Method and system for using natural language techniques to process inputs

• US 8,818,795 B1
• Filed: 07/24/2013
• Issued: 08/26/2014
• Est. Priority Date: 03/14/2013
• Status: Active Grant

• Current Assignee: Yahoo Assets LLC
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Cassimatis, Nicholas Louis, Caneba, Richard Noel, Scally, Jonathan Richard
• Primary Examiner(s): ALBERTALLI, BRIAN LOUIS
• Application Number:
  US13/949,886
• Publication Number:
  US 20140278369A1
• Time in Patent Office: 398 Days
• Field of Search: None
• US Class Current: 704/9
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

Systems and methods are provided for utilizing natural language to process queries. The method for analyzing a linguistic input may include receiving the linguistic input, the linguistic input including at least one word, accessing prestored language data for a language corresponding to the linguistic input, converting the linguistic input into a text-possibility representations based on the received language data, determining a meaning of the text possibility based on the prestored language data, generating at least one semantic structure corresponding to the determined meaning, and determining an action to perform based on the generated at least one semantic structure. The prestored language data may be converted from multiple formats into one or more formats that can be algorithmically processed by a computational device.

Claims

1. A method for analyzing a linguistic input comprising, at a computing device:

receiving the linguistic input, the linguistic input comprising at least one word;

accessing prestored language data for a language corresponding to the linguistic input;

converting the linguistic input into a text possibility based on the received language data;

determining a meaning of the text possibility based on the prestored language data, including;

selecting a subset of the text possibility, and

generating at least one new language construction and/or lexeme based on the subset of the text possibility;

generating at least one semantic structure corresponding to the determined meaning; and

determining an action to perform based on the generated at least one semantic structure.

2. The method of claim 1, wherein the prestored language data includes language constructions and lexemes.

3. The method of claim 2, wherein the language constructions and lexemes associate text patterns with corresponding meanings.

4. The method of claim 1, wherein converting the linguistic input into the text possibility includes determining whether the linguistic input is ambiguous based on the prestored language data.

5. The method of claim 1, further comprising:

receiving a user input; and

converting the user input into the linguistic input,

wherein the linguistic input is converted into the text possibility based on at least one of handwriting recognition, speech recognition, optical character recognition, and a gesture based user interface.

15. The method of claim 1, wherein the linguistic input comprises an input convertible to text including one of a textual input, handwritten, gestural, and a spoken input.

16. The method of claim 1, wherein the prestored language data includes at least one of word semantics, phrase semantics, linguistic statistics, world knowledge, and linguistic taxonomy.

17. The method of claim 1, wherein the determined action is at least one of performing an internet search, composing a message, creating a contact, interacting with software, manipulating, retrieving or storing data producing computer code, interacting with software through APIs, querying or modifying one or more databases, and placing a telephone call.

18. The method of claim 1, wherein the linguistic input is received from a device and the determined action is performed at one of the device and a separate device.

19. At least one non-transitory processor readable storage medium storing a computer program of instructions configured to be readable by at least one processor for instructing the at least one processor to execute a computer process for performing the method as recited in claim 1.

6. A method for analyzing a linguistic input comprising, at a computing device:

receiving the linguistic input, the linguistic input comprising at least one word;

accessing prestored language data for a language corresponding to the linguistic input;

converting the linguistic input into a text possibility based on the received language data;

determining a meaning of the text possibility based on the prestored language data, including;

selecting a subset of the text possibility;

generating at least one new language construction and at least one new lexeme based on the subset of the text possibility;

identifying a known language construction that matches the at least one new language construction;

combining the known language construction that matches the subset of the text possibility with another language construction to generate a meaning possibility;

generating a semantic structure for the meaning possibility of the text possibility;

modifying the semantic structure for the meaning possibility based on the prestored language data; and

determining the action to perform based on the modified semantic structure for the meaning possibility.

7. The method of claim 6, wherein generating at least one new language construction is based on common linguistic properties included in the prestored language data.

8. The method of claim 7, wherein identifying the known language construction that matches the at least one new language construction is based on the common linguistic properties.

9. The method of claim 6, wherein generating the meaning possibility comprises identifying a grammatical form of the subset.

