Optimizing a language model based on a topic of correspondence messages

Patent metadata

Field	Value
Publication number	US-2016170971-A1
Application number	US-201414570934-A
Country	US
Kind code	A1
Filing date	Dec 15, 2014
Priority date	Dec 15, 2014
Publication date	Jun 16, 2016
Grant date	—

Abstract

Technology for optimizing a language model based on a topic identified in correspondence messages. The system may continuously or periodically optimize a language model based on topics identified in past correspondence messages or topics anticipated based on an intended recipient of a correspondence message being drafted. The system can operate in combination or conjunction with a language prediction system, such as a next word prediction application used by a virtual keyboard, thus providing improved language prediction for conversations related to identified topics.

First claim

We claim:

1 . A tangible computer-readable storage medium containing instructions for performing a method of optimizing a language model based on a topic identified in correspondence messages, the method comprising:
maintaining correspondence messages,
wherein the correspondence messages have been transferred from a first party to at least one other party;
receiving an indication to optimize a language model,
wherein the language model is to be optimized based at least in part on a topic identified in correspondence messages;
selecting a language model to be optimized; identifying correspondence messages associated with at least one of the first party or the at least one other party; determining a topic in the correspondence messages associated with at least one of the first party or the at least one other party; identifying a word and/or phrase associated with the determined topic; optimizing the language model,
wherein optimizing the language model includes adjusting a priority in the language model associated with the identified word and/or phrase; and
outputting the language model.

2 . The tangible computer-readable storage medium of claim 1 , wherein the first party is a user of a device operating a language prediction application, and wherein the identified correspondence messages were sent or received by the user.

3 . The tangible computer-readable storage medium of claim 1 , wherein determining a topic in the correspondence messages includes identifying keywords associated with a topic in correspondence messages.

4 . The tangible computer-readable storage medium of claim 1 , wherein determining a topic in the correspondence messages includes comparing, to a threshold value, a frequency that a word or phrase associated with a topic is used.

5 . The tangible computer-readable storage medium of claim 1 ,
wherein the indication to optimize a language model includes information related to a message being drafted by a user, and wherein the information related to the message being drafted by the user includes an intended recipient of the message being drafted.

6 . The tangible computer-readable storage medium of claim 1 , wherein the indication to optimize a language model includes information related to an intended recipient of the message, and wherein the method further comprises:
determining a second topic based at least in part on the intended recipient of the message; and identifying a word and/or phrase associated with the determined second topic,
wherein optimizing the language model further includes adjusting a priority in the language model associated the identified word and/or phrase associated with the determined second topic.

7 . The tangible computer-readable storage medium of claim 1 , wherein the method further comprises:
determining that the topic is no longer active; and adjusting the priority in the language model associated with the identified word and/or phrase to a previous priority level.

8 . The tangible computer-readable storage medium of claim 1 , wherein the indication to optimize a language model is generated by a language prediction application operating on a device.

9 . A system for optimizing a language model based on a topic identified in correspondence messages, the system comprising:
a memory containing computer-executable instructions of:
a message filtering module configured to:
maintain correspondence messages,
wherein the correspondence messages have been transferred from a first party to at least one other party;

identify correspondence messages associated with at least one of the first party or the at least one other party;

a message analysis module configured to:
determine, in the correspondence messages, a topic associated with at least one of the first party or the at least one other party;
identify a word and/or phrase associated with the determined topic;

a language model identification module configured to select a language model to be optimized; and
a language model optimization module configured to:
receive an indication to optimize a language model,
wherein the language model is to be optimized based at least in part on a topic identified in the identified correspondence messages;

optimize the language model,
wherein the language model is optimized by adjusting a priority in the language model associated with the identified word and/or phrase associated with the determined topic; and
output the language model; and

a processor for executing the computer-executable instructions stored in the memory.

10 . The system of claim 9 , wherein the first party is a user of a device operating a language prediction application, and wherein the identified correspondence messages were sent or received by the user.

11 . The system of claim 9 , wherein the message analysis module is further configured to determine a topic in the correspondence messages based at least in part on identifying keywords associated with the topic in correspondence messages.

12 . The system of claim 9 , wherein the message analysis module is further configured to determine a topic in the correspondence messages based at least in part on a comparison, to a threshold value, of a frequency that a word or phrase associated with the topic is used.

13 . The system of claim 9 ,
wherein the indication to optimize a language model includes information related to a message being drafted by a user, and wherein the information related to the message being drafted by the user includes an intended recipient of the message being drafted.

14 . The system of claim 9 , wherein the indication to optimize a language model includes information related to an intended recipient of the message, and wherein:
the message analysis module is further configured to determine a second topic based at least in part on the intended recipient of the message; and identify a word and/or phrase associated with the determined second topic,
wherein the language model optimization module is further configured to optimize the language model by adjusting a priority in the language model associated the identified word and/or phrase associated with the determined second topic.

15 . The system of claim 9 , wherein the message analysis module is further configured to determine that the topic is no longer active; and the language model optimization module is further configured to adjust the priority in the language model associated with the identified word and/or phrase to a previous priority level.

16 . The system of claim 9 , wherein the indication to optimize a language model is generated by a language prediction application operating on a device.

17 . A computer-implemented method for optimizing a language model based on a topic anticipated in a correspondence message being drafted, the method performed by a processor executing instructions stored in a memory, the method comprising:
receiving an indication to optimize a language model,
wherein the language model is to be optimized based at least in part on an anticipated topic of a correspondence message being drafted,
wherein the indication to optimize the language model includes an intended recipient of the correspondence message;
selecting a language model to be optimized; determining an anticipated topic based at least in part on the intended recipient of the correspondence message; identifying a word and/or phrase associated with the determined topic; optimizing the language model,
wherein optimizing the language model includes adjusting a prior

Assignees

• Nuance Communications Inc

Inventors

• Mcsherry Michael
• Balasubramanian Sundar

Classifications

• G06F40/274 Primary

  Converting codes to words; Guess-ahead of partial word inputs · CPC title

• G06F40/284 Primary

  Lexical analysis, e.g. tokenisation or collocates · CPC title

• G06F17/28 Primary

  Physics · mapped topic

• G06F17/277

  Physics · mapped topic