Henk Zeevat - Bayesian Natural Language Semantics and Pragmatics

Bayesian Natural Language Semantics and Pragmatics — read online for free the complete book (whole text) full work

Below is the text of the book, divided by pages. System saving the place of the last page read, allows you to conveniently read the book "Bayesian Natural Language Semantics and Pragmatics" online for free, without having to search again every time where you left off. Put a bookmark, and you can go to the page where you finished reading at any time.

Light

Font size:

↓

↑

Georgia Tahoma Arial Verdana

Reset

Interval:

↓

↑

Bookmark:

Make

1234567...116

Springer International Publishing Switzerland 2015

Henk Zeevat and Hans-Christian Schmitz (eds.) Bayesian Natural Language Semantics and Pragmatics Language, Cognition, and Mind 10.1007/978-3-319-17064-0_1

Perspectives on Bayesian Natural Language Semantics and Pragmatics

Henk Zeevat 1

(1)

ILLC, University of Amsterdam, Amsterdam, The Netherlands

Henk Zeevat

Email:

Abstract

Bayesian interpretation is a technique in signal processing and its application to natural language semantics and pragmatics (BNLSP from here on and BNLI if there is no particular emphasis on semantics and pragmatics) is basically an engineering decision. It is a cognitive science hypothesis that humans emulate BNLSP. That hypothesis offers a new perspective on the logic of interpretation and the recognition of other peoples intentions in inter-human communication. The hypothesis also has the potential of changing linguistic theory, because the mapping from meaning to form becomes the central one to capture in accounts of phonology, morphology and syntax. Semantics is essentially read off from this mapping and pragmatics is essentially reduced to probability maximation within Grices intention recognition. Finally, the stochastic models used can be causal, thus incorporating new ideas on the analysis of causality using Bayesian nets. The paper explores and connects these different ways of being committed to BNLSP.

Keywords

Bayesian interpretation Semantics Pragmatics Stochastic natural language interpretation Presupposition Implicature

Overview

This paper explains six ways in which one can understand the notion of Bayesian Natural Language Semantics and Pragmatics and argues for each of them that they merit consideration since they seem to allow progress in understanding human language interpretation and with respect to problems in the area of NL semantics and pragmatics. The six ways can be listed as in (1).

The paper follows (1) trying in each case to explain the concept and its merits, before embarking on some controversial consequences and some urgent issues. What is the best way to implement NL interpretation is the best way to implement NL interpretation and the brain is also what it is. Therefore (a) and (b) will have to be settled by future research. (c) and (d) however can be seen as a thesis about the nature of NL interpretation. (c) and (d) commit one to a Bayesian analysis of Grices views on meaning and the repercussions of those analysis for the notion of rational NL interpretation. (a) and (b), and especially the characterisation they give of the problem of NL interpretation as finding ones way through massive ambiguity would give important support to this thesis. The more speculative (e) and (f) deal with a counterargument to the thesis, i.e. the argument that it cannot deal with the causal component of the thesis.

Computational BNLSP

Bayesian Interpretation is an important technique in the AI discipline of signal processing, especially in speech recognition and computer vision. Its basis is Bayes theorem, stated in (2). The theorem analyses the posterior probability of hypothesis H given a signal S , p ( H | S ) as the product of the prior probability of the hypothesis p ( H ) and the likelihood p (( S | H ) divided by the probability of the signal p ( S ). It essentially gives an alternative way to find p ( H | S ) if p ( H | S ) cannot be reliably measured in a direct way.

This is simply derived as follows:

Therefore

and

If S is given, the most probable H for a given S written can be computed in a different way given Bayes theorem. , since if S is given. If one is searching for the most probable interpretation H of a signal S , it will therefore suffice to have a good estimation of the probability of H (called the prior, the probability assigned to H before the signal occurred) and of the probability p ( S | H ) (the likelihood of the signal given the interpretation). This way of finding the most probable H is useful and often unavoidable, since it can be difficult or even impossible to directly estimate p ( H | S ) (the posterior probability of the interpretation H , after the occurrence of the signal S ).

The difficulty for direct estimation is especially clear when the set of signals is infinite or continuous, as in speech perception or in vision. For example, in speech perception, one measures the distance between the perceived signal and the predicted signal to estimate likelihood and uses the prediction for the next phoneme from the preceding phoneme sequence obtained from a large body of written text to estimate the prior.

The definition just given makes BNLI a special way of obtaining the most probable interpretation of an utterance and BNLI is thereby a branch of stochastic signal interpretation.

One needs a good argument for using Bayesian interpretation rather than the direct method of estimating the posterior. And one needs to accept that natural language is ambiguous to go stochastic in the first place.

The method by itself does not entail that both likelihood and prior need to be estimated stochastically. The prior can be partly computed by symbolic techniques, e.g. by a theorem prover showing that the prior is not 0 by showing the consistency of the interpretation with the context. Similarly, one can use symbolic techniques (e.g. a text generator in NLIthese are standardly symbolic) to estimate likelihood.

Doing NLI or NLSP with this technique makes BNLSP a branch of stochastic NLSP. Stochastic NLSP obviously helps in obtaining good results in interpretation, since NL utterances severely underdetermine their interpretations in context and other techniques for disambiguation are limited in scope. It is also easy to see that very often disambiguation cannot be explained by an appeal to hard facts and logic, but needs to appeal to what is probable in the world. For example, if I tell you that I took the bus, you will understand me as having used the public transport system and not as having lifted an autobus with my crane. Another basic example is (3).

If you are like me, Tweety is the bird from AI discussions about default reasoning and flies means can fly using his wings. There must be many other Tweeties however and there is no logical reason why the verbal form could not mean uses aeroplanes or can pilot aeroplanes.

Next page

1234567...116

Light

Font size:

↓

↑

Georgia Tahoma Arial Verdana

Reset

Interval:

↓

↑

Bookmark:

Make

Similar books «Bayesian Natural Language Semantics and Pragmatics»

Look at similar books to Bayesian Natural Language Semantics and Pragmatics. We have selected literature similar in name and meaning in the hope of providing readers with more options to find new, interesting, not yet read works.

Frank Brisard

Handbook of Pragmatics: 25th Annual Installment

Chris Cummins

The Oxford Handbook of Experimental Semantics and Pragmatics

Siobhan Chapman

Pragmatics

Börjesson

The Semantics-Pragmatics Controversy

Huang

Pragmatics

Domnita Dumitrescu

L2 Spanish Pragmatics: From Research to Teaching

Pedro José Chamizo Domínguez

Semantics and Pragmatics of False Friends

Ali Almanna

Semantics for Translation Students: Arabic-English-Arabic

Peter Grundy

Doing Pragmatics

Sabbarton-Leary Nigel

The semantics and metaphysics of natural kinds

Betty J. Birner

Introduction to Pragmatics

Emma Borg

Pursuing Meaning

Discussion, reviews of the book Bayesian Natural Language Semantics and Pragmatics and just readers' own opinions. Leave your comments, write what you think about the work, its meaning or the main characters. Specify what exactly you liked and what you didn't like, and why you think so.