed only peripherally, taking a back seat to the more studied members such as Agent and Patient. This dissertation investigates the semantic reality behind the label instrument from the functionalist perspective of Role & Reference Grammar. Starting from a theoretical investigation of what instrumentality truly means when contrasted with related concepts like comitatives, this book explores the morphosyntactic realization of instruments across a wide range of typologically diverse languages. Apart from the standard occurrences of instruments that come to mind from languages such as Latin, German or English, this book delves into several less common constructions that feature the instrument relation. Such constructions include, amongst others, passives with instruments and particularly the Instrument-Subject Alternation, a construction where the instrument seemingly appears as the subject of the sentence. This construction displays variation along three dimensions: 1) The instrument can vary from a very simple tool to a complicated machine, 2) the predicate can vary substantially and 3) languages differ widely with respect to the construction’s acceptability. This makes for a complex playing field where the animacy of the instrument but also the aktionsart class of the predicate play a major role. The last section of this book deals with linking the semantics of instruments and related concepts to their morphosyntactic realizations, including the various encoding strategies that are available in any given language. This book also features a concise introduction to Role & Reference Grammar. Dissertations in Language and Cognition: This series explores issues of mental representation, linguistic structure and representation, and their interplay. The research presented in this series is grounded in the idea explored in the Collaborative Research Center ‘The structure of representations in language, cognition and science’ (SFB 991) that there is a universal format for the representation of linguistic and cognitive concepts.

In an effort to document demographic changes in the academic linguistics community, the Linguistic Society of America (LSA) has commissioned an annual report detailing information about job titles, students, degrees awarded, average salaries, and more. The report offers information about trends in student enrollment and employment by industry. Data for the report was compiled primarily from self-reported information to the LSA's online Directory of Linguistics Programs and Departments. It also includes information based on U.S. federal surveys of PhD linguists and institutions, along with data gleaned from the LSA membership database. An overview of trends in linguistics for this year include: (1) The most common career outcome for linguistics PhDs is a position in higher education; (2) Within higher education, departments report that most members of their faculty are full professors, but the non-professorial category is growing, particularly for women in other full-time and part-time positions; (3) The field of linguistics is growing most rapidly for undergraduates, with an increase of approximately 120 more students awarded BA degrees annually for the past 13 years; (4) Most linguistics undergraduate degrees are awarded to women. Although women represent just over half of graduate students in linguistics, a number which has been increasing over the last 50 years, women comprise over 65 percent of the undergraduate population; and (5) The top five specializations for graduate students in linguistics are: Syntax, Applied Linguistics, Phonology, Semantics, and Language Acquisition. [This edition is based on the original report prepared by Lauren Friedman and Alyson Reed. The data found in this edition were compiled with assistance from Robert Townsend and Dan Foley. For the 2014 report, see ED607300.]

In an effort to document demographic changes in the academic linguistics community, the Linguistic Society of America (LSA) has commissioned an annual report detailing information about job titles, students, degrees awarded, average salaries, and more. The report offers information about trends in student enrollment and employment by industry. Data for the report was compiled primarily from self-reported information to the LSA's online Directory of Linguistics Programs and Departments. It also includes information based on U.S. federal surveys of PhD linguists and institutions, along with data gleaned from the LSA membership database. An overview of trends in linguistics for this year include: (1) The most common career outcome for linguistics PhDs is a position in higher education; (2) Within higher education, departments report that most members of their faculty are full professors, but the non-professorial category is growing, particularly for women in other full-time and part-time positions; (3) The field of linguistics is growing most rapidly for undergraduates, with an increase of approximately 120 more students awarded BA degrees annually for the past 13 years; (4) Most linguistics undergraduate degrees are awarded to women. Although women represent just over half of graduate students in linguistics, a number which has been increasing over the last 50 years, women comprise over 65 percent of the undergraduate population; and (5) The top five specializations for graduate students in linguistics are: Syntax, Applied Linguistics, Phonology, Semantics, and Language Acquisition. [This edition is based on the original report prepared by Lauren Friedman and Alyson Reed. The data found in this edition were compiled with assistance from Robert Townsend and Dan Foley. For the 2014 report, see ED607300.]

