It is with great sadness that we report the passing of Martin Kay in August 2021. Martin was a pioneer and intellectual trailblazer in computational linguistics. He was also a close friend and colleague of many years.Martin was a polyglot undergraduate student of modern and medieval languages at Cambridge University, with a particular interest in translation. He was not (yet) a mathematician or engineer, but idle speculation in 1958 about the possibilities of automating the translation process led him to Margaret Masterman at the Cambridge Language Research Unit, and a shift to a long and productive career.In 1960 he was offered an internship with Dave Hays and the Linguistics Project at The RAND Corporation in California, another early center of research in our emerging discipline. He stayed at RAND for more than a decade, working on basic technologies that are needed for machine processing of natural language. Among his contributions during that period was the development of the first so-called chart parser (Kay 1967), a computationally effective mechanism for dealing systematically with linguistic dependencies that cannot be expressed in context-free grammars. The chart architecture could be deployed for language generation as well as parsing, an important property for Martin’s continuing interest in translation.It was during the years at RAND that Martin found his second calling, as a teacher of computational linguistics, initially at UCLA and then in many other settings. He was a gifted and entertaining speaker and lecturer, able to present complex material with clarity and precision. He took great pleasure in the interactions with his students and the role that he played in helping to advance their careers. He left RAND in 1972 to become a full-time professor and chair of the Computer Science Department at the University of California at Irvine.His time at Irvine was short-lived, as he was attracted back to an open-ended research environment. In 1974 he joined with Danny Bobrow, Ron Kaplan, and Terry Winograd to form the Language Understander project at the recently created Palo Alto Research Center (PARC) of the Xerox Corporation. The group took as a first goal the construction of a mixed-initiative dialog system using state-of-the-art components for knowledge representation and reasoning, language understanding, language production, and dialog management (Bobrow et al. 1977). Martin took responsibility for the language production module, which was initially based on the quite rudimentary technology of the time.That was the beginning of his focus on “reversible grammars,” grammatical rules and representations that could be applied to parse strings into their underlying syntactic representations but also convert underlying representations back to the strings that express them. He and his colleagues at PARC developed the idea of hierarchical attribute-value structures (feature/functional structures) as underlying representations that could be characterized by the primitive predicates of equality and unification. This insight took form in his Functional Unification Grammar (Kay 1979) and in Lexical Functional Grammar (Kaplan and Bresnan 1982), and it also surfaced in the design of Head Driven Phrase Structure Grammar (Pollard and Sag 1987).Reversibility, for translation as well as dialog, was also the motivation at PARC for developing the mathematical, linguistic, and computational concepts that led to the use of bi-directional finite-state transducers for phonological and morphological description (Kaplan and Kay 1994). This technology is still being applied to a wide variety of language processing problems. But for Martin translation was always a central theme, bracketed by his early article “The Proper Place of Men and Machines in Language Translation” (which circulated in research for quite some time before it was finally published [Kay 1997]) and his most recent book (Kay 2017).In 1985 Martin struck a new balance between his commitment to research and his love of teaching by officially dividing his time between his prestigious role as a Research Fellow at PARC and a professorship in the Linguistics Department at Stanford. In addition to his Stanford professorship, he also taught (1998–2014) as an Honorary Professor at Saarland University, offering one or two courses every year. During his stays in Germany, he also advised on ongoing research, and his lectures and discussions helped in the gradual integration of programs in linguistics, computational linguistics, and translation studies.Martin contributed in many other ways to international progress in computational linguistics. In the 1970s and later he was a mainstay lecturer in the International Summer Schools in Computational Linguistics in Italy, and the Nordic summer schools in Scandinavia (actually, he and his wife Iris hosted one Nordic summer school at their home in Menlo Park). He advised research organizations and projects in several countries. He was a specialist advisor to the German Ministry of Education and Research, a reviewer for the two largest European projects in automatic translation, Eurotra and Verbmobil, and a valued advisor for projects at the German Research Center for Artificial Intelligence (DFKI). He also served for many years as chairman of the International Committee for Computational Linguistics (ICCL).Martin received many honors during his lifetime. He is a past President of the Association for Computational Linguistics (ACL). In 2005 he received the ACL Lifetime Achievement Award (Kay 2006). He was awarded honorary doctorate degrees from the University of Gothenburg (1982) and the University of Geneva (2008). He was the recipient of the Okawa Prize in 2019.Martin’s quiet and modest style of personal interaction stood only in apparent contrast to his widely recognized fame as an intellectual leader. His impressive expertise in several disciplines and his diverse intellectual interests made him a wonderful conversation partner for colleagues and friends who were lucky enough to be able to spend time with him. All students and colleagues remember him as a gifted speaker who was able to captivate and convince his audience with excellent didactics, rhetorical sharpness, and his very own sense of humor.We also remember Martin’s wife, Iris Kay, who predeceased him by a few months. Iris was a warm and psychologically insightful figure who played a prominent role in the early social history of computational linguistics, when personal relationships were more immediate and so important. They will both be sorely missed.

