Abstract: Embodiments provide a computer implemented method, in a data processing system comprising a processor and a memory comprising instructions which are executed by the processor to cause the processor to implement a full question generation system, the method comprising ingesting a query dataset derived from one or more search queries entered by one or more users of an internet search engine; identifying questions from the ingested query dataset; separating one or more prior full questions from the ingested dataset; identifying one or more question intent queries from the query dataset; for each identified question intent query; sorting the question intent query into one or more bins based on one or more missing interrogative words; and appending the missing interrogative word and a verb onto the question intent query to generate a full question. The full question generation method can additionally generate un-canonical questions.

Abstract: Examples of techniques for determining a reading speed based on user behavior are disclosed. In one example implementation according to aspects of the present disclosure, a computer-implemented method may include monitoring, by a processing device, data continuously, wherein the data is generated by a sensor of the user device; determining, by the processing device, an orientation and a wobble of the user device based on the data generated by the sensor; determining, by the processing device, a reading speed of a user using the user device based on the orientation and the wobble; detecting, by the processing device, a change in at least one of the orientation and the wobble of the user device based on the data; and adjusting, by the processing device, the reading speed of the user using the user device based on the change in the at least one of the orientation and the wobble.

Abstract: Computer-based processes are disclosed for analyzing and improving document readability. Document readability is improved by using rules and associated logic to automatically detect various types of writing problems and to make and/or suggest edits for eliminating such problems. Many of the rules seek to generate more concise formulations of the analyzed sentences, such as by eliminating unnecessary words, rearranging words and phrases, and making various other types of edits. Proposed edits can be conveyed, e.g., through a word processing platform, by changing the visual appearance of text to indicate how the text would appear with (or with and without) the edit.

Abstract: An electronic payment and authentication system that includes features to verify the authenticity of a payer, validate debit data, and facilitate debit (as opposed to credit) payment transactions. The electronic payment system allows a consumer to purchase items and services via a network while utilizing a debit transaction. The invention provides merchants with the ability to verify the identity of the consumer, determine if the consumer is a good debit transaction risk, and retrieve debit data associated with a particular consumer.

Abstract: A primary ingestion pipeline configured for use in natural language processing includes annotators configured for annotating documents. The annotators and documents to be annotated are evaluated. Based on the evaluations, an ingestion risk score is generated for each document. Each ingestion risk score represents a likelihood that an associated document will not successfully be annotated by the annotators. Each ingestion risk score is compared to a set of risk criteria. Based on the comparisons, a determination is made that each document of a first set of documents satisfies the set of risk criteria. A further determination is made, based on the comparisons, that each document of a second set of documents does not satisfy the set of risk criteria. In response to these determinations, the first set of documents is entered into the primary ingestion pipeline and the second set of documents is provided special handling.

Abstract: A paraphrase generation method according to the present disclosure generates one or more paraphrases of an original text by paraphrasing, within an acceptable limit for accepting paraphrasing, one or more of a plurality of fragments included in the original text into another expression in the language of the original text, the plurality of fragments being obtained by dividing the original text in accordance with a predetermined rule.

Abstract: Intelligent filtering of health-related information includes receiving health-related information including items encoded in one or more external standard terminologies. The health-related information is converted from the external standard terminologies into an internal medical terminology. Items within the health-related information are then identified that are related to a selected term of the internal medical terminology.

Abstract: According to one embodiment, a machine translation apparatus includes the following elements. The machine translation unit performs machine translation on a first text in a first language to generate a first machine translation result in a second language. The retrieval unit retrieves a first question sentence in the first language similar to the first text to obtain a degree of similarity between the first text and the first question sentence. The determination unit determines a first answer sentence in the first language corresponding to the first question sentence to be an output target when the degree of similarity is higher than a threshold and determines the first machine translation result to be an output target when the degree of similarity is lower than the threshold.

