Matching text to images
Text in web pages or other text documents may be classified based on the images or other objects within the webpage. A system for identifying and classifying text related to an object may identify one or more web pages containing the image or similar images, determine topics from the text of the document, and develop a set of training phrases for a classifier. The classifier may be trained and then used to analyze the text in the documents. The training set may include both positive examples and negative examples of text taken from the set of documents. A positive example may include captions or other elements directly associated with the object, while negative examples may include text taken from the documents, but from a large distance from the object. In some cases, the system may iterate on the classification process to refine the results.
1. A method performed at least in part on at least one computer processor, said method comprising:
receiving an image to classify, said image being located within a text document;
identifying a plurality of said text documents comprising said image;
identifying a training set of examples from at least one of said text documents, said training set of examples comprising a subset of said text within said text documents that is classified based on a determination of a measure of proximity to said image;
training a classifier using said training set; and
classifying said text within said text document using said classifier to identify a group of text associated with said image.
2. The method of claim 1 , wherein:
the determination of the measure of proximity to said image is based on a determination of a predefined measure of physical proximity of the subset of said text to said image, in a laid out representation of the text document, in a rendering of the text document for presentation.
3. The method of claim 2 , wherein:
the determination of the measure of proximity to said image is based on a determination that the subset of said text to said image is physically close to said image in the laid out representation, based on a predefined threshold value of physical closeness, wherein the subset of said text is determined as a positive example, if the subset of said text is determined to be physically close to said image.
4. The method of claim 2 , wherein:
the determination of the measure of proximity to said image is based on a determination that the subset of said text to said image is physically far away from said image in the laid out representation, based on a predefined threshold value of physical distance, wherein the subset of said text is determined as a negative example, if the subset of said text is determined to be physically far away from said image.
5. The method of claim 1 , wherein:
the determination of the measure of proximity to said image is based on a determination of a predefined measure of proximity of the subset of said text to said image, based on a count of a number of nodes in a document object model (DOM), in a path from a first node of the DOM that represents the subset of said text, to a second node of the DOM that represents said image.
6. The method of claim 5 , wherein:
the determination of the measure of proximity to said image is based on a determination that the first node is close to the second node said in the DOM, based on a predefined threshold value of physical closeness, based on the count, wherein the subset of said text is determined as a positive example, if the subset of said text is determined to be close to said image.
7. The method of claim 5 , wherein:
the determination of the measure of proximity to said image is based on a determination that the first node is far away from the second node said in the DOM, based on a predefined threshold value of distance, based on the count, wherein the subset of said text is determined as a negative example, if the subset of said text is determined to be far away from said image.
8. The method of claim 1 , wherein said text document comprises a HyperText Markup Language (HTML) document.
9. A system comprising:
a computer processor;
an object classifier operable on said processor, said object classifier that:
receives a set of text documents and identifies a common object to classify, said common object being comprised in each of said text documents;
identifies a training set of examples from at least one of said text documents, said training set of examples comprising a subset of text within said text documents, the subset of text being classified based on a determination of a measure of proximity to the common object;
trains a classifier using said training set; and
classifies said text within said text document using said classifier to identify a group of text associated with said object.
10. The system of claim 9 , wherein:
the determination of the measure of proximity to said common object is based on a determination of a predefined measure of physical proximity of the subset of said text to the common object, in laid out representations of the text documents, in renderings of the text documents for presentation.
11. The system of claim 10 , wherein:
the determination of the measure of proximity to said common object is based on a determination that the subset of said text to said common object is physically close to said common object in the laid out representations, based on a predefined threshold value of physical closeness, wherein the subset of said text is determined as a positive example, if the subset of said text is determined to be physically close to said common object.
12. The system of claim 10 , wherein:
the determination of the measure of proximity to said common object is based on a determination that the subset of said text to said common object is physically far away from said common object in the laid out representations, based on a predefined threshold value of physical distance, wherein the subset of said text is determined as a negative example, if the subset of said text is determined to be physically far away from said common object.
13. The system of claim 9 , wherein:
the determination of the measure of proximity to said common object is based on a determination of a predefined measure of proximity of the subset of said text to the common object, based on a count of a number of nodes in a document object model (DOM), in a path from a first node of the DOM that represents the subset of said text, to a second node of the DOM that represents the common object.
14. The system of claim 13 , wherein:
the determination of the measure of proximity to said common object is based on a determination that the first node is close to the second node said in the DOM, based on a predefined threshold value of physical closeness, based on the count, wherein the subset of said text is determined as a positive example, if the subset of said text is determined to be close to the common object.
15. The system of claim 13 , wherein:
the determination of the measure of proximity to said common object is based on a determination that the first node is far away from the second node said in the DOM, based on a predefined threshold value of physical distance, based on the count, wherein the subset of said text is determined as a negative example, if the subset of said text is determined to be far away from the common object.
16. The system of claim 9 , wherein at least some of said text documents include HyperText Markup Language (HTML) documents.
17. A method performed at least in part on at least one processor, said method comprising:
receiving an image to classify, said image being located within a first web page;
identifying a plurality of web pages comprising said image by transmitting a search request to a search system and returning said plurality of web pages, the first web page included in said plurality of web pages;
identifying a training set of examples from at least one of said web pages, said training set of examples comprising a subset of said text within said plurality of web pages, the subset of said text being classified based on a determination of a measure of proximity to said image, said training set comprising at least one positive example and at least one negative example;
training a classifier using said training set, said classifier being a binary classifier; and
classifying said text within said plurality of web pages using said classifier to identify a group of text associated with said image.
18. The method of claim 17 , wherein:
the determination of the measure of proximity to said image is based on a determination of a predefined measure of physical proximity of the subset of said text to said image, in a laid out representation of the web page, in a rendering of the web page for presentation.
19. The method of claim 17 , wherein:
the determination of the measure of proximity to said image is based on a determination of a predefined measure of proximity of the subset of said text to said image, based on a count of a number of nodes in a document object model (DOM), in a path from a first node of the DOM that represents the subset of said text, to a second node of the DOM that represents said image.
20. The method of claim 17 , wherein said web page comprises a HyperText Markup Language (HTML) document.