This paper attempts to bring information to a broad audience on children's utilization of semantic and syntax context clues in reading. It reviews the research of several contemporary theorists who indicate that semantics and syntax are overlapping processes. The simultaneous use of these processes enable young, particularly struggling readers to compensate for problem-words and difficult language in semantically and syntactically inaccessible texts. English Language Learners' ability to construct meaning from texts written with English vocabulary and sentence structure is also discussed. Finally, the question of how to link theory to practice is raised, and more semantically and syntactically accessible research literature is recommended to aid in this endeavor. (Contains 7 references.) (Author)

This paper attempts to bring information to a broad audience on children's utilization of semantic and syntax context clues in reading. It reviews the research of several contemporary theorists who indicate that semantics and syntax are overlapping processes. The simultaneous use of these processes enable young, particularly struggling readers to compensate for problem-words and difficult language in semantically and syntactically inaccessible texts. English Language Learners' ability to construct meaning from texts written with English vocabulary and sentence structure is also discussed. Finally, the question of how to link theory to practice is raised, and more semantically and syntactically accessible research literature is recommended to aid in this endeavor. (Contains 7 references.) (Author)

This paper attempts to bring information to a broad audience on children's utilization of semantic and syntax context clues in reading. It reviews the research of several contemporary theorists who indicate that semantics and syntax are overlapping processes. The simultaneous use of these processes enable young, particularly struggling readers to compensate for problem-words and difficult language in semantically and syntactically inaccessible texts. English Language Learners' ability to construct meaning from texts written with English vocabulary and sentence structure is also discussed. Finally, the question of how to link theory to practice is raised, and more semantically and syntactically accessible research literature is recommended to aid in this endeavor. (Contains 7 references.) (Author)

This paper attempts to bring information to a broad audience on children's utilization of semantic and syntax context clues in reading. It reviews the research of several contemporary theorists who indicate that semantics and syntax are overlapping processes. The simultaneous use of these processes enable young, particularly struggling readers to compensate for problem-words and difficult language in semantically and syntactically inaccessible texts. English Language Learners' ability to construct meaning from texts written with English vocabulary and sentence structure is also discussed. Finally, the question of how to link theory to practice is raised, and more semantically and syntactically accessible research literature is recommended to aid in this endeavor. (Contains 7 references.) (Author)

The paper attempts to provide an introduction to the techniques and results which have come out of work in computational linguistics which have special relevance to the design of speech understanding systems. The author attempts to trace the development of several important ideas and trends in parsing and syntax and in semantic interpretation.

This report formally defines both a syntax and a semantics for the GSPML visual language. The syntax is defined with a hypergraph grammar and the semantics is defined with Plotkin style structural operational semantics. A decorated trace semantics is derived from the labeled transition system of the structural operational semantics. GSPML itself is motivated by shortcomings in the visual security modeling capabilities of the Model Driven Architecture (MDA).

This report formally defines both a syntax and a semantics for the GSPML visual language. The syntax is defined with a hypergraph grammar and the semantics is defined with Plotkin style structural operational semantics. A decorated trace semantics is derived from the labeled transition system of the structural operational semantics. GSPML itself is motivated by shortcomings in the visual security modeling capabilities of the Model Driven Architecture (MDA).

