Publisher Summary This chapter presents an analysis of ASL, the signed language of the American D... more Publisher Summary This chapter presents an analysis of ASL, the signed language of the American Deaf., phonology, which focuses on the information structure of the sign. General phonological theory should operate in a uniform fashion across modalities and provide theoretical units that play similar, though not identical, roles in the various modalities that may underlie human language. The chapter focuses on analogy that should be established between the coda of a spoken language syllable and the second/weak hand of a two-handed sign. If the proposal is correct, then the cross-modality analogies that are to be made by the theory of phonology are, in a sense, more abstract than previous researchers had been led to believe. If there is in sign language something analogous to the syllable of spoken language, then it does not carry over the sequential character of the syllable of spoken language.
One of the popular beliefs that we found no justification for was that the failure of Generative ... more One of the popular beliefs that we found no justification for was that the failure of Generative Semantics was a consequence of a measurable weakness in its theories in comparison with those of its contemporary rival, Interpretive Semantics. After examining in some detail the arguments of both sides during the Generative Semantics period and attempting to extract from them the elements that bore most directly upon their empirical or logical validity, we concluded that the demise of Generative Semantics must have followed from factors essentially external to theory. More particularly, we concluded that, although each side in the disputes claimed repeatedly to
In this paper, we define a group structure over strings and note briefly that by applying this co... more In this paper, we define a group structure over strings and note briefly that by applying this computation to words, we obtain major steps towards a method for identifying allomorphy and learning morphophonemics. First order differences among a set of words forming a paradigm identifies morphs, while second order differences identifies allomorphy. When this allomorphy appears at morpheme boundary, this can in a wide range of cases be identified as morphophonology. 1 A group structure for strings 1.1 Defining string difference It is common to define a semi-group structure for concatenation of symbols in an alphabet A. This simply assumes a finite set of symbols, which we may indicate mnemonically as a..z or A and which we call letters, plus a binary concatenation operator represented by • when explicit, and simply by juxtaposition when we do not need to be explicit. Concatenation is associative and not commutative. The set of all finite concatenations of elements of A is A, which inc...
The subject which I would like to treat in this paper is the evaluation metric in generative gram... more The subject which I would like to treat in this paper is the evaluation metric in generative grammar. Why? Arguably, the evaluation metric is both the most novel and the most important concept in the development of generative grammar by Noam Chomsky. And yet it is at the same time one of the least recognized and surely most misunderstood of the core concepts of generative grammar. So there you are: the evaluation metric is critically important, it is arguably novel, it is misunderstood, and at some times and in some places, it has even been reviled. What better reasons could there be for spending our time today talking about it? I would like, first, to explain the idea of the evaluation metric in early generative grammar; this will mean exploring the separate ideas of (1) a prior over the set of grammars and (2) a measure of goodness of fit to the data. Second, I will very briefly trace how those two ideas have been developed in the world of machine learning over the last few decade...
The purpose of our experiment was to get an estimate of the usefulness of an automatic morphologi... more The purpose of our experiment was to get an estimate of the usefulness of an automatic morphological analyzer that we have been working on. The automatic morphological analyzer (called AutoMorphology) accepts a text of any length and automatically determines the stems and suffixes of the words in the language. For a full description of the algorithm, see http://humanities.uchicago.edu/faculty/goldsmith/ Linguistica2000/Paper/paper.html. There has been discussion in the IR literature of the tradeoffs between using automatic language-independent means for stemming and language-dependent (hand-written) stemming algorithms{Xu and Croft 1995, Lennon et al.1981}; our goal has been to get the best of both approaches by using an automatic but language-particular stemming algorithm.
The Canadian Journal of Linguistics / La revue canadienne de linguistique, 2006
This chapter explains how Perl works and introduces the edit-run cycle for readers with no backgr... more This chapter explains how Perl works and introduces the edit-run cycle for readers with no background in programming. I begin with how to edit a file using any number of editors, and go on to explain how to compile and run the programs we write. 2.1 Edit and Run Just in case you've never written a computer program in your life, let's go over the basic idea. A programming language allows you to issue instructions to your computer. In effect, it is a lingua franca, a mediating language. You translate your ideas into it and the computer translates it into something it can understand: machine code. The process of writing up your program in the programming language is the edit phase. Once you've written out your program in Perl, you then convert it to machine code and run it using the perl command. This is referred to as the run phase. Let's go through each of these in turn. 2.1.1 Edit You need to create your program using some sort of text editor. In principle, you can use any editor, but it's easiest to use a very simple one. There are a number of possibilities and I list some of them below. The key component is that the file you create should be saved as a text file with the extension ".pl".
