Data processing: speech signal processing – linguistics – language – Linguistics – Natural language
Reexamination Certificate
2000-10-27
2004-11-09
Abebe, Daniel (Department: 2644)
Data processing: speech signal processing, linguistics, language
Linguistics
Natural language
C704S257000
Reexamination Certificate
active
06816831
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention generally relates to language learning apparatuses and methods therefor. More particularly, the invention relates to a language learning apparatus and a method therefor for enabling a system to automatically learn languages through interaction with users and an external environment by using speech processing techniques, dialog processing techniques, robot control technologies, sensor technologies, etc. which represent natural language processing techniques.
2. Description of the Related Art
Generally, algorithms for learning languages are used for outputting language knowledge (vocabularies and syntax rules) by using a set of language information and non-language information as an input.
The language information is information which can be inferred during communication, while the non-language information is information which cannot be directly inferred during communication.
An overview of various conventional language learning methods according to the types of given language information and non-language information and the types of language knowledge to be learned is shown in FIG.
7
.
According to a first method, by supplying “a preposition and nouns (text) placed before and after the preposition” as the language information and “the meaning of the preposition (symbolic representation indicating, for example, the entry number in the dictionary) as the non-language information, “the meaning of the preposition dependent on the nouns before and after (symbolic representation)” is learned as the language knowledge (see C. Harris “A connectionist approach to the story of ‘over’”, Berkeley Linguistic Society, 15, pp. 126-138, 1989).
According to a second method, by supplying “a preposition, a verb, and a noun (text)” as the language information and “the meaning of a preposition (symbolic representation)” as the non-language information, “the meaning of the preposition dependent on the verb and the noun (symbolic representation)” is learned as the language knowledge (see P. Munro et al. “A network for encoding, decoding and translating locative prepositions”, Cognitive Science, Vol. 3, pp. 225-240, 1991).
According to a third method, by supplying “a sentence (text)” as the language information and “whether the reaction made by the system is true or false” as the non-language information, “the contribution rate of the word for the system's reaction”, i.e., “what kind of reaction the system should make in response to a certain word string?” is learned (see A. L. Gorin et al. “Adaptive acquisition of language”, Computer Speech and Language, Vol.5, pp.101-132, 1991).
According to a fourth method, by supplying “a sentence (text)” as the language information and “the semantic representation of the symbolic sentence” as the non-language information, “the meaning of the word (symbolic representation)” is learned as the language knowledge (see J. M. Siskind, “A computation study of cross-situation techniques for learning word-to-meaning mappings”, Cognition, Vol.61, pp.39-91, 1996).
According to a fifth method, by supplying “a sentence (text) and symbolic vocabulary information” as the language information (non-language information is not given), “the syntax rules” are learned as the language knowledge (see Berwick, “The acquisition of syntactic knowledge”, MIT Press, 1985).
According to a sixth method, by supplying “a word (text)” as the language information and “graphics (computer graphics)” as the non-language information, “the meaning of the word (graphic pattern) is learned as the language knowledge (see S. Nakagawa et al. An acquisition system of concept and grammar based on combining with visual and auditory information, Transactions of IPSJ, Vol.10, No.4, pp.129-137, 1994).
According to a seventh method, by supplying “a word (sound)” as the language information and “graphics (computer graphics)” as the non-language information, “the meaning of the word (graphic pattern)” is learned as the language knowledge (see T. Regier, “The Human Semantic Potential”, MIT Press, 1997).
According to an eighth method, by supplying “a sentence (isolated word speech) as the language information and “whether the reaction made by the system is true or false” as the non-language information, “the contribution rate of the word for the system's reaction” is learned as the language knowledge (see A. L. Gorin et al., “An experiment in spoken language acquisition”, IEEE Transactions on speech and audio processing, Vol.2. No.1, pp.224-240, 1994).
In the above-described conventional learning methods, however, when a sentence (text or sound) is given as the language information, and when perceptual information, such as visual information or sensory information, whose meaning is not explicitly given as the non-language information, only the meaning of the word can be learned.
SUMMARY OF THE INVENTION
Accordingly, in view of the above background, it is an object of the present invention to provide a language learning apparatus and a method therefor in which, even in response to perceptual information whose meaning is not explicitly given, the syntax structure can be determined and the syntax rules of the input language can be learned based on the determined syntax structure.
In order to achieve the above object, according to one aspect of the present invention, there is provided a language learning apparatus including a recognition portion for receiving language information and for extracting a word string according to the input language information. A semantic analyzing portion receives perceptual information related to the language information and extracts concepts and a concept representation indicating the relevance of the concepts. A relevance analyzing portion verifies the word string extracted by the recognition portion against the concept representation extracted by the semantic analyzing portion and determines a syntax structure of the word string according to the relevance between the word string and the concept representation.
According to another aspect of the present invention, there is provided a language learning apparatus including a speech recognition portion for receiving speech information and for extracting a word string according to the input speech information. A semantic analyzing portion receives perceptual information related to the speech information and extracts concepts and a concept representation indicating the relevance of the concepts. A relevance analyzing portion verifies the word string extracted by the speech recognition portion against the concept representation extracted by the semantic analyzing portion and determines a syntax structure of the word string according to the relevance between the word string and the concept representation.
According to still another aspect of the present invention, there is provided a language learning apparatus including a syntax-structure analyzing unit and a syntax-rule learning unit. The syntax-structure analyzing unit has a recognition portion for receiving language information and for extracting a word string according to the input language information. A semantic analyzing portion receives perceptual information related to the language information and extracts concepts and a concept representation indicating the relevance of the concepts. A relevance analyzing portion verifies the word string extracted by the recognition portion against the concept representation extracted by the semantic analyzing portion and determines a syntax structure of the word string according to the relevance between the word string and the concept representation. The syntax-rule learning unit receives the syntax structure determined by the relevance analyzing portion of the syntax-structure analyzing unit and learns a syntax rule.
According to a further aspect of the present invention, there is provided a language learning apparatus including a syntax-structure analyzing unit and a syntax-rule learning unit. The syntax-structure analyzing unit has a speech recognition po
Abebe Daniel
Frommer William S.
Frommer & Lawrence & Haug LLP
Sony Corporation
LandOfFree
Language learning apparatus and method therefor does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Language learning apparatus and method therefor, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Language learning apparatus and method therefor will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3363199