Abstract: A method for generating subtitles for audiovisual material received and analyses a text file containing dialogue spoken in audiovisual material and provides a signal representative of the text. The text information and audio signal are aligned in time using time alignment speech recognition and the text and timing information are then output to a subtitle file. Colors can be assigned to different speakers or groups of speakers. Subtitles are derived by receiving and analyzing a text file containing dialogue spoken by considering each word in turn and the next information signal, assigning a score to each subtitle in a plurality of different possible subtitle formatting options which lead to that word. The steps are then repeated until all the words in the text information signal have been used and the subtitle formatting option which gives the best overall score is then derived.

Abstract: A system and method for mining data from stored telephone conversations is provided. Users request advanced data processing on the recorded data, either on the live data stream or the data in storage. Processes search the recorded data for keywords and phrases that the user provides the PTR. User can also request more sophisticated analysis of the recorded data for deeper contextual meaning of the conversations. Context information may include identifying the users, the locations and times referred to by the users during the conference, etc. Additional searches related to the obtained information are performed and the extracted information is compared to similar information obtained from previous meetings. Voice inflections and any emotional stress present in the voices of the users can also be detected and added to the collected information. Search terms can also be highlighted in the results.

Abstract: A device and a method for compressing signal information by removing (thinning out) the signal component of a signal in a specific frequency band. Firstly, an input time-series signal (e.g., a PCM signal) is converted by an analyzer (11) into a spectrum signal. Next, of the bands obtained by dividing the spectrum equally into bands, the band having a predetermined or higher correlation in the spectrum distribution with the lower frequency band is specified as a harmonic band by a frequency band masking unit (12). Then, a removal band from which the spectrum is to be removed is determined from the harmonic band, and the spectrum signal of this removal band, from which the spectrum component has been removed (namely the frequency component has been thinned out), is fed to a synthesizer (13).

Abstract: The invention relates to a method and a transcription system (T) for transcribing dictations, in which a dictation file (5) is converted into a text file (8), and subsequently the text file (8) is compared with the dictation file (5). To increase the speed for the subsequent correction, provision is made that during transcription of the dictation file (5) a confidence value is generated for a transcribed text passage of the text file (8), and a comparison of the text file (8) with the dictation file (5) takes place only in respect of those text passages for which the confidence value of the text passage is below a confidence limit, i.e. a text passage recognized as possibly defective is present.

Abstract: The invention relates to pre-processing of a pronunciation dictionary for compression in a data processing device, the pronunciation dictionary comprising at least one entry, the entry comprising a sequence of character units and a sequence of phoneme units. According to one aspect of the invention the sequence of character units and the sequence of phoneme units are aligned using a statistical algorithm. The aligned sequence of character units and aligned sequence of phoneme units are interleaved by inserting each phoneme unit at a predetermined location relative to the corresponding character unit.

Abstract: A method and apparatus are provided for reducing noise in a signal. Under one aspect of the invention, a correction vector is selected based on a noisy feature vector that represents a noisy signal. The selected correction vector incorporates dynamic aspects of pattern signals. The selected correction vector is then added to the noisy feature vector to produce a cleaned feature vector. In other aspects of the invention, a noise value is produced from an estimate of the noise in a noisy signal. The noise value is subtracted from a value representing a portion of the noisy signal to produce a noise-normalized value. The noise-normalized value is used to select a correction value that is added to the noise-normalized value to produce a cleaned noise-normalized value. The noise value is then added to the cleaned noise-normalized value to produce a cleaned value representing a portion of a cleaned signal.

Abstract: A context-free grammar can be represented by a weighted finite-state transducer. This representation can be used to efficiently compile that grammar into a weighted finite-state automaton that accepts the strings allowed by the grammar with the corresponding weights. The rules of a context-free grammar are input. A finite-state automaton is generated from the input rules. Strongly connected components of the finite-state automaton are identified. An automaton is generated for each strongly connected component. A topology that defines a number of states, and that uses active ones of the non-terminal symbols of the context-free grammar as the labels between those states, is defined. The topology is expanded by replacing a transition, and its beginning and end states, with the automaton that includes, as a state, the symbol used as the label on that transition. The topology can be fully expanded or dynamically expanded as required to recognize a particular input string.

