Abstract: Embodiments of a system and method for generating a classification model with a cost function having different penalties for false positives and false negatives are described. Embodiments may include perform machine learning operations on known duplicates and known non-duplicates to generate a classification model for classifying structured data items as duplicates or non-duplicates. Each duplicate may represent a pair of structured data items describing a common item; each non-duplicate may represent a pair of structured data items describing different items. Generation of the classification model may be performed based on a cost function that penalizes false positive misclassifications within the classification model differently than false negative misclassifications. Embodiments may also include evaluating the classification model to determine whether a candidate structured data item is a duplicate or non-duplicate.

Abstract: Systems and methods are provided for generating query mapping information that associates each of a number of search queries to a corresponding preferred query form to be used in generating search results. A number of queries previously submitted by users may be normalized, then grouped together with other queries sharing the same normalized form. A preferred query form for each group may then be selected. One or more inaccurate mappings may be identified in the initial mapping results based on an analysis of the user behavior of users who previously submitted the search queries included in a given query mapping. A final set of mapped queries may then be generated for use in responding to subsequent search requests, where the final set includes a number of mappings that each associate a particular user-submitted query with a corresponding preferred query form.

Abstract: Disclosed are various embodiments for identifying relevant topics for an item from search queries. Search queries are obtained from users to search a collection of user reviews for a specific item. Relevant topics for the specific item are identified by analyzing the queries. A user interface is generated based at least in part on at least some of the relevant topics.

Abstract: Relevant topics for an item may be extracted from customer review search queries. Customer review search queries are obtained from customers to search a collection of customer reviews for a specific item. Relevant topics for the specific item are extracted by analyzing the queries. A user interface is generated based at least in part on at least some of the relevant topics.

Abstract: Aspects of the present disclosure relate to management of evaluated rule data sets. Specifically, a unreduced evaluated rule data set may contain a number of items to be compared or analyzed according to a number of rules, and may also contain the results of such analysis. An illustrative reduced evaluated data set can include the results of evaluated rules. When utilized in conjunction with an item data set and a rule data set, the information contained within the unreduced evaluated rule data set may be maintained. The reduce memory requirements of the reduced evaluated rule data set may facilitate storage of the reduced evaluated rule data set in faster to access memory, or may facilitate distributed computation of the reduced evaluated rule data set.

Abstract: Aspects of the present disclosure relate to management of evaluated rule data sets. Specifically, a unreduced evaluated rule data set may contain a number of items to be compared or analyzed according to a number of rules, and may also contain the results of such analysis. An illustrative reduced evaluated data set can include the results of evaluated rules. When utilized in conjunction with an item data set and a rule data set, the information contained within the unreduced evaluated rule data set may be maintained. The reduce memory requirements of the reduced evaluated rule data set may facilitate storage of the reduced evaluated rule data set in faster to access memory, or may facilitate distributed computation of the reduced evaluated rule data set.

Abstract: Aspects of the present disclosure relate to management of evaluated rule data sets. Specifically, a unreduced evaluated rule data set may contain a number of items to be compared or analyzed according to a number of rules, and may also contain the results of such analysis. An illustrative reduced evaluated data set can include the results of evaluated rules. When utilized in conjunction with an item data set and a rule data set, the information contained within the unreduced evaluated rule data set may be maintained. The reduce memory requirements of the reduced evaluated rule data set may facilitate storage of the reduced evaluated rule data set in faster to access memory, or may facilitate distributed computation of the reduced evaluated rule data set.

Abstract: Embodiments of a system and method for seeding rule-based machine learning models include generating a set of seed rules including one or more rules resulting from one or more previously performed machine learning operations and one or more randomly or pseudo-randomly generated rules. Embodiments may include performing one or more machine learning operations on the set of seed rules to generate a new set of rules for determining whether a pair of information items have a specific relationship. Generating the new set of rules from the seed rules may be faster than generating a set of rules from random data. Embodiments may also include applying the new set of rules to one or more pairs of information items to identify at least one pair of information items as having the specific relationship. For instance, the rules may be applied to identify pairs of information items that are duplicate pairs.

Abstract: Embodiments of a system and method for identifying and correcting marginal false positives in machine learning models may include, based on reference data that includes pairs of information items and labels indicating whether pairs of information items have a specific relationship, generating a first machine learning model for determining whether pairs of information items have that relationship. Embodiments may include identifying one or more false positive pairs (e.g., a pair of information items that the first machine learning model indicates as having the specific relationship and which are labeled within the reference data as not having that relationship). Embodiments may include selecting identified false positive pairs as candidates for correction.

Abstract: Embodiments of a system and method for identifying structured data items lacking requisite information for rule-based duplicate detection are described. Embodiments may include generating a deficiency score for each of multiple structured data items including applying a set of rules based on duplicate detection techniques to each given structured data item in order to perform a comparison of the given structured data item to itself. The deficiency score of the given structured data item may be based on a result of the comparison. Embodiments may also include, based on the deficiency scores of the structured data items, identifying one or more deficient structured data items having less than a requisite quantity of information for performing duplicate detection on structured data items. Embodiments may also include identifying one or more key attributes missing from some of the one or more deficient structured data items and requesting those key attributes.

Abstract: A device, system, and method are directed towards determining network information. A network address is determined for a possible proxy. A determination is made whether a port on the possible proxy is open and/or if the port supports an HyperText Transfer Protocol (HTTP) proxy request. A request is sent to the possible proxy over the port, the request being configured to be forwarded to a network device. A type of the possible proxy is determined based in part on a behavior of the network device. The behavior may indicate whether the request is received by the network device, or whether the possible proxy obscures an origin of the request. The proxy type may include whether the possible proxy is a non-proxy, an anonymous-proxy, a controlled-proxy, and/or an open-proxy. Various types of network analysis may then be performed using the possible proxy and the determined proxy type.

Abstract: A device, system, and method are directed towards determining network information. A network address is determined for a possible proxy. A determination is made whether a port on the possible proxy is open and/or if the port supports an HyperText Transfer Protocol (HTTP) proxy request. A request is sent to the possible proxy over the port, the request being configured to be forwarded to a network device. A type of the possible proxy is determined based in part on a behavior of the network device. The behavior may indicate whether the request is received by the network device, or whether the possible proxy obscures an origin of the request. The proxy type may include whether the possible proxy is a non-proxy, an anonymous-proxy, a controlled-proxy, and/or an open-proxy. Various types of network analysis may then be performed using the possible proxy and the determined proxy type.