The quality of computer generated speech keeps getting better but people are still able to identify whether speech is produced by a human or not (Bruder & Larrouy-Maestri, 2023). The lack of humanness in synthetic speech might be linked to the naturalness or quality of its prosody. In order to understand this issue we focus on the role of syntactic and semantic aspects of language (which are known to be intrinsically linked to prosody (Cutler et al., 1997)) on the perception of humanness in speech. To do so, we propose an experiment testing the effects of voice type (natural versus synthetic) and of syntactic and semantic information on humanness perception. We created German sentences that were then manipulated to obtain three more conditions conveying syntactic but no semantic information (jabberwocky sentences), no syntactic but semantic information (wordlists), and no syntactic or semantic information (jabberwocky wordlists). This material was generated by text-to-speech tools (Microsoft, Elevenlabs, Murf, and Listnr) and recorded by human speakers. We will ask participants to rate these synthetic and natural spoken sentences on how human the voice sounds to them. Whether syntactic and semantic information interacts with voice category will inform us about the potential differences in processing synthetic and natural speech.

The Semantics of Questions and the Questions of Semantics: Case Studies in the Interrelations of Logic, Semantics, and Syntax. by Jaakko Hintikka Review by: David Harrah

The lexical reader has been modified to allow the PUNDIT Natural Language Processing system to recognize abbreviations ending in periods. PUNDIT previously had not been able to distinguish between periods ending abbreviations and periods ending sentences; consequently, it was not able to parse sentences with internal punctuation. Several meetings have been held to identify a discrete yet useful arena for developing a more flexible control strategy in syntactic and semantic processing of text. Currently, Pundit's control mechanism flows from the syntactic module. (Semantic data is used as a filter on the syntax, via the recently implemented selectional component, but this filtering does not change the control structure). The meetings have focused on the distribution of lexical information, which is dispersed both in the lexicon used by the syntactic parser and in the semantic lexical entries with their associated syntactic mapping rules. One proposal that has emerged would exploit the semantic component's potential to predict verb transitivity on the basis of selectional information and thereby handle the phenomenon of transitivity alternation, i.e., verbs which have transitive and intransitive uses. This would be useful because many verbs in English exhibit transitivity alternations.

In this paper, an attempt is made to explicate the notion "equivalence" in contrastive analysis. It has been suggested that a learner formulates successive hypotheses about the nature of the target language at least partially on the basis of his knowledge of the native language. A deep contrastive study of the two language systems will reveal the areas which need to be focused on in a teaching situation in order to discourage the learner from setting up "false" equivalences between the two systems, or rather, to lead him to set up valid equivalences. The paper advocates a view of contrastive analysis which is based on meaning rather than structure, and also takes into account areas such as pragmatics and conversational implicature in establishing equivalences. The method of such an analysis is illustrated by a detailed discussion of some of the properties of the causative constructions in English and selected South Asian languages. (Author)

Partial contents: Grammar; Syntax/Semantics Interaction; Semantics; Demo Environment; Change in Key Personnel; Darpa Meetings.

During the period 4/29/85 to 4/29/87, Unisys (formerly SDC) and New York University have pursued joint research on natural-language text understanding, as part of the DARPA Strategic Computing Program in support of the Fleet Command Center Battle Management Program (FCCBMP). During this time, two groups have collaborated on Proteus-I, a natural-language text-understanding system that processes the remarks field of messages (casualty reports or CASREPs) about failures of starting air compressors (SACs). The goal of the research is to demonstrate a system capable of understanding paragraph-length messages in a restricted domain. Features of this joint Unisys/NYU system are (1) Creation of an integrated system for the detailed understanding of text; the system includes modules for syntax, semantics, reference resolution, and temporal processing; (2) Grammatical coverage providing the correct parse for 90% of CASREP sentences, including a full treatment of co-ordinate conjunction and an integrated treatment of sentenial fragments characteristic of message traffic; (3) Automatic production of a tabular summary of diagnostic information for sample CASREPs; (4) Construction of a qualitative model to starting air compressor; and (5) Use of the qualitative model to simulate normal and faulty behavior described by natural-language phrases.

We provide the syntax and semantics of the LISA (for "Litmus Instruction Set Architecture") language. The parallel assembly language LISA is implemented in the herd7 tool (http://virginia.cs.ucl.ac.uk/herd/) for simulating weak consistency models.