Abstract The universal generation problem for LFG grammars is the problem of determining whether a given grammar derives any terminal string with a given f-structure. It is known that this problem is decidable for acyclic f-structures. In this brief note, we show that for those f-structures the problem is nonetheless intractable. This holds even for grammars that are off-line parsable.

The formalism for Lexical-Functional Grammar (LFG) was introduced in the 1980s as one of the first constraint-based grammatical formalisms for natural language. It has led to substantial contributions to the linguistic literature and to the construction of large-scale descriptions of particular languages. Investigations of its mathematical properties have shown that, without further restrictions, the recognition, emptiness, and generation problems are undecidable, and that they are intractable in the worst case even with commonly applied restrictions. However, grammars of real languages appear not to invoke the full expressive power of the formalism, as indicated by the fact that algorithms and implementations for recognition and generation have been developed that run—even for broad-coverage grammars—in typically polynomial time. This article formalizes some restrictions on the notation and its interpretation that are compatible with conventions and principles that have been implicit or informally stated in linguistic theory. We show that LFG grammars that respect these restrictions, while still suitable for the description of natural languages, are equivalent to linear context-free rewriting systems and allow for tractable computation.

I want to begin by expressing my deep appreciation to the Nominating and Executive Committees and the Association at large for this tremendous honor. I am especially proud to join my collaborators and friends who previously received the Lifetime Achievement Award: Bill Woods, Martin Kay, Lauri Karttunen, and Joan Bresnan. They have profoundly influenced the contributions for which I think I am being recognized today, as have many other colleagues and students that I have worked with closely over the years.I will follow the tradition set down by most previous LTA recipients by describing how events in my personal history propelled me toward the concepts, theories, and algorithms I have helped to develop and have become known for. This historical perspective is quite different from a paper setting forth some current technical results.Some former recipients, Martin Kay for example, were intrigued by language from a very early age. That was not the case for me. I started as an undergraduate at the University of California Berkeley with a major concentration in physics. I decided at one point that physical reality was a little too messy, and I also discovered that I was not interested in spending long hours in a laboratory. So after a year or two I switched to math. That carried me forward to the middle of my fourth and final year, when I began to contemplate a future as a mathematician. That also seemed a little dry and unappealing. I thought I should look around, even at that late date, and maybe find a specialization that was a little more social and, for me, a little more engaging.I mentioned this to a friend during a brief ride in an elevator. He told me about a professor in the Psychology Department who was offering an individual major in what he called “language behavior.” The major wasn’t actually inside a department because it combined material from a bunch of separate disciplines: psychology, linguistics, philosophy, anthropology, and maybe even mathematics. That sounded pretty interesting, and an important property at that late stage of my undergraduate career was that it had no official prerequisite courses.The professor, Dan Slobin, was an expert in child language acquisition, and, as I later learned, child language acquisition played a prominent ideological role in the newly popular approach to linguistics, transformational grammar. I looked at some of the courses that other students had put into their individual-major schedules and at some of the core readings in psycholinguistics that they had included. The field was brand new, it didn’t have the long history of other disciplines, and therefore not a lot of material to be mastered before jumping in. And I realized that this was a neat application of formal mathematical methods to a collection of humanistic problems. All to the good—I signed up with Dan.I also decided to take the few remaining math courses that I needed to complete that major as well. So at that point I guess I had become, if only by accident, a mathematical psycholinguist. I was awarded a double-major undergraduate degree, and I was admitted to graduate school in social psychology, which is where psycholinguistics was located at Harvard.I encountered a few high-level ideas during that undergraduate period that have reverberated throughout the course of my career. Chomsky (1965) Aspects of the Theory of Syntax had appeared just a few years before. It provided some formal details of the first main revision to the basic machinery of transformational grammar, but that wasn’t what struck me. Rather, there was a simple statement at the beginning that set forth a particular view of