Abstract: Disclosed RNN-implemented methods and systems for abstractive text summarization process input token embeddings of a document through an encoder that produces encoder hidden states; applies the decoder hidden state to encoder hidden states to produce encoder attention scores for encoder hidden states; generates encoder temporal scores for the encoder hidden states by exponentially normalizing a particular encoder hidden state's encoder attention score over its previous encoder attention scores; generates normalized encoder temporal scores by unity normalizing the temporal scores; produces the intra-temporal encoder attention vector; applies the decoder hidden state to each of previous decoder hidden states to produce decoder attention scores for each of the previous decoder hidden states; generates normalized decoder attention scores for previous decoder hidden states by exponentially normalizing each of the decoder attention scores; identifies previously predicted output tokens; produces the intra-decoder at

Abstract: The present disclosure generally relates to handwriting on touch sensitive surfaces. In some examples, text suggestions strokes entered on a touch sensitive surface are viewed and selected in response to a rotatable input mechanism. In some examples, text determined from a set of strokes on the touch sensitive surface is revised based on a subsequently entered stroke on the touch sensitive surface. In some examples, a determination is made whether to include a stroke in a set of strokes based a time between the stroke and the previous stroke. In some examples, determining text based on a set of stroke is interrupted to determined revised text based on the set of strokes and a second stroke.

Abstract: A system and method that translates sentences of natural language text into sets of axioms of formal logic that are consistent with parses resulting from NLP and acquired constraints as they accumulate. The system and method further present these axioms so as to facilitate further disambiguation of such sentences and produces axioms of formal logic suitable for processing by automated reasoning technologies, such as first-order or description logic suitable for processing by various reasoning algorithms, such as logic programs, inference engines, theorem provers, and rule-based systems.

Abstract: Various aspects of the subject technology relate to systems, methods, and machine-readable media for automated quantitative assessment of text complexity. A system may include processing at least one body of text in a text-based query using a natural language processing engine. The processed text may include sub-blocks of text in a predetermined sequence size such as an n-gram. The system may compare reference bases to the processed text, where each reference base is associated with a different natural language. The system determines which of the reference bases has a highest number of matching words within the body of text, and thereby identifies the reference base as the source language of the supplied text. The system then determines an average complexity score for n-gram using a quantitative assessment engine. The system then applies a readability score to the body of text based on the average complexity scores of the n-grams.

Abstract: A targeted summary of textual content tuned to a target audience vocabulary is generated in a digital medium environment. A word generation model obtains textual content, and generates a targeted summary of the textual content. During the generation of the targeted summary, the words of the targeted summary generated by the word generation model are tuned to the target audience vocabulary using a linguistic preference model. The linguistic preference model is trained, using machine learning on target audience training data corresponding to a corpus of text of the target audience vocabulary, to learn word preferences of the target audience vocabulary between similar words (e.g., synonyms). After each word is generated using the word generation model and the linguistic preference model, feedback regarding the generated word is provided back to the word generation model. The feedback is utilized by the word generation model to generate subsequent words of the summary.

Abstract: Apparatuses, methods and storage medium associated with real time closed captioning or highlighting are disclosed herein. In embodiments, an apparatus may include an audio interceptor to intercept an audio portion of an audio/video output stream of a multi-media application; a speech recognizer coupled to the audio interceptor to, in real time, process the audio portion of the audio/video stream, recognize speech within the audio portion, and automatically generate text corresponding to recognized speech, or keywords within the recognized speech; and a controller coupled to the audio interceptor and the speech recognizer, to control operations of the audio interceptor and the speech recognizer. The automatically generated text or keywords may be outputted as closed captions or highlights to complement video content of the audio/video stream. Other embodiments may be disclosed or claimed.

Abstract: This string search method for a computer that searches documents for a specified string includes the following steps: a first step in which the specified string is divided into a plurality of characters or strings, including a first character or string and a second character or string; a second step in which a plurality of sets are created by grouping a plurality of documents in units; a third step in which a determination is made as to whether or not the first character or string appears in that set; a fourth step in which a determination is made as to whether or not the second character or string appears in that set; and a sixth step in which, if first appearance information and second appearance information were both added to one of the plurality of sets, namely a first set, a detailed search is performed on said first set.