Abstract: A pruning method includes representing a set of sequences in a data structure. Each sequence s includes a first symbol w and a context c of at least one symbol. Some of the sequences are associated with a conditional probability p(w|c), based on observations of cw in training data. For others, p(w|c) is computed as a function of the probability p(w|?) of the respective symbol w in a back-off context ?, p(w|?) being based on observations of sequence ?w in the training data. A scoring function ƒ(cw) value is computed for each sequence in the set, based on p(w|c) for the sequence and a probability distribution p(s) of each symbol in the sequence if it is removed from the set of sequences. Iteratively, one of the represented sequences is selected to be removed, based on the computed scoring function values, and the scoring function values of remaining sequences are updated.

Abstract: Techniques for question answering involve receiving, from a user, a text input expressing a question in natural language. The text input may be analyzed, including identifying in the question at least one first portion answerable from at least one structured data source, and at least one second portion answerable from at least one unstructured data source. At least one first query configured for the structured data source(s) may be constructed from the at least one first portion of the question and applied to the structured data source(s) to retrieve first answer information for the at least one first portion of the question. At least one second query configured for the unstructured data source(s) may be constructed from the at least one second portion of the question and applied to the unstructured data source(s) to retrieve second answer information for the at least one second portion of the question.

Abstract: A system capable of performing natural language understanding (NLU) without the concept of a domain that influences NLU results. The present system uses a hierarchical organizations of intents/commands and entity types, and trained models associated with those hierarchies, so that commands and entity types may be determined for incoming text queries without necessarily determining a domain for the incoming text. The system thus operates in a domain agnostic manner, in a departure from multi-domain architecture NLU processing where a system determines NLU results for multiple domains simultaneously and then ranks them to determine which to select as the result.

Abstract: An apparatus comprising a memory and a processor configured for semi-autonomous natural language processing domain adaptation related activities. The processor coupled to the memory and configured to identify a corpus of documents of an evaluation domain and generate a first lexicon based on the corpus of documents of the evaluation domain, and determine a threshold that indicates a sufficiency of domain adaptation of the evaluation domain based at least in part on the first lexicon. The processor is further configured to identify a corpus of documents of a client domain, generate a second lexicon based on the corpus of documents of the client domain, determine a metric associated with the corpus of documents of the client domain and the second lexicon, and determine that domain adaptation of the client domain is complete when the metric exceeds the threshold.

Abstract: Techniques for navigational aid for a hinged device via semantic abstraction are described. Generally, the techniques described herein improve a user experience when the user is navigating through content, such as user-generated content in an electronic document. For example, the techniques described herein semantically abstract authored content in an electronic document to provide abstracted content. In implementations, abstracted content includes abstracted pages that each represent a different section of the authored content. When a user scans through an electronic document, rather than scanning page by page, techniques described herein instead navigate through displays of the abstracted pages. In addition, a hinge between different displays can be used as an input mechanism to control a speed of navigating through the abstracted pages to allow a user to more easily locate specific sections of the authored content.

Abstract: A log record from a host machine node includes an invariant string and a term. A template identifier is selected, from among template identifiers within a template repository, for a template string matching the invariant string. A sampling count threshold is selected from among a set of sampling count thresholds based on the template identifier and the term. A template-term count is obtained based on a number of earlier log records that were received since the count was reset and have a template identifier and a term that match the template identifier and the term of the log record. Based on the template-term count satisfying the sampling count threshold, an index entry is generated in a sampled log records index based on the log record and the template-term count is reset to a defined value. Based on the template-term count not satisfying the sampling count threshold, the template-term count is incremented.

Abstract: Example embodiments provide a system and method for analyzing conversations and determining whether to participate with a response. A networked system receives, over a network, a communication that is a part of a conversation involving one or more users, whereby the networked system is a participant in the conversation. The networked system analyzes the communication including parsing key terms from the communication. The networked system then identifies a sentiment of a user among the one or more users based on the parsed key terms. Based on the identified sentiment, the networked system determines whether to respond to the communication. In response to a determination to respond, the networked system generates a customized response and transmits the customized response, over the network, to a device of the user. The customized response may comprise questions or a set of options related to the conversation.