the relation between a speaker’s knowledge of language, the result of learning a language, and the ability to put that knowledge to use in comprehension and production. He said:No doubt, a reasonable model of language use will incorporate, as a basic component, the generative grammar that expresses the speaker-hearer’s knowledge of the language; but this generative grammar does not, in itself, prescribe the character or functioning of a perceptual model or a model of speech production. (Chomsky 1965, page 9)This is a clear statement of what became known as the Competence Hypothesis. I understood it as making two claims: First, that a representation of linguistic knowledge should play a causal role in the mental processes of everyday communication by language. And second, that linguistic knowledge was only one of many components of a larger model of language performance.The first claim, I thought, distinguished this application of formal descriptions from the way that mathematical characterizations are deployed in most other sciences. The sun, for example, behaves according to the partial differential equations for energy transport, but presumably it does not solve a representation of those equations in order to decide what to do next. But the proposition here is that the psychological system encodes and then interprets something that correlates with generalizations about the properties of language that emerge from linguistic investigation. That would be a very neat state of affairs, if it were true.The second claim places this in the context of another powerful conception, the notion of a “nearly decomposable system” that Herb Simon put forward in a short essay entitled The Architecture of Complexity (Simon 1962). Simon argued that apparently complicated systems, if created artificially or through an evolutionary process, are likely to be recursively decomposable into collections of separate components. Such systems appear complicated because these components interact in non-negligible ways. But those interactions are relatively simple compared to the interaction of elements within the individual components. It is thus fruitful, Simon would say, to construct scientific explanations of such systems by treating each of the components independently at first, and then layering on top an account of their interactions. Chomsky’s statement was with this of decomposable he to the That became the of some of my will have that my not a of or I had in during my had a the of a very in those And I had methods courses in the math But there was as no between and language I had up at an in the of there was a between Berkeley and up at I to to look for a another another friend with to the told me that had a linguistic that maybe I up I in my and for an I that they have an in mathematical of and that I on as a was the of a that Martin Kay, and Lauri was there just that was one of the of He the helped to the of the and was in setting up the He was also a very in many of He was an early for grammar, for example, as a to transformational grammar and And in my case I that he me to about and (Chomsky I of course that were And from Berkeley I thought that the for But I to at a little my first me down the to with Lauri on some I was to a within a I was also in linguistic into some properties of in which and descriptions be not to be I to after a few and told that had that to on a to it to interact with told me to down the in the other and what I from Martin Martin had a that was not on transformational grammar but seemed to the of language It was a system for of into of with But it a very called the to the state of the was a Martin had the but he didn’t have an grammar to That became my for the of the and in some for the of my That was how I about that was my into to an and I to my at was the in child language acquisition at the and language acquisition was a at a first I decided to were to some of the that Lauri and I had worked The what I to and the that I to play the of that I decided I didn’t have the or needed to with I to Chomsky’s in had a of comprehension that transformational grammar as of linguistic They the simple if that were the the of comprehension to be to the of that in the linguistic of a This called the Theory of received some The of that were at but it appeared in early that and two were more for to with simple and the application of an had the of psychological for example, this in the case of the that the of a The and to the late had more or up on the Theory of they had up on Chomsky’s that a model of would a representation of the speaker’s They began to independently of linguistics, to of language by just from to in a collection of what they called that the first in an is as the worked for simple but not for thus for the that to be to The of was that this not how be understood at was no when a to the Theory of Complexity not to a collection of perceptual seemed an The I thought, was not in the in the Competence that knowledge of language should be in Rather, the was in a the transformational were was that but I from my with that there be other and some that are more as a grammar me from to was up at and to a language that would to on the that had from the Bill was the and one of the components was the that he had is an grammar, as