Google Inc.

Daniel Gillick

Amarnag Subramanya

Yuan Zhang

Jason Michael Baldridge

David Joseph Weiss

Jason Riesa

Anton Bakalov

A method for identifying codemixed text includes receiving codemixed text and segmenting the codemixed text into a plurality of tokens. Each token includes at least one character and is delineated from any adjacent tokens by a space. For each token of the codemixed text, the method also includes extracting features from the token and predicting a probability distribution over possible languages for the token using a language identifier model configured to receive the extracted features from the token as feature inputs. The method also includes assigning a language to each token of the codemixed text by executing a greedy search on the probability distribution over the possible languages predicted for each respective token.

Identifying codemixed text

A computer-implemented technique can include obtaining a training corpus including pairs of (i) documents and (ii) corresponding abstracts. The technique can include identifying a set of entity mentions in each abstract and each corresponding document based on their respective part-of-speech (POS) tags and dependency parses. The technique can include clustering the sets of entity mentions referring to a same underlying entity to obtain clusters for each document and each corresponding abstract. The technique can include aligning specific abstract entity mentions to corresponding document entity mentions to obtain a set of aligned abstract and document entities. The technique can include labeling the set of aligned entities as salient and unaligned entities as non-salient to generate a labeled corpus. The technique can also include training features of a classifier using the labeled corpus to obtain a trained classifier.

Techniques for automatically identifying salient entities in documents

Growing community of inventors

Oakland, CA, United States of America

Daniel Gillick