Abstract: A natural language learning method, system, and non-transitory computer readable medium include analyzing a corpus of sentences stored in a database to identify an internal structure of words in the corpus of sentences, creating a plurality of new words that are a combination of the internal structure of a word of the words in the corpus of sentences and the word, clustering the plurality of new words created by the creating that match into a plurality of cluster groups, filtering the plurality of cluster groups to create a partial set of each of the plurality of cluster groups, and performing word embedding processing on the partial set of each of the plurality of cluster groups to obtain vectors for new words.

Abstract: A method and system to identify similar names and addresses from given data set comprising plurality of names and addresses. The invention more specifically addresses the challenge faced in Spanish data quality assurance. The name and data is parsed through parsing engine to parse the plurality of Spanish names and addresses. The parsed Spanish names and addresses are sent to a Probable identification engine to identify the probable matches. The combination of name and address matching process can be used for assuring data quality for Spanish names and addresses. The Spanish name matching process consists of identification of probable matches and finding similarity percentages between those probable. Similarly, the Spanish address matching process consists of identification of probable matches (criteria like same city) and finding similarity percentages between those probable. The system includes a parsing engine, a probable identification engine and a match percentage calculation engine.

Abstract: Aspects of the subject disclosure may include, for example, a computer that performs a statistical natural language processing analysis on a plurality of text documents to determine a plurality of topics, creates a proper subset of topics from the plurality of topics, based on user input, maps one or more topics in the proper subset of topics to each document in the plurality of text documents, thereby creating a plurality of topic-document pairs, identifies n-dimensions of bias for each topic-document pair from the text, creates clusters of topics from the proper subset of topics, and generates presentable content depicting each cluster of the clusters of topics according to a corresponding image configuration. The topics and n-dimensions of bias data can be further analyzed with co-collected structured data for statistical relationships.

Abstract: An electronic device is provided, which includes a housing; a microphone located on or within a predetermined distance of a first portion of the housing; a speaker located on or within a predetermined distance of a second portion of the housing; a communication circuit; a processor electrically connected to the microphone, the speaker, and the communication circuit; and a memory electrically connected to the processor configured to store a message to be provided as a voice through the speaker, wherein the memory stores instructions, wherein the processor is configured to execute the instructions to perform operations comprising: determining time information corresponding to a first part of the message if providing of the message is necessary, outputting the message through the speaker, receiving an input sound through the microphone while at least a part of the message is output, and processing the input sound using the time information to detect at least one word or sentence from the input sound, and the pro

Abstract: A media content recommendation method includes: receiving an input entry parameter, and performing word segmentation on the entry parameter to obtain a word sequence; combining adjacent words in the word sequence to generate a mixed word sequence; acquiring phrases included in the word sequence and the mixed word sequence, and separately allocating a corresponding matching weight coefficient to a phrase; searching for media content including the phrase, acquiring a matching weight coefficient sum of phrases included in the found media content, and screening the found media content according to matching weight coefficient sums and performing recommendation. In addition, a media content recommendation apparatus is also provided.

Abstract: A method comprising receiving a dictionary comprising a plurality of entities, wherein each entity has a length of between 1 and n tokens; constructing a probabilistic data representation model comprising n Bloom filter (BF) pairs indexed from 1 to n; populating said probabilistic data representation model with a data representation of said entities, wherein, with respect to each BF pair indexed i: (i) a first BF is populated with the first i tokens of all said entities having at least i+1 tokens, and (ii) a second BF in populated with all said entities having exactly i tokens; receiving a text corpus, wherein said text corpus is segmented into tokens; and automatically matching each token in said text corpus against said populated probabilistic data representation model, wherein said matching comprises sequentially querying each said BF pair in the order of said indexing, to determine a match.

