International Business Machines Corporation

Globalfoundries Inc.

Tetsuya Nasukawa

Hiroshi Kanayama

Hideo Watanabe

Issei Yoshida

Masayasu Muraoka

Hironori Takeuchi

Shoko Suzuki

Yuta Tsuboi

Raymond Harry Rudy

Shourya Roy

L Venkata Subramaniam

Yohei Ikawa

Miki M Enoki

Ryosuke Kohita

Akiko Murakami

Khan Md Anwarus Salam

Yao-zhong Zhang

Kazuki Sato

Kohichi Kamijoh

Daisuke Takuma

Masaki Ono

Tohru Nagano

Hirofumi Matsuzawa

Jeonghee Yi

Chiaki Oishi

Koichi Takeda

Fusashi Nakamura

Shiho Negishi

Matthew F Hurst

Yuya Unno

Risa Nishiyama

Razvan Constantin Bunescu

Mutsugu Kuboki

Yuki Kurokawa

Takuma Marakami

Diwakar Punjani

Frequent sequences extracted from a set of documents according to a common rule are obtained. Based on comparing occurrence frequencies of various sequences, confidence of the first frequent sequence being a label expression representing a document part in a target document is evaluated. Keywords are extracted from the target document based on evaluation of the confidence.

Text mining based on document structure information extraction

A computer-implemented method is provided for data augmentation. The method includes receiving a set of different base models already pretrained and a set of different test cases. The method further includes collecting a plurality of prediction results of the set of different test cases from the set of different base models. The method also includes identifying a test case as a candidate for the data augmentation based on a number of models in the set of different base models which fail to solve the test case. The method additionally includes augmenting, by a processor device, the identified test case with additional data to form an augmented training dataset. The method further includes retraining at least some of the different base models with the augmented training dataset.

Data augmentation based on failure cases

A method is provided for dictionary expansion. The method acquires an object from a user and adds the object to a set of objects previously acquired from the user that form an expandable dictionary. The method calculates a centroid based on the set. The method calculates a similarity score of each of a plurality of objects relative to the centroid for each of a plurality of object features to calculate a weighted sum of similarity scores for each of the plurality of objects. The method presents candidate objects selected among the plurality of objects based on the weighted sum. The method acquires, from the user, a preferred candidate object among the candidate objects. The method updates weights of the plurality of features to maximize the weighed sum of similarity scores for the preferred candidate object. The method expands the dictionary by adding the preferred candidate object to the expandable dictionary.

User-centric optimization for interactive dictionary expansion

A computer-implemented method is provided for data augmentation. The method includes calculating, by a hardware processor for each of words in a text data, a word replacement probability based on a word occurrence frequency in the text data, wherein the word replacement probability decreases with increasing word occurrence frequency. The method additionally includes selectively replacing at least one of the words in the text data with words predicted therefor by a Bidirectional Neural Network Language Model (BiNNLM) to generate augmented text data, based on the word replacement probability.

Data augmentation by dynamic word replacement

A computer-implemented method for extracting semantic relations is disclosed. In the method, a plurality of hierarchal structures that originates from a corpus of documents is obtained. Each hierarchal structure includes a plurality of elements having respective recitations included in a corresponding document. In the method, for each predetermined relationship between ancestor and descendant elements in the hierarchal structures, a first keyword list is extracted from the ancestor element and a second keyword list is extracted from the descendant element. A statistical index is calculated for each pair of first and second keywords using the first keyword lists and the second keyword lists. The index indicates a strength of association between the first and second keywords. In the method, a candidate list of keyword pairs having semantic relationships is output using the statistical index calculated for each pair.

Extraction of semantic relation

A computer-implemented method is provided. The method includes acquiring a seed word; calculating a similarity score of each of a plurality of words relative to the seed word for each of a plurality of models to calculate a weighted sum of similarity scores for each of the plurality of words; outputting a plurality of candidate words among the plurality of words; acquiring annotations indicating at least one of preferred words and non-preferred words among the plurality of the candidate words; updating weights of the plurality of models in a manner to cause weighted sums of similarity scores for the preferred words to be relatively larger than the weighted sums of the similarity scores for the non-preferred words, based on the annotations; and grouping the plurality of candidate words output based on the weighted sum of similarity scores calculated with updated weights of the plurality of models.