This article describes in detail several explicit computational methods for approaching such ques... more This article describes in detail several explicit computational methods for approaching such questions in phonology as the vowel/consonant distinction, the nature of vowel harmony systems, and syllable structure, appealing solely to distributional information. Beginning with the vowel/ consonant distinction, we consider a method for its discovery by the Russian linguist Boris Sukhotin, and compare it to two newer methods of more general interest, both computational and theoretical, today. The first is based on spectral decomposition of matrices, allowing for dimensionality reduction in a finely controlled way, and the second is based on finding parameters for maximum likelihood in a hidden Markov model. While all three methods work for discovering the fairly robust vowel/consonant distinction, we extend the newer ones to the discovery of vowel harmony, and in the case of the probabilistic model, to the discovery of some aspects of syllable structure.
For each [language], one can ask the question, "Is it a pitch-accent language or a tone language?... more For each [language], one can ask the question, "Is it a pitch-accent language or a tone language?" However, I think that that is a stupid question to ask, since '" various characteristics of pitch-accent systems and of tonal systems are to a fair extent independent of one another and ... there is no reason for squeezing the diversity of phonological systems discussed here into a simple dichotomy.
We explore inflectional morphology as an example of the relationship of the discrete and the cont... more We explore inflectional morphology as an example of the relationship of the discrete and the continuous in linguistics. The grammar requests a form of a lexeme by specifying a set of feature values, which corresponds to a corner |M of a hypercube in feature value space. The morphology responds to that request by providing a morpheme, or a set of morphemes, whose vector sum is geometrically closest to the corner M. In short, the chosen morphemeˆ µ| is the morpheme which maximizes the inner product of µ| and |M. 9. There is a (for us, surprising) similarity between the way this leads us to talk about morphology and the way in which quantum mechanics is described. And as a measurement in physics corresponds to a morpheme realization, we have no reason to view that as a probabilistic operation; instead, a prepared system is realized by the morpheme that is closest to it, not probabilistically by all morphemes onto which it could project.
For each [language], one can ask the question, "Is it a pitch-accent language or a tone lang... more For each [language], one can ask the question, "Is it a pitch-accent language or a tone language? " However, I think that that is a stupid question to ask, since ' " various characteristics of pitch-accent systems and of tonal systems are to a fair extent independent of one another and... there is no reason for squeezing the diversity of phonological systems discussed here into a simple dichotomy. (McCawley 1978, 127-28). 1.
Two philosophers who disagree about a point should, instead of arguing fruitlessly and endlessly,... more Two philosophers who disagree about a point should, instead of arguing fruitlessly and endlessly, be able to take out their pencils, sit down amicably at their desks, and say "Let us calculate." Gottfried Wilhelm von Leibniz (1646 – 1716) Lest men suspect your tale untrue, Keep probability in view.
We address a key problem in grapheme-to-phoneme conversion: the ambiguity in mapping grapheme uni... more We address a key problem in grapheme-to-phoneme conversion: the ambiguity in mapping grapheme units to phonemes. Rather than using single letters and phonemes as units, we propose learning chunks, or subwords, to reduce ambiguity. This can be interpreted as learning a lexicon of subwords that has minimum description length. We implement an algorithm to build such a lexicon, as well as a simple decoder that uses these subwords.
This paper derives from work we have been doing on unsupervised learning of the morphology of lan... more This paper derives from work we have been doing on unsupervised learning of the morphology of languages with rich morphologies, that is, with a high average number of morphemes per word. Our focus in this paper is Swahili, a major Bantu language of East Africa, and our goal is the development of a system that can automatically produce a morphological analyzer of a text on the basis of a large corpus. While a certain amount of work in computational linguistics has already been done on Swahili, our specific goal is a system that can quickly and accurately perform a morphological analysis of any of the approximately 500 Bantu languages when presented with data from it, and little or no computational work currently exists for 99% of them. Our work reported here extends Linguistica, an open source system available at http://linguistica.uchicago.edu.
