Flashcards
Week 2 - Siri,Chatbots & Intelligent conversation
Cards in this deck (15)
- 1What are the levels of functionality in Siri?Speech to Text
Text Analysis
Service Providers - 2What are some of the things Siri can do?Siri can do a variety of task, but based on a limited set of functions:
- interact with calendar
- search contacts
- read & write messages
- interact with maps & location services
- forward search phrases to - 3What is Eliza?ELIZA was a computer program written at MIT and one of the firs primitive attempts at NLP which produced quite realistic responses by analyzing 2 things:
Syntax: Grammatical Analysis of Sentences
Semantics: Logical representation of Meanings - 4What are the key ideas of Shannon and Weaver on Statistics & Language?Shannon: printed English provides the means for statistical language analysis
Weaver: Perfect translations is probably impossible but almost perfect (only x% errors) is almost certainly attainable through the use of statistical semantics. Translation could really be viewed as cryptography. - 5What is Chomsky's critique of Statistical Language modelsThat statistical language models are not based on insight and formulate meaning in the wrong way. Language is innate and cannot be learned from data - its far too complex
- 6Explain Chomsky's rule based approach to languageIt uses POS o Part-Of-Speech tagging to identify constructs in language and analyse how they apply to pre-prgrammed rules of language
- 7What are some of the reasons for the recent resurgence of the Empricism method of NLP?Technological developments are enabling:
1) large scale acquisition: of text
2) coverage: accounting for all phenomenon in a given domain
3) robustness: accommodating data that contains noise which was previously unaccounted for
4) extensibility: expanding or porting a system to new domains - 8What are examples of the challenges in Seech recognition?Homonyms, eg (youth in asia, and euthenasia)
The shortcomings of bi-gram and tri-gram models which only compare 2-3 words beside one another. (give 'the boy eats' + 'the boy ... eats' example - 9Define Speech recognitionMapping continuous speech signals into a sequence of recognizable words
- 10What are bi-grams and tri-gramsbi-gram and tri-gram models are currently used in virtually all speech systems. They are models in which the system remembers and analyses the previous 1 and 2 words respectively, to arrive at the next (2nd/bi or 3rd/tri) word
- 11What is syntactic analysisInvolves determining the grammatical structure of a sentence and determining for example when saw is a noun vs a verb based on the structure of the sentence it is in
- 12Explain the idea of a lexicalized parseThe idea is that all nodes in a tree contain information about the words in a sentence, not just the nodes immediately above the words
- 13What are the 2 sub-tasks of Semantic Analysis?1) Word Disambiguation - figuring which possible meanings for a word are correct (eg pen = writing tool or enclosure for animals)
2) Semantic parsing - mapping natural language into a formal representation of its meaning, ie case-role analysis (agent, patient, source, destination) - 14What is Discourse Analysis?Determining how larger intersentential
context influences the interpretation of a sentence. Includes 2 sub-tasks
(1) coreference resolution
and (2) information extraction. - 15What is machine translationInvolves translating text
from one natural language to another. This can use simple word substitution or use tightly or loosely coupled corpora to determine the semantic relationships between two languages to derive translations
Information
- Description
- 1 Artificial Intelligence in the Marketplace (Week 2 - Siri, Chatbots & Intelligent Conversation) Flashcards on Week 2 - Siri,Chatbots & Intelligent conversation, created by Daniel Gomez-Windshuttle on 09/12/2015.
- Created by
- Daniel Gomez-Windshuttle
- Views
- 1
- Rating