Abstract: The present invention provides a method and device for realizing Chinese character input based on uncertainty information, wherein the method comprises: receiving input information from a user; extracting at least two types of uncertainty information of Chinese characters to be input, from the input information; and, determining the matched Chinese characters according to the at least two types of uncertainty information and outputting the matched Chinese character(s). The device comprises a receiving module, an extracting module and a matching module. The method and device as provided by the present invention allow a user who has incomplete memory of pronunciation or glyph information of Chinese characters to be input to realize correct input of the Chinese characters by defining a certain range for candidate characters corresponding to the Chinese characters to be input, in combination with at least two types of the extracted uncertainty information of the Chinese characters to be input.

Abstract: A method of implementing a controlled vocabulary may include the steps of generating the controlled vocabulary by forming a list of medical terms and associated descriptions; creating a list of codes internal to the controlled vocabulary; and storing the codes, the medical terms, and the descriptions in a format suitable for use in the patient longitudinal electronic medical record environment. In addition, a sound wave may be captured as an electronic file, translated into text, and stored as a patient medical record event. The patient medical record event may be parsed for context and may be tagged with one or more source vocabularies, and a code internal to the source vocabulary may be provided. Event context and source vocabulary findings may then be used as discrete elements to add to the longitudinal patient electronic medical record environment.

Abstract: Provided is a system, method and computer-readable medium for generating data that may be used to train models for a natural language processing application. A system architect creates a plurality of sentence patterns that include entity variables and initiates sentence generation. Each entity is associated with one or more entity data sources. A language generator accepts the sentence patterns as inputs, and references the various entity sources to create a plurality of generated sentences. The generated sentences may be associated with a particular class and therefore used to train one or more statistical classification models and entity extraction models for associated models. The sentence generated process may be initiated and controlled using a user interface displayable on a computing device, the user interface in communication with the language generator module.

Abstract: A device may receive information that identifies a first medical term and may determine whether the first medical term corresponds to a first medical code, which may be a medical code previously input by a first user in association with a second medical term. Or, the first medical code may be included in a dictionary. The device may determine a valid code corresponding to the first medical term based on determining whether the first medical term corresponds to the first medical code. The valid code may be determined based on the first medical code when the first medical term corresponds to the first medical code. The valid code may be determined based on a second medical code, input by a second user, when the first medical term does not correspond to the first medical code. The device may provide information that identifies the valid code.

Abstract: Computer-implemented methods and systems for textual analysis by generating models of the progression of writing and writing abilities are provided. A set of examinee-generated responses for one or more test items may be identified based upon at least one index. One or more data-based profiles may be assembled from the identified set of examinee-generated responses. Writing characteristics may be quantified from the assembled one or more data-based profiles based on the at least one index.

Abstract: A computer system for language modeling may collect training data from one or more information sources, generate a spoken corpus containing text of transcribed speech, and generate a typed corpus containing typed text. The computer system may derive feature vectors from the spoken corpus, analyze the typed corpus to determine feature vectors representing items of typed text, and generate an unspeakable corpus by filtering the typed corpus to remove each item of typed text represented by a feature vector that is within a similarity threshold of a feature vector derived from the spoken corpus. The computer system may derive feature vectors from the unspeakable corpus and train a classifier to perform discriminative data selection for language modeling based on the feature vectors derived from the spoken corpus and the feature vectors derived from the unspeakable corpus.

Abstract: A method for software application optimization using natural language-based queries. The method includes obtaining a user-provided query. The user-provided query includes a constraint to be used for an identification of an application element that matches the constraint, from a set of application elements of a software application. The user-provided query is a string that includes a human language sentence. The method further includes deriving a formalized query from the user-provided query by translating the user-provided query into a syntactic construct of segmented sentence elements and obtaining the application element that matches the constraint. Obtaining the application element that matches the constraint includes deriving a pattern representation of the user-provided query from the formalized query and identifying the application element that matches the pattern representation of the user-provided query from the plurality of application elements.

