Content search in complex language, such as Japanese

• US 20060031207A1
• Filed: 06/08/2005
• Published: 02/09/2006
• Est. Priority Date: 06/12/2004
• Status: Active Grant

• Current Assignee: Getty Images Incorporated
• Original Assignee: Getty Images Incorporated
• Inventors: Bjarnestam, Anna, Carothers, David, MacGuffie, Monika
• Granted Patent: US 7,523,102 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/36 Creation of semantic tools, e.g. ontology or thesauri
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/53 Processing of non-Latin text kana-to-kanji conversion G06F40/129; vowelisation G06F40/232
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99936 Pattern matching access
• CPC Class Codes: Y10S 707/99937 Sorting

Abstract

A search facility provides searching capabilities in languages such as Japanese. The facility may use a vocabulary knowledge base organized by concepts. For example, each concept may be associated with at least one keyword (as well as any synonyms or variant forms) by applying one or more rules that relate to identifying common main forms, script variants, alternative grammatical forms, phonetic variants, proper noun variants, numerical variants, scientific name, cultural relevance, etc. The contents of the vocabulary knowledge base are then used in executing search queries. A user may enter a search query in which keywords (or synonyms associated with those key words) may be identified, along with various stopwords that facilitate segmentation of the search query and other actions. Execution of the search query may result in a list of assets or similar indications being returned, which relate to concepts identified within the search query.

Claims

1. A system for searching content requested using a text-based query including one or more scripts or orthographic forms associated with Japanese, the system comprising:

a hierarchically structured vocabulary knowledge base for storing vocabulary information associated with the one or more scripts or orthographic forms, wherein the vocabulary knowledge base is generated by a method comprising;

assigning an identifier to a semantic concept;

identifying a main orthographic form for the semantic concept, wherein the main orthographic form is based on kanji script, katakana script, hiragana script, or any combination of kanji script, katakana script, and hiragana script;

for at least one of the one or more scripts or orthographic forms, associating at least one synonymous orthographic form with the semantic concept, wherein the synonymous orthographic form is at least partially distinct from the main orthographic form, and wherein the at least one synonymous orthographic form includes any one or more of;

kanji script,

katakana script,

hiragana script,

okurigana variant,

romaji written form,

phonetic variants associated with one or more of kanji script, katakana script, hiragana script, okurigana variant, and romaji written form, and/or

hybrid variants associated with one or more of kanji script, katakana script, hiragana script, okurigana variant, and romaji written form;

storing the identifier, the main orthographic form, and the at least one synonymous orthographic form in a data storage component associated with the system; and

repeating the assigning, the identifying, the associating, and the storing for additional semantic concepts;

an asset repository for storing information associated with searchable assets;

a classification component for classifying the searchable assets in the asset repository to facilitate matching between the searchable assets and the vocabulary information, wherein the matching is based, at least in part, on identifiers assigned to semantic concepts included in the vocabulary knowledge base; and

a search engine for receiving and executing queries for the searchable assets.

2. The system of claim 1 wherein the search engine performs a method comprising:

receiving a search query comprising a textual expression represented using one or more characters in at least one of the one or more scripts or orthographic forms;

normalizing the received search query, wherein the normalizing includes converting any characters with multiple possible representations to a standardized form;

tokenizing the normalized search query, wherein the tokenizing includes separating the textual expression into one or more tokens;

based on a vocabulary knowledge base, determining a match for each token of the tokenized search query;

for each token where the determining of a match is not successful, segmenting the token, wherein the segmenting includes performing additional matching on identified segments of the token;

executing the search using a set of tokens for which a match has been determined; and

displaying a result set from the executed search.

3. The system of claim 1 wherein the method for generating the vocabulary knowledge base further comprises any one or more of the following:

associating one main common form with the semantic concept;

associating one or more script-based forms with the semantic concept;

associating one or more grammatically-based forms with the semantic concept;

associating one or more phonetic forms with the semantic concept;

associating one or more synonymous forms for any proper nouns associated with the semantic concept;

associating one or more synonymous forms for any Arabic numerals and katakana with the semantic concept;

associating a scientific name written using Latin letters with the semantic concept;

associating one or more synonymous forms that are culturally appropriate with the semantic concept;

screening synonymous forms that are not culturally appropriate; and

screening synonymous forms that are not culturally relevant.

