Spelling Correction of Email Queries

• US 20160063094A1
• Filed: 08/28/2014
• Published: 03/03/2016
• Est. Priority Date: 08/28/2014
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Udupa, Raghavendra Uppinakuduru, Bhole, Abhijit Narendra
• Granted Patent: US 10,068,008 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3322 using system suggestions G06F16/3325 takes precedence
• CPC Class Codes: G06F 16/334 Query execution G06F16/335 takes precedence
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06Q 10/107 Computer-aided management of electronic mailing [e-mailing]
• CPC Class Codes: H04L 51/42 Mailbox-related aspects, e.g. synchronisation of mailboxes

Abstract

Techniques and constructs to facilitate spelling correction of email queries can leverage features of email data to obtain candidate corrections particular to the email data being queried. The constructs may enable accurate spelling correction of email queries across languages and domains based on, for example, one or more of a language model such as a bigram language model and/or a normalized token IDF based language model, a translation model such as an edit distance translation model and/or a fuzzy match translation model, content-based features, and/or contextual features. Content-based features can include features associated with the subject line of emails, content including identified phrases, contacts, and/or the number of candidate emails returned. Contextual features can include a time window of subject match and/or contact match, a frequency of emails received from a contact, and/or device characteristics.

Claims

1. A method comprising:

mining at least one email repository;

determining at least two features for candidate correction of an email query as at least two determined features based on at least one of;

content from the email repository;

a language model;

a translation model;

or

context of the email repository;

generating candidate corrections for the email query from the at least two determined features;

ranking the candidate corrections based at least on the determined features; and

identifying at least one correction from the email repository according to ranks associated with the respective candidate corrections.

2. A method as claim 1 recites, wherein at least a part of the email repository is associated with an entity from which the email query is received.

3. A method as claim 1 recites, wherein ranking the candidate corrections is performed according to at least one statistical ranking algorithm.

4. A method as claim 1 recites, wherein the content from the email repository identifies content of email fields including at least one of a contact field, a subject field, or a body field.

5. A method as claim 1 recites, wherein context of respective emails in the email repository includes at least one of:

a time range of recent emails having a subject or contact field associated with determined features, or a frequency of emails received from, or sent to, a contact.

6. A method as claim 1 recites, wherein at least one of the candidate corrections being generated for the email query from the determined features represents a misspelling of a known word.

7. A method as claim 1 recites, further comprising:

matching at least one of the candidate corrections to data from the email repository; and

omitting a candidate correction of the candidate corrections from the ranking in an event the candidate correction is not matched to data from the email repository.

8. A method as claim 1 recites, further comprising identifying at least one email from the email repository containing at least one match to at least one of the candidate corrections without matching the email query.

9. A method as claim 1 recites, further comprising identifying at least one email from the email repository containing at least one match to at least one of the candidate corrections matching the email query.

10. A method as claim 1 recites, wherein the email repository includes multiple subdivisions associated with an entity from which a same email query is received;

and further comprising;

identifying at least a first email from the email repository associated with a first of the multiple subdivisions, the first email containing at least a first match to at least one of the candidate corrections without matching the email query;

identifying at least a second email from the email repository associated with a second of the multiple subdivisions, the second email containing at least a second match to at least one of the candidate corrections without matching the email query; and

the first match differing from the second match.

11. A method as claim 1 recites, wherein the email repository includes multiple subdivisions associated with an entity from which a same email query is received;

and further comprising;

identifying at least a first email from the email repository associated with a first of the multiple subdivisions, the first email containing at least a first match to at least one of the candidate corrections matching the email query;

identifying at least a second email from the email repository associated with a second of the multiple subdivisions, the second email containing at least a second match to at least one of the candidate corrections matching the email query; and

the first match differing from the second match.

12. A method as claim 1 recites, wherein the email repository is associated with at least two entities from which a same email query is received;

and further comprising;

identifying at least a first email from the email repository associated with a first of the at least two entities, the first email containing at least a first match to at least one of the candidate corrections without matching the email query;

identifying at least a second email from the email repository associated with a second of the at least two entities, the second email containing at least a second match to at least one of the candidate corrections without matching the email query; and

the first match differing from the second match.

13. A method as claim 1 recites, wherein the email repository is associated with at least two entities from which a same email query is received;

and further comprising;

identifying at least a first email from the email repository associated with a first of the at least two entities, the first email containing at least a first match to at least one of the candidate corrections matching the email query;

identifying at least a second email from the email repository associated with a second of the at least two entities, the second email containing at least a second match to at least one of the candidate corrections matching the email query; and

the first match differing from the second match.

14. A device comprising:

one or more computer-readable media having thereon a plurality of modules and an email repository;

a processing unit operably coupled to the computer-readable media, the processing unit adapted to execute modules of the plurality of modules comprising;

an feature function module configured to identify features from email data in the email repository, the feature function module including a content-based features module configured to score candidate corrections according to where respective of the candidate corrections occur in content of the email data;

a candidate generation module configured to generate the candidate corrections corresponding to an email query, the candidate corrections being based at least in part on the email data;

a ranking module configured to rank the candidate corrections based at least in part on the features from the email data upon which the candidate corrections are based.

15. A device as claim 14 recites, wherein the content-based features module is configured to treat output of at least one of the language model module or the translation model module as a feature.

16. A device as claim 14 recites, wherein the content-based features module is configured to score content-based features according to at least one of:

subject cover, a portion of mails containing at least one candidate token of the candidate corrections in the subject line;

subject match, a maximum of a portion of candidate tokens of the candidate corrections appearing in any single subject line in the email data;

key phrase cover, a set covering a maximum number of candidate tokens of the candidate corrections contained in phrases in the email data in the email repository;

contact match, a portion of candidate tokens that are present in at least one contact of at least one email account in the email data in the email repository;

or

too many emails, a Boolean function to evaluate whether the candidate returns a number of emails above a user definable or pre-defined threshold.

17. A device as claim 14 recites, wherein the feature function module further includes at least one module comprising:

a language model module configured to score candidate corrections;

a translation model module configured to score lexical similarity of candidate corrections;

or

a contextual features module configured to score candidate corrections according to whether respective of the candidate corrections occur in a defined time window.

18. A device as claim 17 recites, wherein the translation model module is configured to employ at least one translation model of edit distance or fuzzy match.

19. A device as claim 17 recites, wherein the contextual features module is configured to score contextual features according to at least one of:

recent subject match, a portion of candidate tokens appearing in a subject line in the email data in the email repository for a time window of recent emails;

recent contact match, a portion of candidate tokens appearing in a contact in the email data in the email repository for a time window of recent contact interactions;

frequent contact match, a portion of candidate tokens that are present more than a certain number of times in the email data in the email repository for a time window of recent contact interactions and/or more than a certain number of times in the email data in the email repository for a time window of the email data;

or

device match, a portion of candidate tokens present with an indication of input from a particular input type.

20. A computer-readable medium having thereon computer-executable instructions, the computer-executable instructions upon execution configuring a computer to perform operations comprising:

identifying a plurality of tokens in an email query;

identifying features associated with respective of the tokens, the features associated with at least one of email contacts or email content;

creating respective vectors of the features associated with respective of the tokens;

identifying candidate corrections for the respective tokens; and

combining the respective vectors to rank the candidate corrections based at least on contextual aspects of the features.