Abstract: Documents needing to be analyzed for various reasons, such as financial crimes, are ranked by examining the topicality and sentiment present in each document for a given subject of interest. In one approach a given document is classified to determine its category, and entity recognition is used to identify the subject of interest. Passages from the document that relate to the entity are grouped and analyzed for sentiment to generate a sentiment score. Documents are then ranked based on the sentiment scores. In another approach, a classification probability score is computed for each passage representing a likelihood that the passage relates to a category of interest, and the document is ranked based on the sentiment scores and the classification probability scores. The category classification uses an ensemble of natural language text classifiers. One of the classifiers is a naïve Bayes classifier with feature vectors generated using Word2Vec modeling.

Abstract: Sentiment analysis is targeted toward a specific subject of interest (or selected subjects) in a passage of natural language text. A dependency tree is generated for the passage, and subtrees are found that have sentiment polarities which contribute to the subject(s) of interest. A targeted sentiment score is computed for the subject(s) of interest based on sentiment expressed in those subtrees. Consecutively occurring nouns in the passage are collapsed into a noun phrase, as are possessives with ensuing nouns. The sentiment expressed in a given subtree can be modified using various linguistic heuristics. For example, sentiment polarity which is modified by a negation word may be inverted, sentiment polarity which is modified by an intensifying word may be increased, or sentiment polarity which is modified by a diluting word may be decreased.

Abstract: Methods and systems for generating cognitive real-time pictorial summary scenes are disclosed. A method includes: obtaining, by a computing device, a document; training, by the computing device, computer models using natural language processing and deep learning based computer vision; and creating, by the computing device, a first pictorial summary scene that summarizes the document using the computer models.

Abstract: Systems and methods for generating search results based on non-linguistic tokens are disclosed. In embodiments, a computer-implemented method comprises: mapping, by a computing device, a non-linguistic token to documents during indexing based on associations in a token database between the non-linguistic token and variations of a multi-word term; receiving, by the computing device, the multi-word term in a search query; determining, by the computing device, the non-linguistic token associated with the multi-word term using the token database; and generating, by the computing device, search results based on the multi-word term and the non-linguistic token.

Abstract: A method and system for analyzing one or more requirements at one or more levels of abstraction of a project is disclosed. The method includes an act of plotting one or more dependency graphs from the one or more requirements at the one or more levels of abstraction and an act of determining vulnerability and influence of the one or more requirements from the one or more dependency graphs. In a further act, one or more vulnerability scores and one or more influence scores are for the one or more requirements are evaluated from the vulnerability and the influence of the one or more requirements. In a further act, one or more volatile scores for the one or more requirements are determined from the one or more vulnerability scores and the one or more influence scores.