Publisher Summary This chapter presents an analysis of ASL, the signed language of the American D... more Publisher Summary This chapter presents an analysis of ASL, the signed language of the American Deaf., phonology, which focuses on the information structure of the sign. General phonological theory should operate in a uniform fashion across modalities and provide theoretical units that play similar, though not identical, roles in the various modalities that may underlie human language. The chapter focuses on analogy that should be established between the coda of a spoken language syllable and the second/weak hand of a two-handed sign. If the proposal is correct, then the cross-modality analogies that are to be made by the theory of phonology are, in a sense, more abstract than previous researchers had been led to believe. If there is in sign language something analogous to the syllable of spoken language, then it does not carry over the sequential character of the syllable of spoken language.
One of the popular beliefs that we found no justification for was that the failure of Generative ... more One of the popular beliefs that we found no justification for was that the failure of Generative Semantics was a consequence of a measurable weakness in its theories in comparison with those of its contemporary rival, Interpretive Semantics. After examining in some detail the arguments of both sides during the Generative Semantics period and attempting to extract from them the elements that bore most directly upon their empirical or logical validity, we concluded that the demise of Generative Semantics must have followed from factors essentially external to theory. More particularly, we concluded that, although each side in the disputes claimed repeatedly to
In this paper, we define a group structure over strings and note briefly that by applying this co... more In this paper, we define a group structure over strings and note briefly that by applying this computation to words, we obtain major steps towards a method for identifying allomorphy and learning morphophonemics. First order differences among a set of words forming a paradigm identifies morphs, while second order differences identifies allomorphy. When this allomorphy appears at morpheme boundary, this can in a wide range of cases be identified as morphophonology. 1 A group structure for strings 1.1 Defining string difference It is common to define a semi-group structure for concatenation of symbols in an alphabet A. This simply assumes a finite set of symbols, which we may indicate mnemonically as a..z or A and which we call letters, plus a binary concatenation operator represented by • when explicit, and simply by juxtaposition when we do not need to be explicit. Concatenation is associative and not commutative. The set of all finite concatenations of elements of A is A, which inc...
The subject which I would like to treat in this paper is the evaluation metric in generative gram... more The subject which I would like to treat in this paper is the evaluation metric in generative grammar. Why? Arguably, the evaluation metric is both the most novel and the most important concept in the development of generative grammar by Noam Chomsky. And yet it is at the same time one of the least recognized and surely most misunderstood of the core concepts of generative grammar. So there you are: the evaluation metric is critically important, it is arguably novel, it is misunderstood, and at some times and in some places, it has even been reviled. What better reasons could there be for spending our time today talking about it? I would like, first, to explain the idea of the evaluation metric in early generative grammar; this will mean exploring the separate ideas of (1) a prior over the set of grammars and (2) a measure of goodness of fit to the data. Second, I will very briefly trace how those two ideas have been developed in the world of machine learning over the last few decade...
The purpose of our experiment was to get an estimate of the usefulness of an automatic morphologi... more The purpose of our experiment was to get an estimate of the usefulness of an automatic morphological analyzer that we have been working on. The automatic morphological analyzer (called AutoMorphology) accepts a text of any length and automatically determines the stems and suffixes of the words in the language. For a full description of the algorithm, see http://humanities.uchicago.edu/faculty/goldsmith/ Linguistica2000/Paper/paper.html. There has been discussion in the IR literature of the tradeoffs between using automatic language-independent means for stemming and language-dependent (hand-written) stemming algorithms{Xu and Croft 1995, Lennon et al.1981}; our goal has been to get the best of both approaches by using an automatic but language-particular stemming algorithm.
The Canadian Journal of Linguistics / La revue canadienne de linguistique, 2006
This chapter explains how Perl works and introduces the edit-run cycle for readers with no backgr... more This chapter explains how Perl works and introduces the edit-run cycle for readers with no background in programming. I begin with how to edit a file using any number of editors, and go on to explain how to compile and run the programs we write. 2.1 Edit and Run Just in case you've never written a computer program in your life, let's go over the basic idea. A programming language allows you to issue instructions to your computer. In effect, it is a lingua franca, a mediating language. You translate your ideas into it and the computer translates it into something it can understand: machine code. The process of writing up your program in the programming language is the edit phase. Once you've written out your program in Perl, you then convert it to machine code and run it using the perl command. This is referred to as the run phase. Let's go through each of these in turn. 2.1.1 Edit You need to create your program using some sort of text editor. In principle, you can use any editor, but it's easiest to use a very simple one. There are a number of possibilities and I list some of them below. The key component is that the file you create should be saved as a text file with the extension ".pl".