This collection of abstracts is part of a continuing series providing information on recent doctoral dissertations. The 12 titles deal with a variety of topics, including the following: the relationship of text organization to readability and reading ability; how advance organizers, outlining, and other preparatory procedures affect learning and reading comprehension; how syntax and semantics affect the recall of connected discourse; a model of affective and cognitive silent reading strategies; the strategic processes involved in beginning reading; grade level differences in reading comprehension; and differences between recognizing words on a word list and words in context. (RL)

This collection of abstracts is part of a continuing series providing information on recent doctoral dissertations. The 12 titles deal with a variety of topics, including the following: the relationship of text organization to readability and reading ability; how advance organizers, outlining, and other preparatory procedures affect learning and reading comprehension; how syntax and semantics affect the recall of connected discourse; a model of affective and cognitive silent reading strategies; the strategic processes involved in beginning reading; grade level differences in reading comprehension; and differences between recognizing words on a word list and words in context. (RL)

This collection of abstracts is part of a continuing series providing information on recent doctoral dissertations. The 12 titles deal with a variety of topics, including the following: the relationship of text organization to readability and reading ability; how advance organizers, outlining, and other preparatory procedures affect learning and reading comprehension; how syntax and semantics affect the recall of connected discourse; a model of affective and cognitive silent reading strategies; the strategic processes involved in beginning reading; grade level differences in reading comprehension; and differences between recognizing words on a word list and words in context. (RL)

Contents: Grammar; Syntax/Semantics Interaction; Semantics; Work on Multiple Domains; Environment; Sun Interface; Port to the Unisys Explorer.

Progress includes continuing to port Pundit to the Trident domain of maintenance reports while at the same time extending coverage in the RAINFORMS and CASREPS domains. Parallel development in the three domains has had a synergistic effect on the overall system, as features implemented for one domain become available for the others. Keywords: Natural language comprehension, Computer programming.

The PUNDIT User's Guide is intended to provide a concise and general introduction to the facilities of the PUNDIT text-processing system. The intended audience is computational linguists familiar with Quintus Prolog. While this document is not a reference manual, and does not in itself contain sufficient information for you to either extend the system or port it to a new domain, we have tried to cover the operational basics: how to run PUNDIT (section 2) and how to interpret PUNDIT's output (Section 3). In addition, Section 4 documents the two main procedures for accessing the system (parse and pundit), as well as a number of other procedures which we make frequent use of as developers. Appendix A and Appendix B will help you set the system up. Appendix D identifies the core and domain files, and Appendix E lists papers, presentations, and technical documentation available for PUNDIT. The User's Guide is designed to accompany a subset of the text-understanding software which has been developed at the Paoli Research Center, as it exists on the date of publication: the core components of PUNDIT, together with the domain-specific components developed to process Navy tactical messages (RAINFORMs). This domain will be referred to henceforth as the MUCK domain (an acronym for the message understanding conference which occasioned the development of the software).

This report is a detailed empirical examination of Suppes' ideas about the syntax and semantics of natural language, and an attempt at supporting the proposal that model-theoretic semantics of the type first proposed by Tarski is a useful tool for understanding the semantics of natural language. Child speech was selected as the best place to find data on natural language because it presents a view of the real problems represented by natural language, and because it allows the study of the process of language development. The main body of data consists of a series of recordings between a 32-month-old girl, Erica, and several adults. The ERIC corpus is found to be syntactically simpler and semantically more straightforward than adult speech. It is divided into utterances; its vocabulary is compared to ADULT vocabulary; a word frequency count is made; and its words are classified gramatically. A discussion follows of the standard concepts and results of the theory of generative grammars. A grammar for ERICA is devised, with special attention to lexical ambiguity. Mathematical syntax and semantics are discussed, followed by a description of ERICA semantics, with special reference to grammatical and semantic ambiguity. Conclusions include: (1) a reasonable probabilistic grammar for ERICA can be constructed; (2) the grammar GE1 is the best model for lexical disambiguation; (3) the grammar functions reasonably well semantically; (4) the notion of probability can play a key role in the construction of a semantics; and (5) simple set-theoretical functions are often successful in describing the ERICA semantics. (Author/AM)