Abstract: Systems and processes for operating an intelligent automated assistant are provided. In accordance with one example, a method includes, at an electronic device with one or more processors and memory, receiving a natural-language speech input indicative of a request to the digital assistant; obtaining, by the digital assistant, context information; determining, by the digital assistant, a text-to-speech mode from a plurality of text-to-speech modes based on the obtained context information; and providing, by the digital assistant, an audio output with the determined text-to-speech mode, where the audio output is indicative of a speech response to the user request.

Abstract: Dynamic semantic processing of text in a word processing application with engagement of question-answering system. A user provides a text stream to a computer system via an input source. The input text stream includes a first natural language statement. The system determines that the first natural language statement includes a fact-based component expressed in natural language form. The system identifies an initial span and an alternative span of the first natural language component, based on the determining. The system engages a question-answering (QA) system by providing the initial span and the alternative span of the first natural language component to the QA system, and by receiving, in response to the providing, a set of natural language results from the QA system. The system evaluates the initial span and the alternative span of the first natural language component based on the received natural language results.

Abstract: According to one exemplary embodiment, a method for generating an answer in a question answering system is provided. The method may include receiving a question. The method may also include identifying a candidate answer from a corpus. The method may then include determining a plurality of sentences based on the identified candidate answer. The method may further include calculating a similarity value for each sentence within the plurality of sentences based on comparing the plurality of sentences to the candidate answer and the received question. The method may also include identifying at least one sentence within the plurality of sentences with a calculated similarity value that exceeds a threshold value. The method may then include generating the answer based on the candidate answer and the identified at least one sentence.

Abstract: Disclosed RNN-implemented methods and systems for abstractive text summarization process input token embeddings of a document through an encoder that produces encoder hidden states; applies the decoder hidden state to encoder hidden states to produce encoder attention scores for encoder hidden states; generates encoder temporal scores for the encoder hidden states by exponentially normalizing a particular encoder hidden state's encoder attention score over its previous encoder attention scores; generates normalized encoder temporal scores by unity normalizing the temporal scores; produces the intra-temporal encoder attention vector; applies the decoder hidden state to each of previous decoder hidden states to produce decoder attention scores for each of the previous decoder hidden states; generates normalized decoder attention scores for previous decoder hidden states by exponentially normalizing each of the decoder attention scores; identifies previously predicted output tokens; produces the intra-decoder at

Abstract: Systems and methods of the present invention provide for one or more server computers communicatively coupled to a network and configured to: receive a string used to search an available domain name, identify a token within the string, and select an acronym or full string token in an acronym database matching the token. If the token matches acronym or full string tokens in the acronym database, the server generates and transmits one or more available domain name candidates with the acronym or full string tokens to the user's client computer.

Abstract: The present disclosure provides language-based mechanisms for generating acronyms from text input. The language of the text input may be provided or automatically detected. The target acronym length may indicate a maximum length and may vary depending on the input language. The text input may be separated into tokens and organized as a token tree list. Based on the tokens, an acronym may be generated from the available capital words. If there are not enough capital words, all words (e.g., both capitalized and lowercase words) may be used to generate the acronym. If there are not enough words, then all words and segments may be used to generate the acronym. Finally, a background color may be generated based characteristics relating to the text input or the generated acronym. The acronym and background color may be used to create a graphic, such as an icon or thumbnail, for a graphic user interface.

Abstract: In an example embodiment, a method for indexing data is provided. Primary data and auxiliary data is received from a first tenant in a computer network. A first index is then created for the primary data and the auxiliary data from the first tenant. The first index is stored in a first shard. Then primary data and auxiliary data are received from a second tenant in the computer network. A second index is created for the primary data and the auxiliary data from the second tenant. The second index is then stored in the first shard. The first shard is then stored in a first instance of a distributed database, the distributed database comprising a plurality of instances, each instance operating on a different logical or physical device.

Abstract: An approach is provided that receives a word that belongs to a first natural language and retrieves a first set of complexity data pertaining to the word in the first natural language. The approach translates the word to one or more translated words, with each of the translated words corresponding to one or more second natural languages. The approach then retrieves sets of complexity data, with the sets of complexity data corresponding to a different translated word. The approach determines a complexity of the word in the first natural language based on an analysis of the first and second sets of complexity data.

