University of Southern California

Panayiotis Georgiou

Shrikanth Sambasivan Narayanan

Dagen Wang

Murtaza Bulut

Abhinav Sethy

A translation between a source language and a target language is disclosed. The source language items are divided, with primary and secondary source language items or named entities being identified, where the primary and secondary source languages being different from each other and from the target language. The entities in the second source language are translated in a different way. For example, they may be copied into the target language, or translated in a special way that is based on their meaning, e.g, into a term that has a more descriptive meaning in the target language.

Communication system using mixed translating while in multilingual communication

A socially-cognizant translation system that takes social state between speaker and listener into account when making the translation. The translation may be more formal or less formal based on the social relationship between speaker and listener. Profanities in the source speech may be detected, and used to determine the social cognizance. Also, the source speech can be translated without the profanity in the target language, instead using the meaning of the profanity.

Socially cognizant translation by detecting and transforming elements of politeness and respect

Spoken translation system which detects both speech from the information and also detects meta information streams from the information. A first aspect produces an enriched training corpus of information for use in the machine translation. A second aspect uses two different extraction techniques, and combines them by lattice rescoring.

Spoken translation system using meta information strings

Forming and/or improving a language model based on data from a large collection of documents, such as web data. The collection of documents is queried using queries that are formed from the language model. The language model is subsequently improved using the information thus obtained. The improvement is used to improve the query. As data is received from the collection of documents, it is compared to a rejection model, that models what rejected documents typically look like. Any document that meets the test is then rejected. The documents that remain are characterized to determine whether they add information to the language model, whether they are relevant, and whether they should be independently rejected. Rejected documents are used to update the rejection model; accepted documents are used to update the language model. Each iteration improves the language model, and the documents may be analyzed again using the improved language model.

Growing community of inventors

La Crescenta, CA, United States of America

Panayiotis Georgiou