USER INPUT CLASSIFICATION

• US 20090070097A1
• Filed: 11/17/2008
• Published: 03/12/2009
• Est. Priority Date: 03/16/2004
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Zhu, Hongjun, Wu, Jun, Zhu, Huican
• Granted Patent: US 8,660,834 B2
• Time in Patent Office: Days
• US Class Current: 704/2
• CPC Class Codes: G06F 40/129 Handling non-Latin characters, e.g. kana-to-kanji conversion

Abstract

Systems and methods of classifying user input are disclosed. The user input can be, for example, in the form of Roman characters. An ambiguous word (e.g., a word that is a non-pinyin word written in Roman characters and a valid pinyin word) can be identified in the user input. Contextual words (e.g., words adjacent to the ambiguous word) are classified as a pinyin context or a non-pinyin context. The ambiguous word is classified based on the context of the contextual words.

Claims

1. A method of classifying a user input, comprising:

receiving a user input of Roman characters;

determining that the user input includes an ambiguous word, the ambiguous word being a non-pinyin word written in the Roman characters and also being a valid pinyin input of one or more pinyin syllables;

in response to determining that the user input includes an ambiguous word;

classifying one or more words adjacent to the ambiguous word as belonging to a pinyin context or belonging to a non-pinyin context;

classifying the ambiguous word as a pinyin input if the one or more words adjacent to the ambiguous word are classified as belonging to the pinyin context; and

classifying the ambiguous word as a non-pinyin word if the one or more words adjacent to the ambiguous word are classified as belonging to the non-pinyin context.

2. The method of claim 1, wherein the one or more words adjacent to the ambiguous word are classified as belonging to the pinyin context when the one or more words adjacent to the ambiguous word are valid pinyin words and are not valid non-pinyin words.

3. The method of claim 2, wherein the one or more words adjacent to the ambiguous word are classified as belonging to the non-pinyin context when the one or more words adjacent to the ambiguous word are valid non-pinyin words and are not valid pinyin words.

4. The method of claim 3, wherein a valid pinyin word comprises Roman characters that constitute one or more pinyin syllables that map to a Chinese word.

5. The method of claim 3, wherein a valid non-pinyin word comprises Roman characters that constitute a word in the English language.

6. The method of claim 1, further comprising:

determining a probability that the ambiguous word corresponds to a Chinese query; and

classifying the ambiguous word as a pinyin input when the probability satisfies a threshold probability.

7. The method of claim 6, wherein determining a probability that the ambiguous word corresponds to a Chinese query comprises generating possible Chinese word strings for the ambiguous word, each possible Chinese word string having a weight that indicates a probability that the possible Chinese word string corresponds to a pinyin word sting comprising the ambiguous word.

8. The method of claim 7, wherein generating possible Chinese word strings for the ambiguous word comprises:

segmenting the ambiguous word into pinyin candidates, each pinyin candidate comprising a portion of the ambiguous word; and

identifying possible Chinese word strings for each pinyin candidate, each possible Chinese word string having a weight that indicates a probability that the possible Chinese word string corresponds to the pinyin candidate.

9. The method of claim 8, further comprising providing possible Chinese word strings having weights that exceed a predefined threshold to a user interface.

10. The method of claim 1, further comprising:

identifying a subset of non-pinyin words based on a frequency of occurrence of the non-pinyin words in the user inputs;

determining whether each non-pinyin word in the identified subset of non-pinyin words is a valid pinyin input; and

defining each non-pinyin word in the subset that is also a valid pinyin input as an ambiguous word.

11. A system for classifying a user input, comprising:

a user interface to receive a user input of Roman characters and to provide possible Chinese queries that correspond to the user input; and

a pinyin classifier configured to receive the user input from the user interface and determine whether the user input includes an ambiguous word, the ambiguous word being a non-pinyin word written in the Roman characters and also being a valid pinyin input of one or more pinyin syllables, the pinyin classifier further configured to classify one or more words adjacent to the ambiguous word as belonging to a pinyin context or belonging to a non-pinyin context and classify the ambiguous word based on the classification of the one or more adjacent words.

12. The system of claim 11, wherein the pinyin classifier is further configured to classify the ambiguous word as a pinyin input if the one or more words adjacent to the ambiguous word are classified as belonging to the pinyin context and classify the ambiguous word as a non-pinyin word if the one or more words adjacent to the ambiguous word are classified as belonging to the non-pinyin context.

13. The system of claim 12, wherein the pinyin classifier is further configured to classify the one or more words adjacent to the ambiguous word as belonging to the pinyin context when the one or more words adjacent to the ambiguous word are valid pinyin words and not valid non-pinyin words.

14. The system of claim 13, wherein the pinyin classifier is further configured to classify the one or more words adjacent to the ambiguous word as belonging to the non-pinyin context when the one or more words adjacent to the ambiguous word are valid non-pinyin words and not valid pinyin words.

15. The system of claim 14, wherein a valid pinyin word comprises Roman characters that constitute one or more pinyin syllables that map to a Chinese word.

16. The system of claim 14, wherein a valid non-pinyin word comprises Roman characters that constitute a valid word in the English language.

17. The system of claim 11, wherein the pinyin classifier is further configured to determine probabilities that possible Chinese word strings correspond to the ambiguous word and classify the user input based on the probabilities.

18. A computer-implemented method, comprising:

receiving a user input of Roman characters;

identifying pinyin candidates based on the user input, each pinyin candidate comprising a portion of the user input;

determining that the user input includes an ambiguous word, the ambiguous word being a non-pinyin word written in the Roman characters and also being a valid pinyin input of one or more pinyin syllables;

classifying the ambiguous word as a pinyin candidate or a non-pinyin word based on a classification of one or more contextual words that are adjacent to the ambiguous word;

providing the pinyin candidates to a pinyin decoder, the pinyin decoder including an updated Chinese dictionary that includes new valid words identified based on an information gain value;

receiving possible Chinese word strings that correspond to the pinyin candidates from the pinyin decoder, each possible Chinese word string having a weight that indicates a probability that the possible Chinese word string corresponds to the user input; and

providing a predefined number of possible Chinese word strings to a user device based on the weights.

19. The method of claim 18, wherein classifying the ambiguous word comprises:

classifying the one or more contextual words adjacent to the ambiguous word as belonging to a pinyin context or belonging to a non-pinyin context;

classifying the ambiguous word as a pinyin input if the one or more contextual words adjacent to the ambiguous word are classified as belonging to the pinyin context; and

classifying the ambiguous word as a non-pinyin word if the one or more contextual words adjacent to the ambiguous word are classified as belonging to the non-pinyin context.

20. The method of claim 19, wherein the one or more contextual words adjacent to the ambiguous word are classified as belonging to the pinyin context when the one or more contextual words adjacent to the ambiguous word are valid pinyin words and are not valid non-pinyin words.

21. The method of claim 20, wherein the one or more contextual words adjacent to the ambiguous word are classified as belonging to the non-pinyin context when the one or more contextual words adjacent to the ambiguous word are valid non-pinyin words and are not valid pinyin words.

22. The method of claim 21, wherein a valid pinyin word comprises Roman characters that constitute one or more pinyin syllables that map to a Chinese word.

23. The method of claim 21, wherein a valid non-pinyin word comprises Roman characters that constitute a word in the English language.