19 May Deixis and Context
Deixis, is a common type of ambiguity that is mediated at the level of context. When deixis occurs in written language, you can normally resolve questions about the identity of the person(s) referenced in the ‘he’, ‘she’ or ‘they’ pronoun. In speaking, you sometimes have to ask the person who said it. 3-DG places context in the stratum of pragmatics and simply searches neighboring sentences for information that can assist in resolving ambiguous referents. The thematic role slots in the context frame are filled as soon as they are ascertained from information available in the immediate sentence. When a role slot is filled by an ambiguous referent, neighboring sentences are searched for the data necessary to resolve the uncertainty.
|Understanding Context Cross-Reference|
|Click on these Links to other posts and glossary/bibliography references|
|Prior Post||Next Post|
|Discourse Pragmatics||Context Collapse: Communication Without Boundaries|
|Anaphora deixis||Chomsky 1968 Miller 1976|
|ambiguity context||Schank 1972 Lamb 1964|
|association rules||Winograd 1983 Sowa 1984|
Singular pronouns exhibit no syntactic ambiguity, but because they usually refer to unspecified entities, they need to be associated with a referent. This can often be done by ascertaining the role the pronoun fills in relation to its verb, and looking into adjoining phrases or sentences to find the referent based on relative position and context rules. In English, the antecedent can normally be found one or two sentences to the left of the deictic referent. In spoken, and some written Japanese, where both the pronoun and the antecedent are often omitted completely, listeners are often required to infer the identity completely from context.
Plural pronouns exhibit both morphological and semantic ambiguity. This is readily apparent in the English YOU. Syntactically, YOU is clearly a PRONOUN, but its plurality is mediated by context. Depending on the context in which it is used, it could be singular-definite, plural-definite, or even singular-indefinite, as in the phrase “if YOU thought about it” instead of “if ONE thought about it.” This kind of ambiguity can be resolved using deixis and context patterns as pragmatic rules.
Thus we see that language analysis, a process aimed at understanding the intent of the speaker or writer, can be analyzed at many different levels. Rules can be designed to discover the importance of patterns at each level, and between pairs of levels. The patterns can be defined as associations, almost like logical “IF -THEN” statements in which the recognized pattern implies the syntactic, morphological, semantic or pragmatic interpretation of the text. It all fits in context. I’ll address these processes again as I post in the upcoming sections on computing, software design, and knowledge-based processes.
Section 6 Summary
Understanding Context is supposed to be about cybernetics, and here we have wasted a whole section talking about talking. What’s the point? Since communication is the most complex of cognitive activities, we must be able to teach the computer to communicate intelligently before it can become sentient. We need real applications that can speak and listen.
How about machine translation? Fully automatic, high-quality machine translation has not yet been achieved. High-quality machine translation that is not fully automatic has only been achieved at unacceptably high costs. Multilingual systems are weak, and no multi-lingual system exists for omni-directional machine translation across widely divergent language systems such as occidental (western) to oriental (eastern) languages.
It is possible that this is due to weaknesses in the formalisms used to treat human languages. Most grammars focus on syntax. Some deal with both syntax and semantics (often fused to “semantax”), while others attempt to call semantics part of syntax. Many have provisions for morphology, but none has any kind of adequate mechanism for dealing with pragmatics. Most important, there is not a parser available today that has orthogonal mechanisms for dealing with all of these aspects of language, although some have ad-hoc appendages that attempt to solve certain problems associated with pragmatics and semantics.
MIPUS’s designers gave him the ability to understand spoken commands so he could serve as a truly useful household helper. In Section 7, we describe the models they used. In Section 8, we go into greater detail about turning the models into real applications. Finally, in section 9, we introduce the linchpin of intelligent technologies – an object centered computational model that can understand language using 3-DG rules and neuromorphic processes.
These are my favorite references in this subject area:
Language and the Mind (Chomsky, 1968)
On Alternation, Transformation,Realization and Stratification (Lamb, 1964)
Language and Perception (Miller, 1976)
Conceptual Dependency: A Theory of Natural Language Understanding (Schank, 1972)
Conceptual Structures (Sowa, 1984)
Language as a Cognitive Process (Winograd, 1983)
|Click below to look in each Understanding Context section|
|4||Perception and Cognition||5||Fuzzy Logic||6||Language and Dialog||7||Cybernetic Models|
|8||Apps and Processes||9||The End of Code||Glossary||Bibliography|