Context-based review translation

Patent metadata

Field	Value
Publication number	US-12561997-B2
Application number	US-202217857416-A
Country	US
Kind code	B2
Filing date	Jul 5, 2022
Priority date	Jul 5, 2022
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

A translation system provides machine translations of review texts on item pages using context from the item pages outside of the review text being translated. Given review text from an item page, context for machine translating the review text is determined from the item page. In some aspects, one or more keywords are determined based on text, images, and/or videos on the item page. The one or more keywords are used as context by the machine translator to translate the review text from a first language to a second language to provide translated review text, which can be presented on the item page.

First claim

What is claimed is:

1 . One or more computer storage media storing computer-useable instructions that, when used by a computing device, cause the computing device to perform operations, the operations comprising:
receiving review text from an item page for an item offered on a listing platform, the item page being a webpage from the listing platform that presents information describing the item and includes one or more user reviews of the item, the review text comprising a user review of the item in a first language; determining one or more keywords describing the item from the information describing the item located on a portion of the item page outside of the review text, the information describing the item comprising one or more selected from the following: textual content describing the item, one or more images of the item, and one or more videos showing the item; translating, by a machine translator, the review text from the first language to a second language using the one or more keywords describing the item as context to disambiguate one or more terms in the review text and improve translation accuracy when translating the review text to provide translated review text in the second language, the translated review text comprising the user review of the item in the second language; and providing the translated review text for presentation in association with the item page.

2 . The computer storage media of claim 1 , wherein determining the one or more keywords comprises:
accessing item page text outside of the review text; and identifying one or more important words from the item page text as the one or more keywords.

3 . The computer storage media of claim 2 , wherein the item page text comprises at least one selected from the following: an item title, and an item description.

4 . The computer storage media of claim 2 , wherein identifying the one or more important words from the item page text comprises:
computing a term frequency-inverse document frequency (TF-IDF) score for each word from a plurality of words from the item page text; and selecting one or more words having a TF-IDF score above a threshold as the one or more important words.

5 . The computer storage media of claim 1 , wherein determining the one or more keywords comprises:
accessing an image on the item page; identifying, using an object detector, an object in the image and labeling the object with a label identifying the object; and selecting the label as one of the one or more keywords.

6 . The computer storage media of claim 1 , wherein determining the one or more keywords comprises:
accessing a video on the item page; selecting a first frame from the video; identifying, using an object detector, an object in the first frame and labeling the object with a label identifying the object; and selecting the label as one of the one or more keywords.

7 . The computer storage media of claim 6 , wherein selecting the first frame in the video comprises:
comparing features of the first frame to features of one or more other frames to determine the first frame and the one or more other frames correspond to a scene; and selecting the first frame as a representative frame for the scene.

8 . The computer storage media of claim 1 , wherein determining the one or more keywords comprises:
identifying a plurality of potential keywords from the information describing the item determining a relevance score for each potential keyword from the plurality of potential keywords; and selecting a first potential keyword as one of the one or more keywords based on the relevance score for the first potential keyword.

9 . The computer storage media of claim 8 , wherein determining the relevance score for the first potential keyword comprises computing a similarity between the first potential keyword and an item title for an item presented by the item page.

10 . A computer-implemented method comprising:
receiving review text from an item page for an item offered on a listing platform, the item page being a webpage from the listing platform that presents information describing the item and includes one or more user reviews of the item, the review text comprising a user review of the item in a first language; determining, by a text extractor, a first candidate set of one or more keywords describing the item from text on an item page outside of the review text to be translated; determining, by an image extractor, a second candidate set of one or more keywords describing the item from an image of the item on the image page; determining, by a video extractor, a third candidate set of one or more keywords describing the item from a frame of a video showing the item on the item page; selecting, by a keyword selection component, a set of one or more keywords describing the item from the first candidate set, the second candidate set, and the third candidate set; translating, by a machine translator, the review text from the first language to a second language using the set of one or more keywords describing the item as context to disambiguate one or more terms in the review text and improve translation accuracy when translating the review text to provide translated review text in the second language, the translated review text comprising the user review of the item in the second language; and providing, via a user interface component, the translated review text for presentation in association with the item page.

11 . The method of claim 10 , wherein the text extractor determines the first candidate set of one or more keywords by:
identifying one or more important words from the text on the item page outside of the review text as one or more candidate keywords.

12 . The method of claim 11 , wherein the one or more important words are identified based on a term frequency-inverse document frequency (TF-IDF) score determined for each word from a plurality of words from the text from the item page outside of the review text.

13 . The method of claim 10 , wherein the image extractor determines the second candidate set of one or more keywords by:
identifying, using an object detector, an object in the image and labeling the object with a label identifying the object; and selecting the label as a candidate keyword.

14 . The method of claim 10 , wherein the video extractor determines the third set of one or more keywords by:
selecting a first frame from the video; identifying, using an object detector, an object in the first frame and labeling the object with a label identifying the object; and selecting the label as a candidate keyword.

15 . The method of claim 10 , wherein the keyword selector selects the set of one or more keywords by:
determining a relevance score for each keyword from the first candidate set, the second candidate set, and the third candidate set; and selecting a first keyword as one of the one or more keywords based on the relevance score for the first keyword.

16 . A computer system comprising:
a processor; and a computer storage medium storing computer-useable instructions that, when used by the processor, causes the computer system to perform operations comprising: receiving review text from an item page for an item offered on a listing platform, the item page being a webpage from the listing platform that presents information describing the item and includes one or more user reviews of the item, the review text comprising a user review of the item in a first language; determining, by a keyword component, one or more keywords describing the item from text

Assignees

• Adobe Inc

Inventors

• Singhal Gourav
• Jindal Amol

Classifications

• G06V20/46

  Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames · CPC title

• G06F40/279

  Recognition of textual entities · CPC title

• G06F40/58 Primary

  Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation · CPC title

• G06V20/63

  Scene text, e.g. street names · CPC title

• G06N3/08

  Learning methods · CPC title