Abstract: Provided is a keyword presenting system based on a sematic depth structure, including: a communication unit connected with a client via a network; an interface unit configured to receive keyword information inputted from the client and to provide a search interface presenting multiple candidate keywords; a search database configured to save multiple search results and multiple candidate keywords constituting the search interface; a semantic depth database configured to save a semantic depth structure representing a semantic distance between keywords for each candidate keyword; and a semantic depth engine configured to derive multiple associated keywords based on the search database regarding the keyword inputted from the client, and to support a function of gradationally presenting recommended keyword lists for each semantic depth level according to the selection of a user by forming recommended keyword groups for each semantic depth level based on the multiple associated keywords using the semantic depth dat

Abstract: A computer-implemented method for generating explanations for a tax calculation or operation performed by tax preparation software is disclosed. A computing device executes a tax calculation engine in connection with the tax preparation software and operates on a tax calculation graph to perform a tax calculation. The tax calculation graph semantically describes data dependent tax operations comprising functional nodes connected to input nodes by one of a plurality of functions, wherein each tax operation is associated with one or more explanations. An explanation engine is executed to generate a narrative explanation from the one or more explanations associated with one of the tax operations and is presented to the user on the computing device.

Abstract: A method and apparatus for name variant extraction from individual handle identifiers uses a sequential extraction process to construct contextual information. Last name data, first/middle name data, initials, nicknames, and vanity names, along with numerical information indicating dates, may all be captured in extracting information about an individual associated with a particular handle. When multiple possible interpretations result from the analysis, those interpretations are ranked using optimality rules. The resulting data may be used to look up additional information in a consumer database in order to structure a targeted marketing message to the individual associated with the handle.

Abstract: Innovative technologies for reducing network request response times over a server-signed connection are disclosed. The technologies may involve dynamically computing synchronized compression dictionaries using server responses to speculative or “read-ahead” client requests. The technologies operate even when the client is unable to accept the server responses due to server-signing constraints. A server proxy may receive a read-ahead request originating from a client proxy. After receiving a response to the read-ahead request from a server, the server proxy may populate a compression dictionary and forward the read-ahead request to the client proxy. The client proxy may populate its own synchronized compression dictionary using the forwarded read-ahead response. The server proxy and client proxy may use the compression dictionaries to respectively compress and decompress a response to an actual client request that matches or is highly similar to the earlier response to the read-ahead request.

Abstract: A device may obtain a document. The device may identify a skip value for the document. The skip value may relate to a quantity of words or a quantity of characters that are to be skipped in an n-gram. The device may determine one or more skip n-grams using the skip value for the document. A skip n-gram, of the one or more skip n-grams, may include a sequence of one or more words or one or more characters with a set of occurrences in the document. The sequence of one or more words or one or more characters may include a skip value quantity of words or characters within the sequence. The device may extract one or more terms from the document based on the one or more skip n-grams. The device may provide information identifying the one or more terms.

Abstract: Systems and methods for generating a social sketch from social communications are provided. Unlike a typical summary of a subject (or of multiple subjects), a social sketch represents a “snapshot” summary of the social communications of people during a given period of time. The social sketch is generated by obtaining a corpus of social communications and filtering the social communications according to time. The filtered results are clustered according to the subject matter/topics of the social communications. Selected clusters are identified and the topic, representative high-quality social communications from non-experts and experts are extracted from each of the selected clusters and saved as a social sketch corresponding to the time period.

Abstract: An apparatus and a method for preserving privacy in natural language databases are provided. Natural language input may be received. At least one of sanitizing or anonymizing the natural language input may be performed to form a clean output. The clean output may be stored.