10. The method of claim 9, wherein generating the meaning possibility comprises determining whether the subset is ambiguous and removing the ambiguity based on at least one of known language semantics, language statistics, and language taxonomy.

11. The method of claim 10, wherein the ambiguity is related to at least one of the grammatical form of the subset, a phrase attachment to the subset, a semantic type of part of the subset, and a pronoun binding in the subset.

12. The method of claim 9, wherein the grammatical form comprises at least one of a word type and a phrase type, wherein each of the word type or the phrase type is constrained in at least one of a grammatical category, semantic category, morphological structure, and phonological structure of input.

13. The method of claim 6, wherein generating the meaning possibility is based on at least one of the combined language constructions, a pronoun binding in the subset, a conjunction in the subset, and an ellipsis in the subset.

14. The method of claim 6, wherein the subset comprises at least one word.

20. A method for analyzing a linguistic input comprising, at a computing device:

accessing prestored data, the prestored data containing at least one of language constructions, lexemes, world knowledge, statistical regularities, and contextual information;

generating a new language construction and lexeme based on the prestored data;

generating an additional language construction and lexeme based on a combination of prior language constructions and lexemes;

converting the prestored data into a different format;

storing the converted data in the different format;

receiving the linguistic input, the linguistic input comprising at least one word;

converting the linguistic input into a text possibility based on the converted data comprising the new and additional language constructions and lexemes;

determining a meaning of the text possibility based on the converted data;

generating at least one semantic structure corresponding to the determined meaning; and

determining an action to perform based on the generated at least one semantic structure.

21. The method of claim 20, wherein each of the language constructions and lexemes include at least one word type paired with a semantic structure, the word type comprising at least one of syntactic and semantic information.

22. The method of claim 20, wherein the prestored data includes at least one of categories of linguistic and nonlinguistic data, word associations, world knowledge, linguistic knowledge, statistical regularities, and an analog signal to text-possibility representation conversion module.

23. The method of claim 20, wherein generating the new language construction and lexeme comprises:

selecting one of the plurality of prestored language constructions and/or lexemes;

determining a syntactic form of the selected construction;

generating a new syntactic form based on the form of the selected construction; and

generating a new semantic structure based on the selected construction and the generated syntactic form.

24. The method of claim 23, wherein the syntactic form of the selected constructions or lexemes comprises at least one of a word type and a phrase type that is constrained in at least one of a grammatical category, semantic category, morphological structure, and phonological structure of input.

25. The method of claim 23, wherein generating the new language construction further comprises:

generating a semantic representation for the generated new syntactic form; and

combining the generated new syntactic form and the semantic representation to form the new language construction.

26. The method of claim 20, wherein generating the additional language construction is based on a determination of whether one or more prior language constructions or lexemes together match one of the plurality of prestored language constructions and lexemes.

27. The method of claim 20, wherein determining a possible syntactic structure and semantic structure of the linguistic input is based on a computational process utilizing an interrelationship between the prestored language constructions, lexemes, world knowledge, statistical regularities, and contextual information.

28. A system for analyzing a linguistic input comprising:

one or more processors communicatively coupled to a network;

wherein the one or more processors are configured to;

receive the linguistic input, the linguistic input comprising at least one word;

access prestored language data for a language corresponding to the linguistic input;

convert the linguistic input into a text possibility based on the received language data;

determine a meaning of the text possibility based on the prestored language data;

generate at least one semantic structure corresponding to the determined meaning;

generate a new language construction and/or lexeme based on the text possibilty; and

determine an action to perform based on the generated at least one semantic structure.

29. A system for analyzing a linguistic input comprising:

one or more processors communicatively coupled to a network;

wherein the one or more processors are configured to;

access prestored data, the prestored data containing at least one of language constructions, lexemes, world knowledge, statistical regularities, and contextual information;

generate a new language construction and lexeme based on the prestored data;

generate an additional language construction and lexeme based on a combination of prior language constructions and lexemes;

convert the prestored data into a different format;

store the converted data in the different format;

receive the linguistic input, the linguistic input comprising at least one word;

convert the linguistic input into a text possibility based on the converted data comprising the new and additional language constructions and lexemes;

determine a meaning of the text possibility based on the converted data;

generate at least one semantic structure corresponding to the determined meaning; and

determine an action to perform based on the generated at least one semantic structure.