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.

Abstract: Systems and methods for representing and diagnosing interaction sequences in accordance embodiments of the invention are disclosed.

Abstract: Systems and methods for identifying and locating related content using natural language processing are generally disclosed herein. One embodiment includes an HTML5/JavaScript user interface configured to execute scripting commands to perform natural language processing and related content searches, and to provide a dynamic interface that enables both user-interactive and automatic methods of obtaining and displaying related content. The natural language processing may extract one or more context-sensitive key terms of text associated with a set of content. Related content may be located and identified using keyword searches that include the context-sensitive key terms. For example, text associated with video of a first content, such as text originating from subtitles or closed captioning, may be used to perform searches and locate related content such as a video of a second content, or text of a third content.

Abstract: A clausifier and method of extracting clauses for spoken language understanding are disclosed. The method relates to generating a set of clauses from speech utterance text and comprises inserting at least one boundary tag in speech utterance text related to sentence boundaries, inserting at least one edit tag indicating a portion of the speech utterance text to remove, and inserting at least one conjunction tag within the speech utterance text. The result is a set of clauses that may be identified within the speech utterance text according to the inserted at least one boundary tag, at least one edit tag and at least one conjunction tag. The disclosed clausifier comprises a sentence boundary classifier, an edit detector classifier, and a conjunction detector classifier. The clausifier may comprise a single classifier or a plurality of classifiers to perform the steps of identifying sentence boundaries, editing text, and identifying conjunctions within the text.

Abstract: Provided are techniques for determining a sentiment of an electronic message. The electronic message is parsed to identify one or more sub-constructs. For at least one of the sub-constructs that is not false-positive, a sentiment indicator is assigned from a set of types of sentiment indicators, and a score is assigned for the sentiment indicator. A final score is obtained for at least one type of sentiment indicator in the electronic message by summing scores for that type of sentiment indicator. Based on the final score for the at least one type of sentiment indicator, a sentiment of the electronic message is identified.

Abstract: Systems and methods are provided for automatically generating a coherence score for a text using a scoring model. A lexical chain is identified within a text to be scored, where the lexical chain comprises a set of words spaced within the text. A discourse element is identified within the text, where the discourse element comprises a word within the text. A coherence metric is determined based on a relationship between the lexical chain and the discourse element. A coherence score is generated using a scoring model by providing the coherence metric to the scoring model.

Abstract: A structure and method for crowdsourcing includes evaluating a metric related to a content to be translated, determining a priority for the content based on the metric related to the content, and queuing the content for crowdsourcing based on the priority determined from the metric.

Abstract: A method for processing natural language includes generating a first layer of a multi-layer knowledge network including a plurality of word nodes arranged to represent a word or an entity name, generating a second layer of the multi-layer knowledge network with a natural language dataset, the second layer including one or more instance nodes arranged to represent a word or an entity of the natural language dataset, each of the instance nodes being linked by one or more semantic or syntactic relations to form one or more sub-graphs, and, referencing the first layer of the multi-layer knowledge network with the second layer of the multi-layer knowledge network by establishing a reference between each of the word nodes and each of the instance nodes when the word or the entity name represented by each word node is associated with the word or the entity represented by the instance node.

Abstract: Systems, methods, and computer-readable storage media for providing for intelligent switching of languages and/or pronunciations in a text-to-speech system. As the system receives text, the text is analyzed to identify portions which should have speech constructed using a pronunciation distinct from the remaining portions of the text. The text-to-speech system uses multiple pronunciation dictionaries to generate and produce speech corresponding to the text, where the identified portions of the text are in a different language or have a different accent from the remainder of the text. Having generated speech corresponding to the text in multiple languages, accents, or dialects, the system combines the portions, then communicates the speech to the text recipient.

Abstract: Embodiments of the invention relate to ambiguity detection. In one embodiment, an object and a topical domain associated with the object are obtained. In this embodiment, the object includes at least one term. At least one of a plurality of information sources is analyzed based on the at least one term and the topical domain. A determination is made that object is one of ambiguous and unambiguous based on analyzing at least one of the plurality of information sources.