Abstract: A keyboard specifically designed to reduce errors in medical coding and aid in the coding process by making it easier to locate only the keys needed to accomplish coding. The keyboard moving the I and O to a different location than the remaining 24 alphabetic characters, and arranges those 24 characters in alphabetic sequence. The keyboard may have a first key group having the alphabetic keys inclusive of all the letters of the alphabet except for the letters I and O, a second key group including the numeric keypad and the letters I and O, and a third key group having only function keys. The keyboard increases coding volume by requiring less time than searching a standard QWERTY keyboard for the necessary keys and consequently helps to cut down on coding errors.

Abstract: A system, method, and computer program product are provided for identifying significant records (e.g. journeys, etc.). In use, a plurality of records including a plurality of events are stored. Further, an indication is received regarding at least one subset of the records. Thereafter, at least a portion of the events of the at least one subset of the records are processed to generate mean information such that the records of the at least one subset may be compared, utilizing the mean information. This may be accomplished while avoiding, at least in part, a covariance calculation, or in such a manner that the comparison is independent of such covariance calculation. A result of the comparison may be displayed.

Abstract: Systems and methods for disambiguation of an ambiguous entity in a search query based on the gaze of a user. These systems and methods may be implemented by a media guidance application (e.g., executed by user equipment associated with the user). In some aspects, the media guidance application may monitor the gaze of the user and attempt to disambiguate the ambiguous entity based on an area of the screen the user viewed while issuing the search query. If the media guidance application receives an indication that it did not disambiguate the ambiguous entity in the search query correctly, the media guidance application may increase the area of the screen that the user viewed in order to find an additional entity. This may allow the media guidance application to quickly and accurately find the correct answer to the user's search query.

Abstract: Methods, apparatus, systems and articles of manufacture to identify a count of n-grams appearing in a corpus are disclosed herein. An example method includes identifying a token that frequently begins a suffix found in the corpus. First suffixes and second suffixes are identified within the corpus, the first suffixes begin with the token and the second suffixes do not begin with the token. A first counting algorithm is performed to identify a first count of n-grams appearing in the first suffixes. A second counting algorithm is performed to identify a second count of n-grams appearing in the second suffixes. The second counting algorithm is different from the first counting algorithm.

Abstract: Disclosed is a method and apparatus for responding to an inquiry from a client via a network. The method and apparatus receive the inquiry from a client via a network. Based on the inquiry, question-answer pairs retrieved from the network are analyzed to determine a response to the inquiry. The QA pairs are not predefined. As a result, the QA pairs have to be analyzed in order to determine whether they are responsive to a particular inquiry. Questions of the QA pairs may be repetitive and similar to one another even for very different subjects, and without additional contextual and meta-level information, are not useful in determining whether their corresponding answer responds to an inquiry.

Abstract: Exemplary embodiments relate to techniques for improving a machine translation system. The machine translation system may include one or more models for generating a translation. The system may generate multiple candidate translations, and may present the candidate translations to different groups of users, such as users of a social network. User engagement with the different candidate translations may be measured, and the system may determine which of the candidate translations was most favored by the users. For example, in the context of a social network, the number of times that the translation is liked or shared, or the number of comments associated with the translation, may be used to determine user engagement with the translation. The models of the machine translation system may be modified to favor the most-favored candidate translation. The translation system may repeat this process to continue to tune the models in a feedback loop.

Abstract: According to an embodiment, a generation device includes a receiver and a generator. The receiver is configured to receive a first model that converts subwords serving as elements of words into the words. The generator is configured to produce, on the basis of the first model, a first finite state transducer that includes a first path having transitions converting one or more subwords into one or more words and a second path, whose first state is the first state of the first path, having cyclic paths to which the subwords are assigned and a transition to which a class classifying a word is assigned.

