Abstract: A mobile device application prompts and conducts audio and/or video tests using a microphone on a smartphone, tablet or laptop in order to record and analyze a patient's speech, cough, breathing and other sounds in order to diagnose the patient with Covid 19, another ailment, or as having normal ranges not indicative of disease. The mobile device's tests and protocols use program instructions, AI processing and other automated tools to facilitate the speed and reliability of the testing.

Abstract: A semantic analysis can be performed to determine an intent of a received query. The intent can relate to a primary object of the query, which can be identified through the semantic analysis. Other attributes can be determined from the query that help to focus the object of the intent. A query vector is generated, based on the intent and primary object, and used to search a multi-dimensional semantic space including semantic representations of possible matches. The attributes are used to adjust the query vector in the semantic space. Objects having vectors ending proximate the query vector are identified as potential search results, with the distance from the query vector being used as a ranking mechanism. If refinement is needed, a dialog is used to obtain additional information from the user. Once results are obtained with sufficient confidence, results can be returned as search results for the query.

Abstract: In a flow of computer actions, a computer system (110) receives a request involving a machine translation. In performing the translation (160, 238), or in using the translation in subsequent computer operations (242, 1110), the computer system takes into account known statistical relationships (310), obtained from previously accumulated click-through data (180), between a machine translation performed in a flow, the flow's portions preceding the translation, and success indicators pertaining to the flow's portion following the translation. The statistical relationships are derived by data mining of the click-through data. Further, normal actions can be suspended to use a random option to accumulate the click-through data and/or perform statistical AB testing. Other features are also provided.

Abstract: A method of propagating annotations of content items to a search query is disclosed. A strength of a correspondence between a search query and a listing of an item on a network-based publication system is determined. The strength of the correspondence is based on an analysis of a set of actions by a set of users who submitted the search query. A set of annotations is generated. The set of annotations is propagated to a search engine and used to enhance search results.

Abstract: Information is provided a user of a mobile device. Media content is sensed by the mobile device and sent to an information processing server. The information processing server also obtains media content from another source and associates at least a portion of the obtained media content with a product or service offer. The sensed media content is correlated to the obtained media content such that an associated buy or service offer is selected. The buy or service offer is sent to the mobile device for display to the user.

Abstract: According to various embodiments, the Query Context Translation Engine identifies a topic of a search query history received during a current user session. The search query history in a first language. The Query Context Translation Engine identifies, in a translation table, target text that corresponds with a query in the search query history, the target text comprising at least one word. The Query Context Translation Engine obtains at least one search result based on a translation of the target text in a second language.

Abstract: A user query for items is received in a first language and translated from the first language to a second language. A result set in the second language that meets the query is obtained and is translated into the first language for presentation to the user. User feedback is used to build an ontology for optimizing the translation from the first language to the second language based on query context and the feedback. Query context may include information determined by learning semantic relationships between keywords in the query. Optimizing may include building an ontology used by a machine translator to translate key words from the first language to the second language. The number of items in the result set are measured or information is abstracted from the feedback and correlated to ontological information of the result set. The system adapts to changes in meanings in the first language over time.

Abstract: A method of propagating annotations of content items to a search query is disclosed. A strength of a correspondence between a search query and a listing of an item on a network-based publication system is determined. The strength of the correspondence is based on an analysis of a set of actions by a set of users who submitted the search query. A set of annotations is generated. The set of annotations is propagated to a search engine and used to enhance search results.

Abstract: A method of forming parallel corpora comprises receiving sets of items in first language and second languages, each of the sets having one or more associated descriptions and metadata. The metadata is collected from the two sets of items and are aligned using the metadata. The aligned metadata are mapped from the first language to the second language for each of the sets. The descriptions of two items are fetched and the structural similarity of the descriptions is measured to assess whether two items are likely to be translations of each other. For mapped items with structurally similar descriptions, the mapped item descriptions are formed into respective sentences in first language and in the second language. The sentences are parallel corpora which may be used to translate an item from the first language to the second language, and also to train a machine translation system.

Abstract: According to various embodiments, the Query Context Translation Engine identifies a topic of a search query history received during a current user session. The search query history in a first language. The Query Context Translation Engine identifies, in a translation table, target text that corresponds with a query in the search query history, the target text comprising at least one word. The Query Context Translation Engine obtains at least one search result based on a translation of the target text in a second language.

Abstract: In various example embodiments, a system and method for a Listing Engine that translates a first listing from a first language to a second language. The first listing includes an image(s) of a first item. The Listing Engine provides as input to an encoded neural network model a portion(s) of a translated first listing and a portions(s) of a second listing in the second language. The second listing includes an image(s) of a second item. The Listing Engine receives from the encoded neural network model a first feature vector for the translated first listing and a second feature vector for the second listing. The first and the second feature vectors both include at least one type of image signature feature and at least one type of listing text-based feature.