Abstract: An approach is provided that returns a simplified set of text to a user of a natural language processing (NLP) system with the simplified set of text having a complexity appropriate to the reading level of the user. The approach receives a word that belongs to a first natural language and retrieves a first set of complexity data pertaining to the word in the first natural language. The approach translates the word to one or more translated words, with each of the translated words corresponding to one or more second natural languages. The approach then retrieves sets of complexity data, with the sets of complexity data corresponding to a different translated word. The approach determines a complexity of the word in the first natural language based on an analysis of the first and second sets of complexity data.

Abstract: In some embodiments, a device displays functionality information in response to receiving an indication of a first input for which a contact meets functionality display criteria. In some embodiments, a device generates a user interface that includes a navigation bar including images from different positions in a respective content item, and a representation of an adjacent content item. In some embodiments, a device moves a selection-indicator in a user interface by a predefined amount in response to receiving an indication of a first input that meets unitary movement criteria. In some embodiments, a device interprets movement of a contact of an input based at least in part on a grip of a user. In some embodiments, a device displays a plurality of character selection options when a text entry field is not tolerant of character ambiguity and a first input corresponds to a plurality of candidate characters.

Abstract: A computer-implemented method includes receiving a search request at a central information provider, generating a query containing media-related terms relating to the search request, and submitting the query to a search engine having access to media-related search results, wherein the query is broader than the received search request, and transmitting media-related search results generated in response to the query by the search engine, to a remote device.

Abstract: The systems and methods are directed towards embodiments for incorporating alternative sources of information, other than current online sources, for use with intelligent personal assistants. These alternative sources are processed and prioritized by the intelligent personal assistant when performing queries by the user. Furthermore, the intelligent personal assistant is capable of performing other actions related to the alternative source of information. In another embodiment, the user is capable of providing both audio commands and motion-based commands to the intelligent personal assistant to instruct the intelligent personal assistant to perform specific processes whereby the audio commands correspond to a first set of processes and the motion-based commands correspond to a distinct second set of processes.

Abstract: A method for context-aware translation is provided. The present invention may include receiving a program integrated information file (PII) associated with a graphical user interface object of with a software product. The present invention may also include indexing each element string within the PII file. The present invention may further include generating a base language general availability build of the graphical user interface object using the PII file. The present invention may also include retrieving a document object model (DOM) of each element within the base language general availability build. The present invention may further include creating a string relation table for the base language general availability build using the DOM for each element and each indexed element string within the PII file. The present invention may also include generating a string relation translation memory table for at least one language.

Abstract: A content management system (CMS) and a translation management system (TMS) can utilize content dimensions for content items to manage and translate the content items between languages. Machine and human translations of complex dynamic content can also be improved by pre-rendering the content to remove localization-related syntax prior to machine or human translation. Content items can also be scored as to their suitability for localization prior to translation, and translation can be skipped for content items that do not have a sufficiently high score. Semantic and natural language processing (NLP) techniques can also be utilized for content categorization and routing. Translations of content items can also be continuously refined and higher quality re-translated content can be provided in an automated fashion.

Abstract: In an aspect, a computerized method for generating processed files of deposition testimony transcript designations may include accessing a file containing designations of contents of a textual transcript, quarantining errors in the designations, and generating a processed file containing processed designations of contents of the textual transcript having quarantined errors removed therefrom. In another aspect, a computerized method of generating designations for a deposition testimony transcript may include accessing designation information regarding designations made with respect to text of the deposition testimony transcript, accessing rules for generating designations based on the designation information, and generating the designations based on the rules.

Abstract: Methods, systems, and computer readable medium for providing a translated message in telecommunications. A request is received for translating a current message in a first language into a current message in a second language destined for a user. Whether the current message in the first language has been previously translated into a corresponding message in the second language is determined. If the corresponding message in the second language exists, the current message in the second language is generated based on the corresponding message in the second language previously translated. The current message is then sent in the second language to the user as a response to the request.