in of a collection of each of which how a of a be and a that for one to be in are with that in called and are forward to what on later a a final in the is to a to the deep that a transformational grammar to that over the of an grammar in the and that was my to the first of the system and and the and other for the second and and and grammar are at in LTA paper and The and of the grammar the of the to relatively simple descriptions of complicated also had the order of for the transformational the is for That seemed on the that a encodes separate this the of the language to as many generalizations as the of the language on the at the order of is in the set from the a through the of the only the and that are at each This seemed to with the view that the language is an evolutionary between the that be to to and to an early paper describing how the and a simple grammar be to model the of some of the perceptual I that those be by the order of each state and for when a does not to a colleagues and I the approach in a different The in are in of their and they the But they in their psychological some and a the to the in and The different in up in that in the early is before the later one is that the is a psychological and set to what called the set of was to me a The in are the for the of the different properties of told and the is at the in but a representation of the in the is should be at in and the should at in this was a as a of and a of these with more in the of and for this with other and a different my I also a particular for with components that were relatively and be the of (Simon 1962). Such a model would a grammar in some as a of linguistic a that to that an of order and to a presumably a of that into of individual components should be as to how they into an but they should also be as in their The grammar as a of linguistic knowledge should be a of linguistic the should of a of linguistic and me to some basic to I the state of an was by on the and in a collection of were to at the of a in of linguistic I that was in this because it was in the or because there was no in a for some of the to the for from at the of each and to those at of be and many be by the representation at each of and for the of to deep I the collection of as a as the of the the from a to an representation for a with an that the was by the of this are the of what became the of and the of other second a in the of and that be to set and the of during a of is the way of the in to in the that the for the and the first that the is the and the is the This guess is when the is by an as in a is encountered after as in the to an set up to that the are that the the what was as the the and the after by the final the the the and are the with the and the of the as in of this is that it no over or of just the ability to The perceptual be to this simple down in the of didn’t was by wasn’t in the with the not the the main is or and the in the is by the of the main The of those be if the is to be The for the is by the in that that of the main and in when the first is understood as the of the of the first is this would follow from the of each with in the order of in But the of is and there is no for the to with through from these and other that the set of and and should be by one to two provided they do not and that that be and I called this the It that two have the and This was the of the of and the in and other in that early account of perceptual on a of each state in the grammar. and a quite that that this was not a about the psychological reality of transformational grammar, they to were to in as in and in are are are to They into the details of their and that a of one is to after of the with of the It that comprehension be influenced by very would not want to that the knowledge of is and Rather, order should not be in the grammar but as an and This is the separate of the I have an I think such a for that be on and deep learning if and are or even account also on the of the from a be very and the that would be It was known that be for reasonable by a for previously and partial That is a very for It seemed reasonable to think that the psychological system would also such a to the of I down the to and then combined it with to the partial Martin and I later a and more of this which later called an He this also in LTA paper was the of my at I have mentioned a few core ideas that from these and the for with separate process, and as an representation for a for on and and the of for previously recognized ideas are the for of my later on the He was the for at and He to a language at in He me to join the when it looked I was about to my a little of I how long would have before to their He that would be one of their and that had a I thought I I that he also Martin Kay, that at the in the of I to forward with at but I that it was to have an in linguistic and is the language started set to on different components of an and worked on knowledge and Martin and I on and This was in of a in to the that at were to the separate components into a after a few years of the components a to with the non-negligible interactions of components that the of individual The result was the early system by and I had the at of being to and linguistic over years were that was not being and that