Automatic Computer Reading and Human Reading (original) (raw)
Related papers
Parsing English. Course Notes for a Tutorial on Computational Semantics, March 17-22, 1975
1975
The course in parsing English is essentially a survey and comparison 'of several of the principal systems used for understanding natural language. The basic procedure of parsing is described. The discussion of the principal systems is based on the idea that "meaning is procedures," that is, that the procedures of application give a parsed structure its significance.' Natural language systems should be content-rather than structure-motivated, i.e. they should be concerned with linguistic problems revealed by parsing rather than with the relation of the proposed structure of the system to the structures of other systems. Within this framework, Winograd's understanding system, SHRDLU, is described and discussed, as are the second generation systems of Simmons, Schank, Colby and Wilks. A subsequent discussion compares all these systems. Concluding remarks outline immediate problems, including the need for a good memory model and the use of texts, rather than individual example Sentences, for investigation. (CLK) * * via the ERIC Document Reproduction Service (EDRS). EDRS is pot * responsible for the quality of the original document. Reproductions * * supplied by EDRS are the best that can be made from the original.
Human Sentence Processing: A Semantics-Oriented Parsing Approach
2007
and many others for valuable comments and discussions about various aspects of this thesis. Extra special thanks go to Barbara Hemforth again, who kept many undesirable distractions and burdens away from me during the hot phase of writing this thesis. I owe a special debt of gratitude to Nick Ketley, who was there when I needed him to put a nice Australian accent into my English writings. I would also like to thank Thomas Mulack and Christoph Hölscher for their support in running the eye-tracking experiments. Thilo Weigel, Andreas Kleegraf, Ari Vollweiter, and Martin Ernst Kraus did a wonderful job in adjusting the eye-tracker software to our purposes. Thanks and apologies to all those important people I forgot because I'm currently too tired to remember. 1. Actually, there are about 45 interpretations for the written version and more than 100 interpretations for the spoken version of the sentence (Altmann, personal communication) Processing models developed to explain the observed preferences differ in the extent to which they consider grammatical knowledge relevant for language processing. On one side, models have been developed which totally leave aside linguistic analyses of the language input and instead use agrammatical heuristics (Bever,1970; Schank, 1972; Herrmann, 1985). Theo Herrmann (1985), for example, claims that "in the normal case" a certain number of learned standard schemata about the canonical order-1.3.1 Serialism, parallelism, guidance and filter Although we seem to have access to human preferences when it comes to ambiguities, we still do not know when exactly during processing the preferred analysis is chosen. It might be the case that the parser calculates all syntactic alternatives at the location of an ambiguity in the first stage, but then decides to pursue only one of them, probably the best one, in the further analysis. That is to say that the parser uses certain heuristics in the second stage to filter out only very few or even only a single alternative. In that sense, PHA could be directly implemented as a second stage heuristic, filtering out the structural alternative favored by PHA. Note that the processes at both stages could apply at one and the same location in the sentence. Therefore, it 1.4 Overview The thesis is organized as follows: in chapter 2, I will present "state of the art" approaches in psycholinguistic research on human sentence processing, focussing on theories or models. Each model will be discussed in the light of general evidence on attachment preferences. The collection of models described includes Parametrized Head Attachment which has been motivated by evidence from self-paced reading studies on attachment preferences in verb-second and verb-final clauses. Since button pressing techniques suffer from several methodological weaknesses, as discussed in chapter 2, I will present data from eye-tracking studies on a similar set of materials as the self-paced reading studies in chapter 4.1.
Natural Language Understanding by Computer The Next Step
1986
A critical review of selected current research work in natural language understanding by computer is given. directing attention to some of the limitations of these working systems. Then some suggestions for possible "next " research goals are presented. followed by a discussion of two suggestions to facilitate meeting these goals.
… on Computers in …, 2003
This paper presents a method of ambiguity resolution for natural language processing in intelligent English learning support systems. Ambiguity of interpretation of sentences is one of the most important problems for intelligent language learning support systems which allow learners input composed sentences freely. Our system has a question and answer function which asks learners the contents of a story. Our method for the function resolves structural and semantic ambiguity by using the degree of agreement between results of natural language processing: syntactic and semantic information, of learners' answers and a story.
A new approach to text understanding
1992
This paper first briefly describes the architecture of PLUM, BBN's text processing system, and then reports on some experiments evaluating the effectiveness of the design at the component level. Three features are unusual in PLUM's architecture: a domainindependent deterrninistie parser, processing of (the resulting) fragments at the semantic and discourse level, and probabilistie models.