Abstract: A cascaded finite-state-transducer array includes a plurality of finite-state-transducers, the finite-state-transducers being distributed in space. The finite-state-transducer array is configured with dedicated data transfer channels between the finite-state-transducers to transfer specific data types. Each data stream on a dedicated data transfer channel may transmit a particular data type, which may be sorted in increasing order of start offsets or token IDs.

Abstract: Embodiments of the present invention address deficiencies of the art in respect to collaborative information object management and provide a method, system and computer program product for annotating collaborative information structures. In an embodiment of the invention, a method for annotating collaborative information structures can be provided. The method can include creating a collaborative information structure document (ISD) with each of an object section and an annotation section, adding a collaborative object in a collaborative computing environment to the object section of the collaborative ISD, adding an annotation for the collaborative ISD to the annotation section of the collaborative ISD, and storing the collaborative ISD for use as a collaborative object in the collaborative computing environment.

Abstract: Methods, systems, and computer readable medium for providing translated web content with translated URLs. A request is received content in a second language translated from content in a first language. The content in the first language is obtained from the first Internet source. One or more links are identified from the content in the first language. Each original URL is divided into one or more translatable URL components in the first language. Whether the one or more translatable URL components in each original URL have been previously translated into the second language and stored as translated URL components is determined. If there is at least one translatable URL component previously translated and stored as a corresponding translated URL component, a translated URL in the second language for each original URL is generated. The content in the second language with translated URLs is created.

Abstract: Technology is described for refining a language model for a language recognition system based on aggregating and analyzing word tag metadata from multiple users of the language. The technology allows a user to mark a word or phrase in a selected language (e.g., as offensive or misspelled, or as a part of speech or other category), combines information collected from multiple users of the selected language, and updates the user's language model based on the combined information from multiple users of the selected language.

Abstract: A method includes accessing information identifying multiple files and identifying classification data for the multiple files, where the classification data indicates, for a particular file of the multiple files, whether the particular file includes malware. The method also includes generating a sequence of entropy indicators for each of the multiple files, each entropy indicator of the sequence of entropy indicators for the particular file corresponding to a chunk of the particular file. The method further includes generating n-gram vectors for the multiple files, where the n-gram vector for the particular file indicates occurrences of groups of entropy indicators in the sequence of entropy indicators for the particular file. The method also includes generating and storing a file classifier using the n-gram vectors and the classification data as supervised training data.

Abstract: Systems and methods provide a media guidance application that recognizes a plurality of natural language search queries for identifying a set of search results. For example, a user may want to determine when the Yankees are playing their next baseball game. The user may structure their query in multiple ways, such as, “When are the Yankees playing?” “What time is the Yankees game?” “When is the next Yankees baseball game?” The user would expect the same result, a description of when the Yankees are playing, regardless of how the query is structured. The systems and methods enable a user to use a plurality of search queries when searching for items or information to get desired results.

Abstract: An information processing apparatus includes a memory, a receiver, and circuitry. The memory stores, for each one of a plurality of character strings, display number information indicating a total number of times the character string has been displayed at a user terminal, the user terminal being operated by a user. The receiver receives, from the user terminal, information on one of the plurality of character strings that is displayed at the user terminal. The circuitry updates the display number information for the displayed character string, based on the information on the displayed character string that is received by the receiver, determines whether the displayed character string is to be translated, based on the updated display number information, and sends a request to manually translate the displayed character string, based on determination indicating that the displayed character string is to be translated.

Abstract: Various technologies described herein pertain to training and utilizing a general, statistical framework for modeling translation via Markov random fields (MRFs). An MRF-based translation model can be employed in a statistical machine translation (SMT) system. The MRF-based translation model allows for arbitrary features extracted from a phrase pair to be incorporated as evidence. The parameters of the model are estimated using a large-scale discriminative training approach based on stochastic gradient ascent and an N-best list based expected Bilingual Evaluation Understudy (BLEU) as an objective function.

Abstract: In one example embodiment, a server may divide documents into a plurality of morphological segments, parse, from the plurality of morphological segments, a plurality of named biological entities, tag the plurality of named biological entities and the plurality of morphological segments, generate a database that includes the tagged named biological entities and the tagged morphological segments, receive a first search term including one or more of the tagged named biological entities or one or more of the tagged morphological segments, search the database for at least one result phrase that includes at least one occurrence of the first search term, receive an input to retrieve a second search term including at least one morphological segment from one of the result phrases, store the second search term as a mining search term, and mine for the one or more documents based on the mining search term.