This report is a detailed empirical examination of Suppes' ideas about the syntax and semantics of natural language, and an attempt at supporting the proposal that model-theoretic semantics of the type first proposed by Tarski is a useful tool for understanding the semantics of natural language. Child speech was selected as the best place to find data on natural language because it presents a view of the real problems represented by natural language, and because it allows the study of the process of language development. The main body of data consists of a series of recordings between a 32-month-old girl, Erica, and several adults. The ERIC corpus is found to be syntactically simpler and semantically more straightforward than adult speech. It is divided into utterances; its vocabulary is compared to ADULT vocabulary; a word frequency count is made; and its words are classified gramatically. A discussion follows of the standard concepts and results of the theory of generative grammars. A grammar for ERICA is devised, with special attention to lexical ambiguity. Mathematical syntax and semantics are discussed, followed by a description of ERICA semantics, with special reference to grammatical and semantic ambiguity. Conclusions include: (1) a reasonable probabilistic grammar for ERICA can be constructed; (2) the grammar GE1 is the best model for lexical disambiguation; (3) the grammar functions reasonably well semantically; (4) the notion of probability can play a key role in the construction of a semantics; and (5) simple set-theoretical functions are often successful in describing the ERICA semantics. (Author/AM)

This report is a detailed empirical examination of Suppes' ideas about the syntax and semantics of natural language, and an attempt at supporting the proposal that model-theoretic semantics of the type first proposed by Tarski is a useful tool for understanding the semantics of natural language. Child speech was selected as the best place to find data on natural language because it presents a view of the real problems represented by natural language, and because it allows the study of the process of language development. The main body of data consists of a series of recordings between a 32-month-old girl, Erica, and several adults. The ERIC corpus is found to be syntactically simpler and semantically more straightforward than adult speech. It is divided into utterances; its vocabulary is compared to ADULT vocabulary; a word frequency count is made; and its words are classified gramatically. A discussion follows of the standard concepts and results of the theory of generative grammars. A grammar for ERICA is devised, with special attention to lexical ambiguity. Mathematical syntax and semantics are discussed, followed by a description of ERICA semantics, with special reference to grammatical and semantic ambiguity. Conclusions include: (1) a reasonable probabilistic grammar for ERICA can be constructed; (2) the grammar GE1 is the best model for lexical disambiguation; (3) the grammar functions reasonably well semantically; (4) the notion of probability can play a key role in the construction of a semantics; and (5) simple set-theoretical functions are often successful in describing the ERICA semantics. (Author/AM)

This report is a detailed empirical examination of Suppes' ideas about the syntax and semantics of natural language, and an attempt at supporting the proposal that model-theoretic semantics of the type first proposed by Tarski is a useful tool for understanding the semantics of natural language. Child speech was selected as the best place to find data on natural language because it presents a view of the real problems represented by natural language, and because it allows the study of the process of language development. The main body of data consists of a series of recordings between a 32-month-old girl, Erica, and several adults. The ERIC corpus is found to be syntactically simpler and semantically more straightforward than adult speech. It is divided into utterances; its vocabulary is compared to ADULT vocabulary; a word frequency count is made; and its words are classified gramatically. A discussion follows of the standard concepts and results of the theory of generative grammars. A grammar for ERICA is devised, with special attention to lexical ambiguity. Mathematical syntax and semantics are discussed, followed by a description of ERICA semantics, with special reference to grammatical and semantic ambiguity. Conclusions include: (1) a reasonable probabilistic grammar for ERICA can be constructed; (2) the grammar GE1 is the best model for lexical disambiguation; (3) the grammar functions reasonably well semantically; (4) the notion of probability can play a key role in the construction of a semantics; and (5) simple set-theoretical functions are often successful in describing the ERICA semantics. (Author/AM)