not have to for had become a for many language Martin and I argued at to and on the most and for And had and about the of and on as the basic of and something to the as the for property it to an to of the that were being for some early to speech But had different and about how to from Martin was to the that the grammar be as and that of a the grammar recursively with of the The is to the of and as would Herb Simon as a I a separate grammar as a system of in the use of to the of and only as the I thought of the as just a that or not and not an by a of I would that I the model of the Martin the Martin into what became I will how the of from the approach that I was on approach that the needed a for and for and were but set a more a one that would not only for but also for that the of an approach that would the of and just as as at that were in the as of They started from and and in an were of the that a in a language be to in if it after a context in a language and before a context in needed an for individual and then the This to be a The for individual the looked to be more complicated the needed for that be Syntax is is had to be a realized that the of such a grammar is if an individual is not to to at a in a or in a of the realized that the that of each individual be by a and that a of be by the of their that there a to each was not the as how to construct such a for a with that on and for a of to the with and and up with what looked the But even a simple would up with quite a few and a large of and would a or or maybe even a that the the was to The was to and only the of the context in and and to that the only when by the This is what years of had one of those in on a was to a a into an a needed a that would and in a context or beginning with a not just some of in that there was an to the simple and between and in first order or to simple in the is the in of these and these over by to properties of and the are and the in are a language that a only if in is by a a for the in the language by is by a in are and the of these two and only where are And of course there is a that that language. this were to an for of and that the had the This high-level also to construct a for for and the and in the early the realized that the also be as a relation and be at the top of the was then no for separate This worked for most of the with simple or to be a It was in and in to the of as a or as a collection of at and later a The if the of is as a early in the as the very The order of be because of the of and the result is a with of and the and and The even for is an from to and a this an way of the generalizations in a collection of to in of with the but with different of different for and different of in to up these because of other of it this was not only as a component, it had many the algorithms and with of large worked with a to the in for the of to a from other to of the I to point that in about years of with created a of and even on this as in also the in that were to most of the of that for and in in Martin and I to a of and Kay are that it was recognized as a for the was a and from the of year after I to I to to in a of that and were at They had received a large from to stage a to the of the of the They me and my other and and a of where Joan and I for the first a over the course of a year, with each a or to their and to a and I to transformational grammar in one another Joan of as a way of describing many within a in the and at in the and to some and develop some that and even the of as for I realized that solve some and that in after had to in in the of the in in and and with quite a few were to course was called and I the approach within the and formal that I was from the early course was on transformational grammar. I in course and in and at the of the interactions as in more and more ways. of were years how was when Joan and I deep into a when should have at the from the at But when Joan and I decided to for to during the year, and for me to have a at in the of decided to and a course called that would psychology, and from a no started with the I had at by of a The was to that linguistic generalizations be in that and those by had to some had to in to the had just the from helped to the that at the core of the I a to the of the formal set The and the deep in are of different in a and they are for the of the are as in the mathematical for collections of and the that of the The on the are at the to a that an the of a that from of the to of the on a the to the on the to in this and to the to the from this be a model for the of the equations in and the from the other The in this The to from the that the historical is here as mathematical and the to is the formal of as it from that course and and this of the is after This of formal descriptions from and was and in and it as a for the interactions between and other relatively of linguistic and and use of a between and to descriptions of and a to in the approach to another application of this and have the interaction of and by of a particular of and are of the and it from and other this were even to a particular That created the and to for This is the that I started to on after I to and that I in with was as an and a was to the grammar. algorithms were combined in the by to the in