Abstract: A method for replying to comments is disclosed. The method may comprise obtaining, by a terminal device, a request to reply to a plurality of comments on a published post; determining, by the terminal device, a first target group from the plurality of comments, the first target group including reply targets having similar contents; and sending, by the terminal device, a reply to each of the reply targets.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing word sequences using neural networks. One of the methods includes receiving a first sequence of words arranged according to a first order; and for each word in the first sequence, beginning with a first word in the first order: determining a topic vector that is associated with the word; generating a combined input from the word and the topic vector, and processing the combined input through one or more sequence modeling layers to generate a sequence modeling output for the word; and processing one or more of the sequence modeling outputs through an output layer to generate a neural network output for the first sequence of words.

Abstract: Systems and processes are disclosed for predicting words in a text entry environment. Candidate words and probabilities associated therewith can be determined by combining a word n-gram language model and a character m-gram language model. Based on entered text, candidate word probabilities from the word n-gram language model can be integrated with the corresponding candidate character probabilities from the character m-gram language model. A reduction in entropy can be determined from integrated candidate word probabilities before entry of the most recent character to integrated candidate word probabilities after entry of the most recent character. If the reduction in entropy exceeds a predetermined threshold, candidate words with high integrated probabilities can be displayed or otherwise made available to the user for selection. Otherwise, displaying candidate words can be deferred (e.g., pending receipt of an additional character from the user leading to reduced entropy in the candidate set).

Abstract: Methods and apparatus related to determining reply content for a reply to an electronic communication. Some implementations are directed generally toward analyzing a corpus of electronic communications to determine relationships between one or more original message features of “original” messages of electronic communications and reply content that is included in “reply” messages of those electronic communications. Some implementations are directed generally toward providing reply text to include in a reply to a communication based on determined relationships between one or more message features of the communication and the reply text.

Abstract: Embodiments described herein relate generally to content analysis technologies and natural language processing (NLP). In particular, devices, systems, and methods may implement a reverse sentence reconstruct (RSR) utility, and a sentence vectorization technique (SVT) utility. A computer server may be configured to receive a feeling classification request with text data elements, and in response, generate a feeling classification response indicating feeling for the text data elements using the RSR utility and the SVT utility.

Abstract: Cross-lingual automatic query annotation technique is described, for example, to classify online search queries in Arabic as being of commercial intent, without the need to use human judged Arabic queries. In examples, a query classifier available for a source language (such as English) is used to produce a query classifier for a target language (such as Arabic, German, French). In various examples, a target language query log and target language documents may be used to enable target language and target culture dependent queries to be classified. In various examples a click graph with edges weighted by click frequency is used to infer class membership of unlabeled target language queries from target language documents. In examples the target language documents may be classified using a supervised or semi-supervised classifier. In various examples the automatically labeled target language queries are used to train a target language query classifier for information retrieval and/or advertising.

Abstract: Technology is disclosed to select a preferred machine translation from multiple machine translations of a content item, each machine translation from the multiple machine translations created for the same target language. Each machine translation is assigned a score based on feedback from a user group that receives the machine translation. The machine translation with the highest score is identified as the preferred machine translation, and is provided in response to subsequent requests for translations of the content item. If there is no preferred translation, the several top scoring machine translations are provided to a larger group of users for further scoring. This process may be repeated until either a clearly preferred translation is identified, a maximum number of iterations is reached, or a maximum number of scoring users is reached.

Abstract: Schema-less databases can make data modeling and data management difficult and can detrimentally affect integration with an RDBMS. Inferring a schema from a schema-less database can improve integration by indicating a structure or organization of data in the schema-less database. A schema analyzer can infer a schema by processing data of the schema-less database to identify statistically significant data fields. The schema analyzer then creates a schema that comprises the statistically significant data fields. A data modeler can use the resulting schema along with a schema for a RDBMS to generate a unified data model. A user may submit a query based on the unified data model to obtain results from both databases. The data modeler translates the query from the unified model to be compatible with each of the schemas so that data may be written to or retrieved from each of the schema-less database and the RDBMS.