Abstract: Systems and methods can determine product lines product searches. One or more computing devices can receive a product query of search terms. The product query may be classified to identify a product category. A brand may be identified for the product query. The brand may be selected from a list of known brands for the product category. One or more unknown product line terms may be identified within the product query. A metric may be computed to indicate how well the unknown product line terms correspond to an actual product line within the brand. The metric may be compared to a specified threshold. The unknown product line terms may be designated as a new product line of the brand if the metric favorably compares to the specified threshold. A product search may be performed on the product query. Product search results may be returned according to the product search.

Abstract: Rendering graphical user interfaces display query results based on latent intent of users comprises a query network system. The query server stores data for each result of a group of search results, the product data comprising one or more items of metadata that are usable by the one or more computing devices and are not presented with the results for display to a user computing device. The server correlates items of metadata associated with results selected after first query to determine latent intent of the user. The server receives a second query and determines that the second query includes terms related to the first query. The server determines a latent intent of the second query based on the correlation and provides instructions to the user computing device to render a graphical user interface, the graphical user interface comprising the query results.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for predicting object identity using an ensemble of predictors. In one aspect, a method includes selecting candidate objects that likely match a received object that is to be identified, from a database of objects, and providing attributes of the received object compared with those of the candidates to an ensemble of predictors having respective properties. Based on previous training, each predictor can predict a most likely candidate. From among the most likely candidates, a previously trained support vector machine can select a potential match candidate. If a score that the support vector machine associates with the potential match candidate, that is representative of the potential match candidate's likelihood to match the received candidate satisfies a threshold, then the potential match candidate can be determined to be the received candidate.

Abstract: Systems and methods can determine product lines product searches. One or more computing devices can receive a product query of search terms. The product query may be classified to identify a product category. A brand may be identified for the product query. The brand may be selected from a list of known brands for the product category. One or more unknown product line terms may be identified within the product query. A metric may be computed to indicate how well the unknown product line terms correspond to an actual product line within the brand. The metric may be compared to a specified threshold. The unknown product line terms may be designated as a new product line of the brand if the metric favorably compares to the specified threshold. A product search may be performed on the product query. Product search results may be returned according to the product search.

Abstract: The aesthetic quality of a picture is automatically inferred using visual content as a machine learning problem using, for example, a peer-rated, on-line photo sharing Website as data source. Certain visual features of images are extracted based on the intuition that they can discriminate between aesthetically pleasing and displeasing images. A one-dimensional support vector machine is used to identify features that have noticeable correlation with the community-based aesthetics ratings. Automated classifiers are constructed using the support vector machines and classification trees, with a simple feature selection heuristic being applied to eliminate irrelevant features. Linear regression on polynomial terms of the features is also applied to infer numerical aesthetics ratings.

Abstract: Systems and methods can determine categories for product searches. One or more computing devices can receive a product query of search terms. The product query can be classified to identify a product category. The search terms may be verified against an ambiguous term list for the product category. The search terms may also be verified against an attribute list for the product category. The product query may be classified as fully understood in response to all of the search terms matching either the ambiguous term list or the attribute list for the product category. A product search may be performed on the product query. The product search may be informed by the product category when the product query has been classified as fully understood. Search results may be generated and returned according to the product search.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for predicting object identity using an ensemble of predictors. In one aspect, a method includes selecting candidate objects that likely match a received object that is to be identified, from a database of objects, and providing attributes of the received object compared with those of the candidates to an ensemble of predictors having respective properties. Based on previous training, each predictor can predict a most likely candidate. From among the most likely candidates, a previously trained support vector machine can select a potential match candidate. If a score that the support vector machine associates with the potential match candidate, that is representative of the potential match candidate's likelihood to match the received candidate satisfies a threshold, then the potential match candidate can be determined to be the received candidate.

Abstract: The aesthetic quality of a picture is automatically inferred using visual content as a machine learning problem using, for example, a peer-rated, on-line photo sharing Website as data source. Certain visual features of images are extracted based on the intuition that they can discriminate between aesthetically pleasing and displeasing images. A one-dimensional support vector machine is used to identify features that have noticeable correlation with the community-based aesthetics ratings. Automated classifiers are constructed using the support vector machines and classification trees, with a simple feature selection heuristic being applied to eliminate irrelevant features. Linear regression on polynomial terms of the features is also applied to infer numerical aesthetics ratings.

Abstract: In a system and method for the generation of attack-resistant, user-friendly, image-based CAPTCHAs (Completely Automated Public test to Tell Computers and Humans Apart), controlled distortions are applied to randomly chosen images and presented to a user for annotation from a given list of words. An image is presented that contains multiple connected but independent images with the borders between them distorted or otherwise visually obfuscated in a way that a computer cannot distinguish the borders and a user selects near the center of one of the images. The distortions are performed in a way that satisfies the incongruous requirements of low perceptual degradation and high resistance to attack by content-based image retrieval systems. Word choices are carefully generated to avoid ambiguity as well as to avoid attacks based on the choices themselves.

Abstract: A principled, probabilistic approach to meta-learning acts as a go-between for a ‘black-box’ image annotation system and its users. Inspired by inductive transfer, the approach harnesses available information, including the black-box model's performance, the image representations, and a semantic lexicon ontology. Being computationally ‘lightweight.’ the meta-learner efficiently re-trains over time, to improve and/or adapt to changes. The black-box annotation model is not required to be re-trained, allowing computationally intensive algorithms to be used. Both batch and online annotation settings are accommodated. A “tagging over time” approach produces progressively better annotation, significantly outperforming the black-box as well as the static form of the meta-learner, on real-world data.

Abstract: The aesthetic quality of a picture is automatically inferred using visual content as a machine learning problem using, for example, a peer-rated, on-line photo sharing Website as data source. Certain visual features of images are extracted based on the intuition that they can discriminate between aesthetically pleasing and displeasing images. A one-dimensional support vector machine is used to identify features that have noticeable correlation with the community-based aesthetics ratings. Automated classifiers are constructed using the support vector machines and classification trees, with a simple feature selection heuristic being applied to eliminate irrelevant features. Linear regression on polynomial terms of the features is also applied to infer numerical aesthetics ratings.

Abstract: In a system and method for the generation of attack-resistant, user-friendly, image-based CAPTCHAs (Completely Automated Public test to Tell Computers and Humans Apart), controlled distortions are applied to randomly chosen images and presented to a user for annotation from a given list of words. An image is presented that contains multiple connected but independent images with the borders between them distorted or otherwise visually obfuscated in a way that a computer cannot distinguish the borders and a user selects near the center of one of the images The distortions are performed in a way that satisfies the incongruous requirements of low perceptual degradation and high resistance to attack by content-based image retrieval systems. Word choices are carefully generated to avoid ambiguity as well as to avoid attacks based on the choices themselves.