4. In a computer system, a method for searching for content identified using text and symbols of a language having multiple written variants of at least some semantic concepts, the method comprising:

receiving a search query comprising a textual expression represented using one or more characters in at least one script or orthographic form associated with the language;

normalizing the received search query, wherein the normalizing includes converting any characters with multiple possible representations to a standardized form;

tokenizing the normalized search query, wherein the tokenizing includes separating the textual expression into one or more tokens;

based on contents of a vocabulary knowledge base, determining a match for each token of the tokenized search query;

for each token where the determining of a match is not successful, segmenting the token, wherein the segmenting includes performing additional matching on identified segments of the token;

executing the search using a set of tokens for which a match has been determined; and

displaying a result set from the executed search.

5. The method of claim 4 wherein the language is Japanese.

6. The method of claim 4 wherein the language is a language other than Japanese.

7. The method of claim 4 wherein the segmentation includes one or more of:

identifying the token as including a kanji group with trailing hiragana and checking for matches in the vocabulary knowledge base;

identifying the token as including only kanji, only hiragana, or only katakana and checking for matches in the vocabulary knowledge base;

identifying the token as including only non-kanji, only non-hiragana, or only non-katakana and checking for matches in the vocabulary knowledge base;

identifying the token as including a hiragana character group that contains one or more stopwords; and

segmenting any kanji-only character groups that are not successfully matched.

8. A system for searching for content identified using text and symbols of a complex language associated with multiple written forms, the system comprising:

a vocabulary knowledge base for storing vocabulary information associated with the complex language, wherein the vocabulary knowledge base is generated or updated by a repeatable method comprising;

assigning an identifier to a semantic concept;

identifying a main written form for the semantic concept, wherein the main written form is based on at least one of the multiple written forms;

for at least one of the multiple written forms associated with the complex language, associating at least one synonymous written form with the semantic concept, wherein the synonymous written form is at least partially distinct from the main written form; and

storing the identifier, the main written form, and the at least one synonymous written form in a data storage component associated with the system;

an asset repository for storing information associated with searchable assets; and

a search engine for receiving and executing queries for the searchable assets, wherein the execution is based, at least in part, on the contents of the vocabulary knowledge base.

9. The system of claim 8, further comprising, a classification component for classifying the searchable assets in the asset repository to facilitate matching between the searchable assets and the vocabulary information.

10. The system of claim 8 wherein the asset repository stores data associated with images.

11. The system of claim 8 wherein the asset repository stores data associated with documents.

12. The system of claim 8 wherein the vocabulary knowledge base is a subcomponent of a meta-data repository, and wherein the meta-data repository further comprises a concept-to-asset repository, for storing information associated with relationships between concepts from the vocabulary database and assets from the asset repository.

13. The system of claim 8 wherein the vocabulary knowledge base is structured hierarchically, such that concepts located farther down in the hierarchy provide more specificity than concepts located higher up in the hierarchy.

14. The system of claim 8 wherein the identified main written form is tagged as a keyword for searching.

15. The system of claim 8 wherein the language is Japanese, and wherein the multiple written forms include, kanji script, katakana script and hiragana script, okurigana variant, and romaji written form.

16. A method for generating or updating a vocabulary repository used in executing queries for retrieving content, wherein the queries are provided in a complex language associated with multiple written forms, the method comprising:

identifying a main form for a semantic concept, wherein the main form is based on at least one of the multiple written forms;

assigning an identifier to a semantic concept, wherein the assigned identifier is associated with at least the main form;

associating at least one synonymous variant with the semantic concept, wherein the at least one synonymous written variant is at least partially distinct from the main form; and

storing the identifier, the main form, and the at least one synonymous variant in a data storage component accessible to a search engine configured to search for assets when provided with search queries in the complex language.

17. The method of claim 16, further comprising repeating the assigning, the identifying, the associating, and the storing for additional semantic concepts.

18. The method of claim 16, wherein identifying a main form for a semantic concept and associating at least one synonymous variant with the semantic concept includes:

associating one or more script based forms with the semantic concept;

associating one or more grammatically based forms with the semantic concept;

associating one or more phonetic forms with the semantic concept;

associating one or more synonymous forms for any proper nouns associated with the semantic concept;

associating one or more synonymous forms for any Arabic numerals and katakana with the semantic concept;

associating a scientific name with the semantic concept using Latin characters;

associating one or more synonymous forms that are culturally appropriate;

screening synonymous forms that are not culturally appropriate; and

screening synonymous forms that are not culturally relevant.

19. A computer-readable medium containing a data structure associated with a semantic concept that is representable in a complex language, the data structure comprising:

identifier information associated with the semantic concept;

information identifying a main textual form for the semantic concept, wherein the main textual form is based on a first script type associated with the complex language; and

information associating at least one synonymous textual form with the semantic concept, wherein the synonymous textual form is based on a second script type associated with the complex language,

wherein the second script type is at least partially distinct from the first script type, and

wherein the data structure is configured for facilitating matching assets with search queries based on the complex language.