Abstract: An approach to classify different defect records by mapping plain language phrases to a taxonomy. The approach includes a method that includes receiving, by at least one computing device, a defect record associated with a defect. The method further includes receiving, by the least one computing device, a plain language phrase or word. The method further includes mapping, by the least one computing device, the plain language phrase or word to a taxonomy. The method further includes classifying, by the least one computing device, how the defect was at least one of detected and resolved using the taxonomy.

Abstract: Systems, methods, and apparatus for accessing distributed models in automated machine processing, including using large language models in machine translation, speech recognition and other applications.

Abstract: A computer-implemented system and method for developing ethnic and cultural emoticons that are downloadable or uploadable to smart devices or devices, such as laptops, smartphones, and tablet devices, for fast and efficient communications between smart device or other users is disclosed. The computer-implemented system and method also provides for updating cultural or ethnic dictionaries on a periodic basis to reflect the changing nature of language being used by ethnic and cultural groups so that effective communications can be carried out as these changes take place. The computer-implemented system and method include at least a system server connected to the Internet or similar wireless network and one or more databases connected to the system server that will store the ethnic and cultural dictionaries.

Abstract: A translation window is opened in conjunction with a primary window, such as a Web page window containing Web pages hosted on the Internet. The translation window and primary window are automatically adjusted in size and position so that they fit on one user-viewable screen without overlapping. The translation window is linked to a translation dictionary database accessible through the Internet which provides accurate and comprehensive definitions of the words that are identified to be translated.

Abstract: An apparatus for detecting grammatical errors includes a sentence analyzer to break up a sentence into units of morphemes, tag the morphemes with parts of speech, and analyze a syntactic structure of the sentence based on the tagged parts of speech; a first error detector to generate part-of-speech sequences using n-grams of the tagged parts of speech, and detect first grammatical errors by analyzing the generated part-of-speech sequences based on grammatical rules; a second error detector to generate morpheme sequences by binding the morphemes in a preset window (n-window) size, and detect second grammatical errors according to frequencies of appearance of morpheme sequences identical to the generated morpheme sequences by searching examples from an example-based index database (DB); and an integrated error detector to determine final grammatical errors in the sentence by combining the detected first grammatical errors and the detected second grammatical errors.

Abstract: A method and system for recognizing chemical names in a Chinese document. The method includes: receiving a Chinese document including chemical names; recognizing chemical name segments in the document; recognizing non-chemical name segments in the document; and combining the chemical name segments to get chemical names based on the recognized chemical name segments and non-chemical name segments. Specific embodiments of the present invention can effectively recognize chemical names from a chemical document.

Abstract: An alert processing system and method are adapted for processing device alerts. The system includes a routing device in communication with a printer. The routing device receives at least one alert description in a source language transmitted from the printer. The routing device identifies a set of words derived from the alert description related to a condition of the associated device. The routing device compares the set of words, in a target language, to a categorization model and, based on the comparison, categorizes the set of words into to one of a predetermined set of alert categories.

Abstract: A system and method for extracting facts from documents. A fact is extracted from a first document. The attribute and value of the fact extracted from the first document are used as a seed attribute-value pair. A second document containing the seed attribute-value pair is analyzed to determine a contextual pattern used in the second document. The contextual pattern is used to extract other attribute-value pairs from the second document. The extracted attributes and values are stored as facts.

Abstract: Alternative textual interpretations of each sequence of inputs detected within an auto-correcting keyboard region are determined. Actual keystroke contract locations may occur outside the boundaries of specific keyboard key regions associated with the actual characters of word interpretations proposed for selection. The distance from each contact location to each corresponding intended character may increase with the expected frequency of the intended word. An intended word is selected from among generated interpretations and is automatically accepted for output.

Abstract: Implementations of the present disclosure are directed to a method, a system, and a computer program storage device for detecting a language in a text message. A plurality of different language detection tests are performed on a message associated with a user. Each language detection test determines a set of scores representing a likelihood that the message is in one of a plurality of different languages. One or more combinations of the score sets are provided as input to one or more distinct classifiers. Output from each of the classifiers includes a respective indication that the message is in one of the different languages. The language in the message may be identified as being the indicated language from one of the classifiers, based on a confidence score and/or an identified linguistic domain.