Abstract: In a flow of computer actions, a computer system (110) receives a request involving a machine translation. In performing the translation (160, 238), or in using the translation in subsequent computer operations (242, 1110), the computer system takes into account known statistical relationships (310), obtained from previously accumulated click-through data (180), between a machine translation performed in a flow, the flow's portions preceding the translation, and success indicators pertaining to the flow's portion following the translation. The statistical relationships are derived by data mining of the click-through data. Further, normal actions can be suspended to use a random option to accumulate the click-through data and/or perform statistical AB testing. Other features are also provided.

Abstract: In various example embodiments, a system and method for a Listing Engine that translates a first listing from a first language to a second language. The first listing includes an image(s) of a first item. The Listing Engine provides as input to an encoded neural network model a portion(s) of a translated first listing and a portions(s) of a second listing in the second language. The second listing includes an image(s) of a second item. The Listing Engine receives from the encoded neural network model a first feature vector for the translated first listing and a second feature vector for the second listing. The first and the second feature vectors both include at least one type of image signature feature and at least one type of listing text-based feature.

Abstract: Deep Neural Networks (DNN) are time shifted relative to one another and trained. The time-shifted networks may then be combined to improve recognition accuracy. The approach is based on an automatic speech recognition (ASR) system using DNN and using time shifted features. Initially, a regular ASR model is trained to produce a first trained DNN. Then a top layer (e.g., SoftMax layer) and the last hidden layer (e.g., Sigmoid) are fine-tuned with same data set but with a feature window left- and right-shifted to create respective second and third left-shifted and right-shifted DNNs. From these three DNN networks, four combination networks may be generated: left- and right-shifted, left-shifted and centered, centered and right-shifted, and left-shifted, centered, and right-shifted. The centered networks are used to perform the initial (first-pass) ASR. Then the other six networks are used to perform rescoring.

Abstract: A hybrid frame, phone, diphone, morpheme, and word-level Deep Neural Networks (DNN) in model training and applications-is based on training a regular ASR system, which can be based on Gaussian Mixture Models (GMM) or DNN. All the training data (in the format of features) are aligned with the transcripts in terms of phonemes and words with the timing information and new features are formed in terms of phonemes, diphones, morphemes, and up to words. Regular ASR produces a result lattice with timing information for each word. A feature is then extracted and sent to the word-level DNN for scoring Phoneme features are sent to corresponding DNNs for training. Scores are combined to form the word level scores, a rescored lattice and a new recognition result.

Abstract: A user query for items is received in a first language and translated from the first language to a second language. A result set in the second language that meets the query is obtained and is translated into the first language for presentation to the user. User feedback is used to build an ontology for optimizing the translation from the first language to the second language based on query context and the feedback. Query context may include information determined by learning semantic relationships between keywords in the query. Optimizing may include building an ontology used by a machine translator to translate key words from the first language to the second language. The number of items in the result set are measured or information is abstracted from the feedback and correlated to ontological information of the result set. The system adapts to changes in meanings in the first language over time.

Abstract: A method of forming parallel corpora comprises receiving sets of items in first language and second languages, each of the sets having one or more associated descriptions and metadata. The metadata is collected from the two sets of items and are aligned using the metadata. The aligned metadata are mapped from the first language to the second language for each of the sets. The descriptions of two items are fetched and the structural similarity of the descriptions is measured to assess whether two items are likely to be translations of each other. For mapped items with structurally similar descriptions, the mapped item descriptions are formed into respective sentences in first language and in the second language. The sentences are parallel corpora which may be used to translate an item from the first language to the second language, and also to train a machine translation system.

Abstract: A user query for items is received in a first language and translated from the first language to a second language. A result set in the second language that meets the query is obtained and is translated into the first language for presentation to the user. User feedback is used to build an ontology for optimizing the translation from the first language to the second language based on query context and the feedback. Query context may include information determined by learning semantic relationships between keywords in the query. Optimizing may include building an ontology used by a machine translator to translate key words from the first language to the second language. The number of items in the result set are measured or information is abstracted from the feedback and correlated to ontological information of the result set. The system adapts to changes in meanings in the first language over time.

Abstract: Deep Neural Networks (DNN) are time shifted relative to one another and trained. The time-shifted networks may then be combined to improve recognition accuracy. The approach is based on an automatic speech recognition (ASR) system using DNN and using time shifted features. Initially, a regular ASR model is trained to produce a first trained DNN. Then a top layer (e.g., SoftMax layer) and the last hidden layer (e.g., Sigmoid) are fine-tuned with same data set but with a feature window left- and right-shifted to create respective second and third left-shifted and right-shifted DNNs. From these three DNN networks, four combination networks may be generated: left- and right-shifted, left-shifted and centered, centered and right-shifted, and left-shifted, centered, and right-shifted. The centered networks are used to perform the initial (first-pass) ASR. Then the other six networks are used to perform rescoring.

Abstract: In various example embodiments, a system and method for a Listing Engine that translates a first listing from a first language to a second language. The first listing includes an image(s) of a first item. The Listing Engine provides as input to an encoded neural network model a portion(s) of a translated first listing and a portions(s) of a second listing in the second language. The second listing includes an image(s) of a second item. The Listing Engine receives from the encoded neural network model a first feature vector for the translated first listing and a second feature vector for the second listing. The first and the second feature vectors both include at least one type of image signature feature and at least one type of listing text-based feature.