Method And System For Enhancing Voice Search Query On A Digital Platform

Flipkart Internet Private Limited
Granted App No 202241063463 Patent 561754 Communication

Patent information

IP India · live record
Application number
202241063463
Filing date
07 November 2022
Status
Granted
Legal status
Inforce
Publication number
04/2023
Publication type
INA
Invention field
Communication
Patent number
561754
Grant date
03 March 2025
Next renewal due
07 November 2027

Abstract

The present disclosure encompasses systems and methods for generating enhanced voice search query on an e-commerce platform. The method comprises receiving, at a transceiver unit [104], an audio query from a user on the digital platform. The method further comprises generating, by a processing unit [106], one or more text tokens based on the received audio query and sending the received audio query to a pre-trained first sub-system [108], and the converted text tokens to a pre-trained second sub-system [110]. The pre-trained first sub-system [108] then identifies one or more attributes of the user from the received audio query and generates a speaker embedding for the user based on the identified attributes. The pre-trained second sub-system [110] generates an enhanced query based on the speaker embedding and the text tokens.

Applicants

1
India
FLIPKART INTERNET PRIVATE LIMITED Buildings Alyssa, Begonia & Clover, Embassy Tech Village, Outer Ring Road, Deverabeesanahalli Village, Bengaluru - 560103, Karnataka, India

Inventors

2
India
Raviraj JoshiFlat A-1104 Polite Harmony, Akurdi-Chikhali Road, Nr Sane Chowk, Chikhali Pune Maharashtra India 411062
India
Gaurav GuptaFlat No 6 Ghati Jin Complex, Nutan nagar Khargone Madhya Pradesh India 451001

Claims

13
1 A method for enhancing voice search query on a digital platform, the method comprising: - receiving, by a transceiver [104], an audio query from a user on the digital platform; - generating, by a processing unit [106], one or more text tokens based on the received audio query; - sending, by the processing unit[106], the received audio query to a pre-trained first sub-system [108], and the converted text tokens to a pre-trained second sub-system [110]; - identifying, by the pre-trained first sub-system [108], one or more attributes of the user from the received audio query; - generating, by the pre-trained first sub-system [108], a speaker embedding for the user based on the identified attributes; - generating, by a second sub-system [110], an enhanced query based on the speaker embedding and the one or more text tokens.
2 The method as claimed in claim 1, wherein the one or more attributes include age, gender, demography, language.
3 The method as claimed in claim 1, wherein the pre-trained first subsystem [108] is trained based on a machine learning model.
4 The method as claimed in claim 1, wherein the pre-trained second subsystem [108] is trained based on a sequence to sequence model.
5 The method as claimed in claim 1, the method further comprising: - generating search results based on the enhanced query; and - providing, via a user interface, the generated search results on the digital platform.
6 The method as claimed in claim 1, wherein enhancing is based on a set of historical speaker embeddings and a set of historical vector sequences associated with the user, stored in a memory unit [114], in an event the voice of the same user is identified by the processing unit[106].
7 A system for enhancing voice search query on a digital platform, the system comprising: - a transceiver [104] configured to receive an audio query from a user on the digital platform; - a processing unit [106] configured to: o generate one or more text tokens based on the received audio query; and o send the received audio query to a pre-trained first subsystem [108], and the one or more text tokens to a pre-trained second sub-system [110]; wherein, - the pre-trained first sub-system [108] is configured to: o identify one or more attributes of the user from the received audio query; and o generate a speaker embedding for the user based on the identified one or more attributes; and - the pre-trained second sub-system [110] is configured to: o generate one or more vector sequences based on the received audio query; and o generate an enhanced query based on the speaker embedding and the one or more text tokens;
8 The system as claimed in claim 7, wherein the one or more attributes include age, gender, demography, language,....
9 The system as claimed in claim 7, wherein the pre-trained first subsystem [108] is trained based on a machine learning model.
10 The system as claimed in claim 7, wherein the pre-trained second subsystem [108] is trained based on a sequence to sequence model.
11 The system as claimed in claim 7, wherein the processing unit [106] is configured to generate search results based on the enhanced query.
12 The system as claimed in claim 11, the system further comprising a user interface [112] configured to provide the generated search results on the digital platform.
13 The system as claimed in claim 7, wherein enhancing is based on a set of historical speaker embeddings and a set of historical vector sequences associated with the user, stored in a memory unit [114], in an event the voice of the same user is identified by the processing unit[106].

Specification

Show full textCollapse

We Claim:

1. A method for enhancing voice search query on a digital platform, the method comprising:

- receiving, by a transceiver [104], an audio query from a user on the digital platform;

- generating, by a processing unit [106], one or more text tokens based on the received audio query;

- sending, by the processing unit[106], the received audio query to a pre-trained first sub-system [108], and the converted text tokens to a pre-trained second sub-system [110];

- identifying, by the pre-trained first sub-system [108], one or more attributes of the user from the received audio query;

- generating, by the pre-trained first sub-system [108], a speaker embedding for the user based on the identified attributes;

- generating, by a second sub-system [110], an enhanced query based on the speaker embedding and the one or more text tokens.

2. The method as claimed in claim 1, wherein the one or more attributes include age, gender, demography, language.

3. The method as claimed in claim 1, wherein the pre-trained first subsystem [108] is trained based on a machine learning model.