20. The computer-readable medium of claim 19 wherein the computer-readable medium is a logical node in a computer network receiving the contents.

21. The computer-readable medium of claim 19 wherein the computer-readable medium is a computer-readable disk.

22. The computer-readable medium of claim 19 wherein the computer-readable medium is a memory of a computer system.

23. A computer-implemented method for executing a search query, the method comprising:

receiving a search query including a textual expression, wherein the textual expression is written in a language that at least occasionally lacks discrete boundaries between words or autonomous language units;

determining whether the textual expression comprises a keyword or synonym associated with a structured vocabulary knowledge base for storing vocabulary information associated a language having multiple orthographic forms or scripts; and

if the textual expression does not comprise a keyword or synonym associated with a structured vocabulary knowledge base, performing segmentation on the textual expression, wherein the segmentation includes systematically splitting the textual expression into two or more segments, and identifying at least one keyword from the vocabulary knowledge based on the textual expression and the two or more segments.

24. The method of claim 23 wherein the systematic splitting includes identifying predetermined stopwords that are not keywords, wherein at least some of the predetermined stopwords are prepositions.

25. The method of claim 23 wherein the systematic splitting includes splitting the textual expression into the longest segments possible in an effort to preserve the intended meaning of the textual expression.

26. The method of claim 23 wherein the systematic splitting includes an application of both linguistic rules and contrived rules.

27. The method of claim 23 wherein the systematic splitting includes identifying predetermined stopwords that are not keywords, wherein at least some of the predetermined stopwords represent Boolean values used in executing the search query.

28. The method of claim 23 wherein the language is Japanese, and wherein determining whether the textual expression in its entirety comprises a keyword or synonym associated with the structured vocabulary knowledge base includes checking for matches in kanji groups with trailing hiragana.

29. The method of claim 23 wherein the language is Japanese, and wherein determining whether the textual expression in its entirety comprises a keyword or synonym associated with the structured vocabulary knowledge base includes checking for matches in character groups of only kanji, only hiragana, only katakana, and only non-kanji non-hiragana non-katakana characters.

30. The method of claim 23 wherein the systematic splitting includes segmenting a textual expression WXYZ as:

WXY, XYZ, WX, XY, YZ, W, X, Y, Z, wherein W, X, Y, and Z each represent a character or autonomous grouping of characters associated with the language.

31. The method of claim 23 wherein the search query is received via a network connection.

32. A structured vocabulary stored in a computer memory usable with a computer system, comprising:

multiple digital media assets; and

multiple keywords, wherein the keywords are ordered in the structured vocabulary according to inherent relationships between the terms, such that at least some of the digital media assets are associated with a keyword and at least one variant form of the keyword, and wherein each digital media asset is associated with an identifier and can be identified by matching a requested term with a keyword or at least one variant form of the keyword associated with one of the identifiers.

33. The structured vocabulary of claim 32 wherein the structured vocabulary is portable and not limited for use within any specific computer application.

34. A method in a first computer system for retrieving a media content unit from a second computer system having a plurality of media content units that have been classified according to keyword terms of a structured vocabulary, comprising:

sending a request for a media content unit, the request specifying a search term; and

receiving an indication of at least one media content unit that corresponds to the specified search term, wherein the search term is located within the structured vocabulary and is used to determine at least one media content unit that corresponds to the search term, and wherein orthographic variations of the search term are automatically provided to assist in determining the at least one media content unit that corresponds to the search term.

35. The method of claim 34 wherein the keyword terms of the structured vocabulary are ordered such that the relationship of each term to each other term is inherent in the ordering.

36. A computer-implemented method to search for data in a database, the method comprising:

receiving a search request, wherein the request includes a search query term;

automatically determine any orthographic variations of the search query term; and

conducting a search of the database based on the search query term and any orthographic variations of the search query term.

37. The method of claim 36 wherein the search request comprises at least one orthographic form of Japanese text.

38. The method of claim 36 wherein the search request comprises at least one orthographic form of Korean text.

39. The method of claim 36 wherein the search request comprises at least one orthographic form of Chinese text.

40. The method of claim 36 wherein conducting the search occurs in the context of searching for images cataloged in an image repository.

41. The method of claim 36 wherein conducting the search occurs in the context of searching for products in an e-commerce or Internet retail setting.

42. The method of claim 36 wherein conducting the search occurs in the context of searching for words or items within a document.

43. The method of claim 36 wherein conducting the search occurs in the context of a free-text search engine.

44. The method of claim 36 wherein conducting the search occurs in the context of retrieving documents from a document archive.