International Business Machines Corporation

Massachusetts Institute of Technology

Dan Gutfreund

Ehud Aharoni

Noam Slonim

Ruty Rinott

Ran Levy

Tamar Lavee

Daniel Hershcovich

Yonatan Bilu

Quanfu Fan

Indrajit Bhattacharya

David Carmel

Abhijit Mishra

Amrita Saha

Md Faisal Mahbub Chowdhury

Vikas Raykar

Roy N Bar-Haim

Chen Yanover

Lena Dankin

Tete Xiao

Godbole Shantanu

Francesco Dinuzzo

Bolei Zhou

Sagar Manohar

Abhijit S Mudigonda

Techniques are described for reducing the number of parameters of a deep neural network model. According to one or more embodiments, a device can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a structure extraction component that determines a number of input nodes associated with a fully connected layer of a deep neural network model. The computer executable components can further comprise a transformation component that replaces the fully connected layer with a number of sparsely connected sublayers, wherein the sparsely connected sublayers have fewer connections than the fully connecter layer, and wherein the number of sparsely connected sublayers is determined based on a defined decrease to the number of input nodes.

Restructuring deep neural networks to reduce the number of parameters

An embodiment of the present invention generates natural language content from a set of keywords in accordance with a template. Keyword vectors representing a context for the keywords are generated. The keywords are associated with language tags, while the template includes a series of language tags indicating an arrangement for the generated natural language content. Template vectors are generated from the series of language tags of the template and represent a context for the template. Contributions from the contexts for the keywords and the template are determined based on a comparison of the series of language tags of the template with the associated language tags of the keywords. One or more words for each language tag of the template are generated to produce the natural language content based on combined contributions from the contexts for the keywords and the template.

Natural language text generation from a set of keywords using machine learning and templates

Provided are embodiments including a computer-implemented method for performing recognition. The computer-implemented method includes receiving video data, and performing, at a pre-attention prediction module, a pre-attention prediction for the video data to generate first prediction priors. The computer-implemented method also includes receiving, at a dual attention module, data including the video data and data from the pre-attention prediction to generate attention maps, wherein the attention maps indicate a region of interest of a frame of the video data, wherein the dual attention module generates enhanced feature representations, and performing, at a post-attention prediction module, a post-attention prediction from data from the dual attention module based at least in part on the enhanced feature representation. Also provided are embodiments for a system and a computer program produce for performing recognition.

Context-aware action recognition by dual attention networks

A method comprising using at least one hardware processor for: receiving a topic under consideration (TUC) and content relevant to the TUC; detecting one or more claims relevant to the TUC in the content, based on detection of boundaries of the claims in the content; and outputting a list of said detected one or more claims.

Automatic detection of claims with respect to a topic

A method comprising using at least one hardware processor for receiving a topic under consideration (TUC); providing the TUC as input to a claim function, wherein the claim function is configured to mine at least one content resource, and applying the claim function to the at least one content resource, to extract claims with respect to the TUC; and providing the TUC as input to a classification function, and applying the classification function to one or more claims of the extracted claims, to output corresponding one or more classification tags, wherein each classification tag is associated with its corresponding claim.

Automatic construction of arguments

A method comprising using at least one hardware processor for: receiving a context; identifying evidence with respect to the context in at least one content resource, wherein the identifying comprises: identifying context-free features that generally characterize evidence in the at least one content resource, and identifying context features indicative of the relevance of text segments in the at least one content resource to the context; and outputting a list of said identified evidence.

Context-dependent evidence detection

A method comprising using at least one hardware processor for: identifying relations between pairs of claims of a set of claims; aggregating the claims of the set of claims into a plurality of clusters based on the identified relations; generating a plurality of arguments from the plurality of clusters, wherein each of the arguments is generated from a cluster of the plurality of clusters, and wherein each of the arguments comprises at least one claim of the set of claims, scoring each possible set of a predefined number of arguments of the plurality of arguments, based on a quality of each argument of the predefined number of arguments and on diversity between the predefined number of arguments; and generating a speech, wherein the speech comprises a top scoring possible set of the possible set of the predefined number of arguments.

Automatic generation of a speech by processing raw claims to a set of arguments

A method comprising using at least one hardware processor for: receiving (a) a proposition and (b) a plurality of claims; identifying a local claim polarity of each claim of the plurality of claims with respect to the proposition; calculating a pairwise claim polarity agreement score for each pair of claims of the pairs of claims reflecting the likelihood of said each pair of claims to have the same claim polarity, wherein the pairwise claim polarity agreement score is associated with each claim of the pair of claims; and determining a global claim polarity for each claim of the plurality of claims based on the local claim polarity of the claim and pairwise claim polarity agreement scores associated with said each claim.

Claim polarity identification

Growing community of inventors

Raanana, Israel

Dan Gutfreund