This report is a detailed empirical examination of Suppes' ideas about the syntax and semantics of natural language, and an attempt at supporting the proposal that model-theoretic semantics of the type first proposed by Tarski is a useful tool for understanding the semantics of natural language. Child speech was selected as the best place to find data on natural language because it presents a view of the real problems represented by natural language, and because it allows the study of the process of language development. The main body of data consists of a series of recordings between a 32-month-old girl, Erica, and several adults. The ERIC corpus is found to be syntactically simpler and semantically more straightforward than adult speech. It is divided into utterances; its vocabulary is compared to ADULT vocabulary; a word frequency count is made; and its words are classified gramatically. A discussion follows of the standard concepts and results of the theory of generative grammars. A grammar for ERICA is devised, with special attention to lexical ambiguity. Mathematical syntax and semantics are discussed, followed by a description of ERICA semantics, with special reference to grammatical and semantic ambiguity. Conclusions include: (1) a reasonable probabilistic grammar for ERICA can be constructed; (2) the grammar GE1 is the best model for lexical disambiguation; (3) the grammar functions reasonably well semantically; (4) the notion of probability can play a key role in the construction of a semantics; and (5) simple set-theoretical functions are often successful in describing the ERICA semantics. (Author/AM)

This report is a detailed empirical examination of Suppes' ideas about the syntax and semantics of natural language, and an attempt at supporting the proposal that model-theoretic semantics of the type first proposed by Tarski is a useful tool for understanding the semantics of natural language. Child speech was selected as the best place to find data on natural language because it presents a view of the real problems represented by natural language, and because it allows the study of the process of language development. The main body of data consists of a series of recordings between a 32-month-old girl, Erica, and several adults. The ERIC corpus is found to be syntactically simpler and semantically more straightforward than adult speech. It is divided into utterances; its vocabulary is compared to ADULT vocabulary; a word frequency count is made; and its words are classified gramatically. A discussion follows of the standard concepts and results of the theory of generative grammars. A grammar for ERICA is devised, with special attention to lexical ambiguity. Mathematical syntax and semantics are discussed, followed by a description of ERICA semantics, with special reference to grammatical and semantic ambiguity. Conclusions include: (1) a reasonable probabilistic grammar for ERICA can be constructed; (2) the grammar GE1 is the best model for lexical disambiguation; (3) the grammar functions reasonably well semantically; (4) the notion of probability can play a key role in the construction of a semantics; and (5) simple set-theoretical functions are often successful in describing the ERICA semantics. (Author/AM)

This report is a detailed empirical examination of Suppes' ideas about the syntax and semantics of natural language, and an attempt at supporting the proposal that model-theoretic semantics of the type first proposed by Tarski is a useful tool for understanding the semantics of natural language. Child speech was selected as the best place to find data on natural language because it presents a view of the real problems represented by natural language, and because it allows the study of the process of language development. The main body of data consists of a series of recordings between a 32-month-old girl, Erica, and several adults. The ERIC corpus is found to be syntactically simpler and semantically more straightforward than adult speech. It is divided into utterances; its vocabulary is compared to ADULT vocabulary; a word frequency count is made; and its words are classified gramatically. A discussion follows of the standard concepts and results of the theory of generative grammars. A grammar for ERICA is devised, with special attention to lexical ambiguity. Mathematical syntax and semantics are discussed, followed by a description of ERICA semantics, with special reference to grammatical and semantic ambiguity. Conclusions include: (1) a reasonable probabilistic grammar for ERICA can be constructed; (2) the grammar GE1 is the best model for lexical disambiguation; (3) the grammar functions reasonably well semantically; (4) the notion of probability can play a key role in the construction of a semantics; and (5) simple set-theoretical functions are often successful in describing the ERICA semantics. (Author/AM)