Abstract: In an aspect, a computerized method for generating processed files of deposition testimony transcript designations may include accessing a file containing designations of contents of a textual transcript, quarantining errors in the designations, and generating a processed file containing processed designations of contents of the textual transcript having quarantined errors removed therefrom. In another aspect, a computerized method of generating designations for a deposition testimony transcript may include accessing designation information regarding designations made with respect to text of the deposition testimony transcript, accessing rules for generating designations based on the designation information, and generating the designations based on the rules.

Abstract: A computer-implemented method for detecting character encodings of text streams may include 1) identifying a request to identify a character encoding of a text stream, 2) dividing the text stream to identify a plurality of n-grams within the text stream, 3) identifying a plurality of vectors within a vector space, each vector representing a potential character encoding by a plurality of expected n-grams within the potential character encoding, 4) generating, based on the plurality of n-grams within the text stream and on an inverse frequency of appearance of each of the plurality of n-grams within the plurality of vectors, a representative vector describing the text stream, and 5) identifying the character encoding of the text stream based on identifying a similarity between the representative vector and a neighboring vector within the plurality of vectors representing the character encoding. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A method for context-aware translation is provided. The present invention may include receiving a program integrated information file (PII) associated with a graphical user interface object of with a software product. The present invention may also include indexing each element string within the PII file. The present invention may further include generating a base language general availability build of the graphical user interface object using the PII file. The present invention may also include retrieving a document object model (DOM) of each element within the base language general availability build. The present invention may further include creating a string relation table for the base language general availability build using the DOM for each element and each indexed element string within the PII file. The present invention may also include generating a string relation translation memory table for at least one 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: In an approach for evaluating performance of machine translation, a processor receives a first document in a source language. A processor translates the first document in the source language to a second document in a target language, based, at least in part, on a first quantity of information. A processor evaluates the second document in the target language, based, at least, on one or more aspects of the translation. A processor determines, based, at least in part, on the evaluation, the second document in the target language meets a predetermined threshold.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for storing, in an index associated with a document, a particular term that occurs in the document, wherein the particular term comprises n words, and wherein n is greater than 1; identifying a substitute term of the particular term; and in response to identifying the substitute term of the particular term, storing, in the index associated with the document, (i) the substitute term of the particular term, and (ii) data indicating that the substitute term spans the n words of the particular term.

Abstract: Chinese name transliteration is described. A Chinese name including a surname followed by a given name is received. The received Chinese name is parsed into a surname and a given name by: identifying any dual-character surnames in the received Chinese name, by referencing a list of known Chinese surnames; identifying any single-character surnames in the received Chinese name, by referencing the list of known Chinese surnames; and identifying, among the remaining characters in the received Chinese name, any characters not belonging to a dual-character or a single-character surname as a given name. The identified surname and the identified given name are transliterated into a Mandarin Pinyin representation. The Mandarin Pinyin representation of the surname and the given name is normalized into a pre-determined representation.

Abstract: Chinese name transliteration is described. A Chinese name including a surname followed by a given name is received. The received Chinese name is parsed into a surname and a given name by: identifying any dual-character surnames in the received Chinese name, by referencing a list of known Chinese surnames; identifying any single-character surnames in the received Chinese name, by referencing the list of known Chinese surnames; and identifying, among the remaining characters in the received Chinese name, any characters not belonging to a dual-character or a single-character surname as a given name. The identified surname and the identified given name are transliterated into a Mandarin Pinyin representation. The Mandarin Pinyin representation of the surname and the given name is normalized into a pre-determined representation.

Abstract: Methods and arrangements for automatic ground truth generation of medical image collections. Aspects include receiving a plurality of imaging studies, wherein each imaging study includes one or more images and a textual report associated with the one or more images. Aspects also include selecting a key image from each of the one or more images from each of the plurality of imaging studies and extracting one or more discriminating image features from a region of interest within the key image. Aspects further include processing the textual report associated with the one or more images to detect one or more concept labels, assigning an initial label from the one or more concept labels to the one or more discriminating image features, and learning an association between each of the one or more discriminating image features and the one or more concept labels.