Word grouping using a plurality of models

A computer-implemented method for identifying an expression for a target concept, includes: obtaining a set of texts as a target set of texts, with each text being associated with one of images relevant to a target concept. Candidate expressions for the target concept are extracted from the target set of texts. The candidate expressions are characteristic of the target set of texts. Each image relevant to one of the candidate expressions is collected by using an image search engine. A target expression for the target concept is selected from the candidate expressions based on a comparison result of the target concept and the collected images.

Identifying expressions for target concept with images

A computer-implemented method for finding an asymmetric relation between a plurality of target words is disclosed. The method includes preparing a plurality of image sets, each of which includes one or more images relevant to a corresponding one of the plurality of the target words. The method also includes obtaining a plurality of object labels for each of the plurality of image sets. The method further includes computing a representation for each of the plurality of the target words using the plurality of the object labels obtained for each of the plurality of image sets. The method includes further determining whether there is an asymmetric relation between the plurality of the target words using representations computed for the plurality of the target words.

Finding of asymmetric relation between words

Text analysis includes determining one or more global analysis parameters based on backtranslation of a first corpus between supported languages. A new text analysis model is determined for an unsupported language based on the one or more global analysis parameters and a text analysis model for a first supported language. An input text is analyzed in the unsupported language with the new text analysis model.

Text analysis in unsupported languages using backtranslation

A technique for use in analyzing multidimensional data is disclosed. In the technique, a subset of texts specified by a textual feature is selected from the multidimensional data. Each text of the subset is projected into a target image based on the corresponding spatial information to obtain a spatial distribution map for the textual feature. The similarity between the spatial distribution map for the textual feature and each property distribution map for each predefined property is determined. For the similarity exceeding a threshold, the textual feature is outputted as a notable textual feature.

Method for identifying concepts that cause significant deviations of regional distribution in a large data set

The present invention provides a method for selecting an identified message group from among the plurality of messages sent by a plurality of users of social media. For the messages sent by users other than the original user that quote the message sent by the original message user, the method provides: performing a primary evaluation of the potential for being further resent; identifying a single user who has sent a message for which the primary evaluation result was higher than a predetermined first threshold; identifying a message group similar to a resent messages group, based on a resent messages group that quotes the message send by the single user; performing a secondary evaluation for the potential that the similar message group will be further resent; and selecting a message group for which the secondary evaluation result was higher than a predetermined second threshold.

Method, computer program and computer for detecting trends in social media

A method for identifying synonymous expressions includes determining synonymous expression candidates for a target expression. A plurality of target images related to the target expression and a plurality of candidate images related to each of the synonymous expression candidates are identified. Features extracted from the plurality of target images are compared with features extracted from the plurality of candidate images using a processor to identify a synonymous expression of the target expression.

Discriminating synonymous expressions using images

A method, device, and computer program product for detecting negative opinions in social media, computer program product, and computer. Negative opinions in social media can be precisely detected at an early stage. A method for processing, with a computer, a plurality of messages sent by a plurality of users over time includes the following steps: obtaining a plurality of messages, each including a specific proper noun; determining a politeness level of each of the plurality of messages, each including the specific proper noun; and calculating a proportion of messages having a politeness level lower than a certain threshold with respect to the plurality of messages, each including the specific proper noun.

Method for detecting negative opinions in social media, computer program product and computer

A computer implemented method, system, and product for finding correspondence between terms in two different languages. The method includes the steps of: creating a technical term set and a general term set for each of i) a first language and ii) a second language, creating two bipartite graphs, where each graph corresponds to one of the two languages, and connects the technical term set and general term set of each language, respectively, with weighted links based on corpus information, creating a third bipartite graph by creating weighted links between general terms in the first language and general terms in the second language by using a translation dictionary, creating an association matrix M corresponding to the three bipartite graphs, calculating a similarity matrix Q by calculation of an inverse matrix, and outputting correspondence between the technical term sets of the first and second language on basis of the similarity matrix.

Growing community of inventors

Kanagawa-ken, Japan

Tetsuya Nasukawa