This article describes in detail several explicit computational methods for approaching such ques... more This article describes in detail several explicit computational methods for approaching such questions in phonology as the vowel/consonant distinction, the nature of vowel harmony systems, and syllable structure, appealing solely to distributional information. Beginning with the vowel/ consonant distinction, we consider a method for its discovery by the Russian linguist Boris Sukhotin, and compare it to two newer methods of more general interest, both computational and theoretical, today. The first is based on spectral decomposition of matrices, allowing for dimensionality reduction in a finely controlled way, and the second is based on finding parameters for maximum likelihood in a hidden Markov model. While all three methods work for discovering the fairly robust vowel/consonant distinction, we extend the newer ones to the discovery of vowel harmony, and in the case of the probabilistic model, to the discovery of some aspects of syllable structure.
For each [language], one can ask the question, "Is it a pitch-accent language or a tone language?... more For each [language], one can ask the question, "Is it a pitch-accent language or a tone language?" However, I think that that is a stupid question to ask, since '" various characteristics of pitch-accent systems and of tonal systems are to a fair extent independent of one another and ... there is no reason for squeezing the diversity of phonological systems discussed here into a simple dichotomy.
We explore inflectional morphology as an example of the relationship of the discrete and the cont... more We explore inflectional morphology as an example of the relationship of the discrete and the continuous in linguistics. The grammar requests a form of a lexeme by specifying a set of feature values, which corresponds to a corner |M of a hypercube in feature value space. The morphology responds to that request by providing a morpheme, or a set of morphemes, whose vector sum is geometrically closest to the corner M. In short, the chosen morphemeˆ µ| is the morpheme which maximizes the inner product of µ| and |M. 9. There is a (for us, surprising) similarity between the way this leads us to talk about morphology and the way in which quantum mechanics is described. And as a measurement in physics corresponds to a morpheme realization, we have no reason to view that as a probabilistic operation; instead, a prepared system is realized by the morpheme that is closest to it, not probabilistically by all morphemes onto which it could project.
For each [language], one can ask the question, "Is it a pitch-accent language or a tone lang... more For each [language], one can ask the question, "Is it a pitch-accent language or a tone language? " However, I think that that is a stupid question to ask, since ' " various characteristics of pitch-accent systems and of tonal systems are to a fair extent independent of one another and... there is no reason for squeezing the diversity of phonological systems discussed here into a simple dichotomy. (McCawley 1978, 127-28). 1.
Two philosophers who disagree about a point should, instead of arguing fruitlessly and endlessly,... more Two philosophers who disagree about a point should, instead of arguing fruitlessly and endlessly, be able to take out their pencils, sit down amicably at their desks, and say "Let us calculate." Gottfried Wilhelm von Leibniz (1646 – 1716) Lest men suspect your tale untrue, Keep probability in view.
We address a key problem in grapheme-to-phoneme conversion: the ambiguity in mapping grapheme uni... more We address a key problem in grapheme-to-phoneme conversion: the ambiguity in mapping grapheme units to phonemes. Rather than using single letters and phonemes as units, we propose learning chunks, or subwords, to reduce ambiguity. This can be interpreted as learning a lexicon of subwords that has minimum description length. We implement an algorithm to build such a lexicon, as well as a simple decoder that uses these subwords.
This paper derives from work we have been doing on unsupervised learning of the morphology of lan... more This paper derives from work we have been doing on unsupervised learning of the morphology of languages with rich morphologies, that is, with a high average number of morphemes per word. Our focus in this paper is Swahili, a major Bantu language of East Africa, and our goal is the development of a system that can automatically produce a morphological analyzer of a text on the basis of a large corpus. While a certain amount of work in computational linguistics has already been done on Swahili, our specific goal is a system that can quickly and accurately perform a morphological analysis of any of the approximately 500 Bantu languages when presented with data from it, and little or no computational work currently exists for 99% of them. Our work reported here extends Linguistica, an open source system available at http://linguistica.uchicago.edu.
Uploads
Papers by John Goldsmith