a This was but not even and are was to the of The of by equations to appear and in And those emerge from the of by a grammar that be first the to that even from of a be But this is a if equations are to as thought should be the case for language. at a that for the of in equations their years of the to a very simple with a and is a of for which there is a and complete The of that only on the of of equations are in relatively the for a of is to the from to for not just to up It also for when components of different formal are in a decomposable It is not to and of when they are a and and take of this to of as they first from to and then to knowledge that of and also the for at on within one before the of other components or of context be into and I also for with the of and at a that in a different way for the that are not and The result was a and for and with called it the I am with on a mathematical of the that a formal for the of the in the mathematical and properties of the for and and with colleagues at the University of and the in created the was to in for a of and to to and are the many in many who worked to those and to the I also that worked with as a and one of the early of the grammar. And an important later to for and was an for of by and and the grammar were the language for the years later at a by have in my career to on in psychology, linguistics, and with some of the very the I the Theory and at The and LTA many of the other I have The I for the to be very I have here just some of the contributions to and that were to over the years of That is quite a long in years.I for when language seemed to for and that the of the of I have more as a technical at and that language are needed to the of the systems that are in their everyday The in the of deep learning with the of that be are a major But to be such systems of many different and from many different and of it be in interactions. by even when not be the most or way of the of language, interaction application or the at on interaction are more or only from the knowledge and that have by scientific and and from that be to one of my as for of language should that systems, and other are decomposable and that their are of I think the scientific and will be to and the non-negligible interactions of these components. It will take in to construct systems that are and to interact want to by the and the Association for this tremendous honor.

This article describes an approach to Lexical-Functional Grammar (LFG) generation that is based on the fact that the set of strings that an LFG grammar relates to a particular acyclic f-structure is a context-free language. We present an algorithm that produces for an arbitrary LFG grammar and an arbitrary acyclic input f-structure a context-free grammar describing exactly the set of strings that the given LFG grammar associates with that f-structure. The individual sentences are then available through a standard context-free generator operating on that grammar. The context-free grammar is constructed by specializing the context-free backbone of the LFG grammar for the given f-structure and serves as a compact representation of all generation results that the LFG grammar assigns to the input. This approach extends to other grammatical formalisms with explicit context-free backbones, such as PATR, and also to formalisms that permit a context-free skeleton to be extracted from richer specifications. It provides a general mathematical framework for understanding and improving the operation of a family of chart-based generation algorithms.

We present a stochastic parsing system consisting of a Lexical-Functional Grammar (LFG), a constraint-based parser and a stochastic disambiguation model. We report on the results of applying this system to parsing the UPenn Wall Street Journal (WSJ) treebank. The model combines full and partial parsing techniques to reach full grammar coverage on unseen data. The treebank annotations are used to provide partially labeled data for discriminative statistical estimation using exponential models. Disambiguation performance is evaluated by measuring matches of predicate-argument relations on two distinct test sets. On a gold standard of manually annotated f-structures for a subset of the WSJ treebank, this evaluation reaches 79% F-score. An evaluation on a gold standard of dependency relations for Brown corpus data achieves 76% F-score.

The questions before this panel presuppose a distinction between parsing and interpretation. There are two other simple and obvious distinctions that I think are necessary for a reasonable discussion of the issues. First, we must clearly distinguish between the static specification of a process and its dynamic execution. Second, we must clearly distinguish two purposes that a natural language processing system might serve: one legitimate goal of a system is to perform some practical task efficiently and well, while a second goal is to assist in developing a scientific understanding of the cognitive operations that underlie human language processing. I will refer to parsers primarily oriented towards the former goal as Practical Parsers (PP) and refer to the others as Performance Model Parsers (PMP). With these distinctions in mind, let me now turn to the questions at hand.