4. The method as claimed in claim 1, wherein the pre-trained second subsystem [108] is trained based on a sequence to sequence model.

5. The method as claimed in claim 1, the method further comprising:

- generating search results based on the enhanced query; and

- providing, via a user interface, the generated search results on the digital platform.

6. The method as claimed in claim 1, wherein enhancing is based on a set of historical speaker embeddings and a set of historical vector sequences associated with the user, stored in a memory unit [114], in an event the voice of the same user is identified by the processing unit[106].

7. A system for enhancing voice search query on a digital platform, the system comprising:

- a transceiver [104] configured to receive an audio query from a user on the digital platform;

- a processing unit [106] configured to:

o generate one or more text tokens based on the received audio query; and

o send the received audio query to a pre-trained first subsystem [108], and the one or more text tokens to a pre-trained second sub-system [110];

wherein,

- the pre-trained first sub-system [108] is configured to:

o identify one or more attributes of the user from the received audio query; and

o generate a speaker embedding for the user based on the identified one or more attributes; and

- the pre-trained second sub-system [110] is configured to:

o generate one or more vector sequences based on the received audio query; and

o generate an enhanced query based on the speaker embedding and the one or more text tokens;

8. The system as claimed in claim 7, wherein the one or more attributes include age, gender, demography, language,....

9. The system as claimed in claim 7, wherein the pre-trained first subsystem [108] is trained based on a machine learning model.

10. The system as claimed in claim 7, wherein the pre-trained second subsystem [108] is trained based on a sequence to sequence model. 11. The system as claimed in claim 7, wherein the processing unit [106] is configured to generate search results based on the enhanced query.

12. The system as claimed in claim 11, the system further comprising a user interface [112] configured to provide the generated search results on the digital platform.

13. The system as claimed in claim 7, wherein enhancing is based on a set of historical speaker embeddings and a set of historical vector sequences associated with the user, stored in a memory unit [114], in an event the voice of the same user is identified by the processing unit[106].

Documents

Sign in to open
Orders
Section
Application documents
1 202241063463-STATEMENT OF UNDERTAKING (FORM 3) [07-11-2022(online)].pdf 07 Nov 2022
2 202241063463-REQUEST FOR EXAMINATION (FORM-18) [07-11-2022(online)].pdf 07 Nov 2022
3 202241063463-REQUEST FOR EARLY PUBLICATION(FORM-9) [07-11-2022(online)].pdf 07 Nov 2022
4 202241063463-PROOF OF RIGHT [07-11-2022(online)].pdf 07 Nov 2022
5 202241063463-POWER OF AUTHORITY [07-11-2022(online)].pdf 07 Nov 2022
6 202241063463-FORM-9 [07-11-2022(online)].pdf 07 Nov 2022
7 202241063463-FORM 18 [07-11-2022(online)].pdf 07 Nov 2022
8 202241063463-FORM 1 [07-11-2022(online)].pdf 07 Nov 2022
9 202241063463-FIGURE OF ABSTRACT [07-11-2022(online)].pdf 07 Nov 2022
10 202241063463-DRAWINGS [07-11-2022(online)].pdf 07 Nov 2022
11 202241063463-DECLARATION OF INVENTORSHIP (FORM 5) [07-11-2022(online)].pdf 07 Nov 2022
12 202241063463-COMPLETE SPECIFICATION [07-11-2022(online)].pdf 07 Nov 2022
13 202241063463-Request Letter-Correspondence [09-11-2022(online)].pdf 09 Nov 2022
14 202241063463-Power of Attorney [09-11-2022(online)].pdf 09 Nov 2022
15 202241063463-Form 1 (Submitted on date of filing) [09-11-2022(online)].pdf 09 Nov 2022
16 202241063463-Covering Letter [09-11-2022(online)].pdf 09 Nov 2022
17 202241063463-Correspondence_Form-1 And POA_21-11-2022.pdf 21 Nov 2022
18 202241063463-FER.pdf 11 Jul 2023
19 202241063463-FER_SER_REPLY [29-12-2023(online)].pdf 29 Dec 2023
20 202241063463-CLAIMS [29-12-2023(online)].pdf 29 Dec 2023
21 202241063463-US(14)-HearingNotice-(HearingDate-14-02-2025).pdf 29 Jan 2025
22 202241063463-Correspondence to notify the Controller [31-01-2025(online)].pdf 31 Jan 2025
23 202241063463-FORM-26 [10-02-2025(online)].pdf 10 Feb 2025
24 202241063463-Written submissions and relevant documents [21-02-2025(online)].pdf 21 Feb 2025
25 202241063463-PatentCertificate03-03-2025.pdf 03 Mar 2025
26 202241063463-IntimationOfGrant03-03-2025.pdf 03 Mar 2025
Search strategy
1SearchHistoryE_10-07-2023.pdf

ERegister / renewals

Inforce
3rd Renewed 05 Mar 2025 Period 07 Nov 2024 – 07 Nov 2025 · CBR 12880 · Certificate 16890 ₹4,000
4th Renewed 08 Aug 2025 Period 07 Nov 2025 – 07 Nov 2026 · CBR 46957 · Certificate 58377 ₹4,000
5th Renewed 08 May 2026 Period 07 Nov 2026 – 07 Nov 2027 · CBR 38992 · Certificate 37588 ₹4,000