Abstract: A regular rule learning method, system, and non-transitory computer readable medium, include an analyzing circuit configured to analyze a corpus of sentences stored in a database to discover lexical features and conjunctively create a regular set of rules based on the discovered lexical features and syntactical features.

Abstract: In one embodiment, a satellite microphone assembly for use in teleconferencing or other audio based communications comprises a base housing a microphone and volume control electronics. The base is enclosed by a cylinder, rotatable about the base and having a top surface configured to be an actuatable button to toggle a mute/unmute status of the microphone. The cylinder has a sidewall configured to be engaged by a user such that the user can rotate the cylinder. An optical sensor is supported by the base, and is configured to detect a rotation of the cylinder and to output information about a direction and a degree of rotation of the cylinder to the volume control electronics, causing a rotation of the cylinder to affect the volume level of a speaker.

Abstract: Approaches for translating a transliterated search query are provided. An approach includes receiving a search query containing a transliterated word. The approach also includes determining a source language corresponding to the transliterated word. The approach further includes converting the transliterated word to a word in the source language. The approach additionally includes translating the word in the source language to a word in a target language. The approach also includes performing a search using the word in the target language.

Abstract: A system includes a processor and a non-transitory computer readable medium storing instructions translatable by the processor. The instructions when translated by the processor cause the system to determine an exemplar reference which identifies a managed object stored in a repository residing in an enterprise computing environment. The system can determine a translation group utilizing the exemplar reference. The managed object may represent a content item. The translation group may include the content item and the content item may be multilingual enabled. The system can determine an effective locale for a request for content received from a client device communicatively connected to a web server in the enterprise computing environment, determine a language that is appropriate for the effective locale, and respond to the request for content with the content item or a translation of the content item that is in the language appropriate for the effective locale.

Abstract: The present invention includes a method of providing users with formatted access to XML and similar documents received from trading partners using a general purpose viewer or browser. Particular aspects of the present invention are described in the claims, specification and drawings.

Abstract: A object selector for supporting at least one selectable view is described. The object selector includes: a selectable view accessor configured for accessing a set of selectable views, wherein each selectable view of the set of selectable views is configured for displaying data associated with at least one selected item; an integrated selectable view renderer configured for rendering at least one integrated selectable view by integrating the at least one selected item into at least one selectable view of the set of selectable views; and an integrated selectable view displayer configured for displaying the at least one integrated selectable view.

Abstract: Mechanisms are provided for implementing a natural language content (NLC) disambiguation engine. The NLC disambiguation engine operates to identify, in a corpus of natural language content, a portion of natural language content (NLC) that is determined to be ambiguous with regard to the context of the portion of NLC. The NLC disambiguation engine compares the ambiguous content portion to private content information associated with a source of the ambiguous content portion. The NLC disambiguation engine identifies a domain of the ambiguous content portion based on a domain of a matching portion of the private content information. The NLC disambiguation engine performs a clarifying operation that clarifies the ambiguous content portion based on the identified domain of the matching portion of the private content information to thereby generate a clarified content portion for processing by a cognitive operation of a cognitive system.

Abstract: Methods, systems, and devices are described for deriving an identifier encoded in a visible light communication (VLC) signal. One method includes capturing a first part of the VLC signal; extracting, from the first part of the VLC signal, a first pattern of bits representing at least a portion of the identifier encoded in the VLC signal; comparing the first pattern of bits to different portions of a plurality of identifiers; and identifying, based at least in part on the comparing, a subset of the plurality of identifiers as candidate matches to the identifier encoded in the VLC signal.

Abstract: Disclosed herein are systems, methods, and non-transitory computer-readable storage media for learning latent representations for natural language tasks. A system configured to practice the method analyzes, for a first natural language processing task, a first natural language corpus to generate a latent representation for words in the first corpus. Then the system analyzes, for a second natural language processing task, a second natural language corpus having a target word, and predicts a label for the target word based on the latent representation. In one variation, the target word is one or more word such as a rare word and/or a word not encountered in the first natural language corpus. The system can optionally assigning the label to the target word. The system can operate according to a connectionist model that includes a learnable linear mapping that maps each word in the first corpus to a low dimensional latent space.