Abstract: In one embodiment, the invention provides a method for building a voice response system. The method comprises developing voice content for the voice response system, the voice content including prompts and information to be played to a user; and integrating the voice content with logic to define a voice user-interface that is capable of interacting with the user in a manner of a conversation in which the voice user-interface receives an utterance from the user and presents a selection of the voice content to the user in response to the utterance.

Abstract: A method of presenting a multi-modal help dialog move to a user in a multi-modal dialog system is disclosed. The method comprises presenting an audio portion of the multi-modal help dialog move that explains available ways of user inquiry and presenting a corresponding graphical action performed on a user interface associated with the audio portion. The multi-modal help dialog move is context-sensitive and uses current display information and dialog contextual information to present a multi-modal help move that is currently related to the user. A user request or a problematic dialog detection module may trigger the multi-modal help move.

Abstract: A method is provided for customizing a multi-media message created by a sender for a recipient, in which the multi-media message includes an animated entity audibly presenting speech converted from text by the sender. At least one image is received from the sender. Each of the at least one image is associated with a tag. The sender is presented with options to insert the tag associated with one of the at least one image into the sender text.

Abstract: The display language conversion system of the present invention comprises a first data base 303a, a second data base 303b, an image display processing means 51 and an another-language display processing means 52. The image display processing means 51 displays control data read from the first data base 303a, image data and one language data read from the position information data and the second data base 303b, on a display 3. When the display mode and the display language are specified by the display by the control data displayed on the display 3 and the mouse, and the blowoff frame or the predetermined commentary display region is region specified by the input means, the another-language display processing means 52 displays another language related to the second data base by the region specification in accordance with the display mode.

Abstract: A method, computer readable medium and system are provided which retrieve hint sentences from a sentence database in response to a query. An input component receives the query having terms. A search engine expands the query by including synonyms of the terms to obtain expanded terms. The search engine then combines the expanded terms to form dependency triples from the expanded terms. From the formed dependency triples, dependency triples which are not found in a dependency triples database are discarded to obtain remaining dependency triples from the expanded terms. The search engine then searches the sentence database using the remaining dependency triples as search parameters.

Abstract: A spoken dialog system and method having a dialog management module are disclosed. The dialog management module includes a plurality of dialog motivators for handling various operations during a spoken dialog. The dialog motivators comprise an error-handling, disambiguation, assumption, confirmation, missing information, and continuation. The spoken dialog system uses the assumption dialog motivator in either a-priori or a-posteriori modes. A-priori assumption is based on predefined requirements for the call flow and a-posteriori assumption can work with the confirmation dialog motivator to assume the content of received user input and confirm received user input.

Abstract: A method automatically determines groups of words or phrases that are descriptive names of a small set of documents, as well as infers concepts in the small set of documents that are more general and more specific than the descriptive names, without any prior knowledge of the hierarchy or the concepts, in a language independent manner. The descriptive names and the concepts may not even be explicitly contained in the documents. The primary application of the invention is for searching of the World Wide Web, but the invention is not limited solely to use with the World Wide Web and may be applied to any set of documents. Classes of features are identified in order to promote understanding of a set of documents. Preferably, there are three classes of features. “Self” features or terms describe the cluster as a whole. “Parent” features or terms describe more general concepts. “Child” features or terms describe specializations of the cluster.

Abstract: A method is provided for parsing text in a corpus. The method includes hypothesizing a possible new entry for a dictionary based on a first segment of text. A successful parse is then formed for the first segment of text using the possible new entry. Based on the successful parse, the dictionary is changed to include the new entry. The new entry in the dictionary is then used to parse a second segment of text.

Abstract: A system for understanding entries, such as speech, develops a classifier by employing prior knowledge with which a given corpus of training entries is enlarged threefold. A rule is created for each of the labels employed in the classifyier, and the created rules are applied to the given corpus to create a corpus of attachments by appending a weight of ?p(x), or 1??p(x), to labels of entries that meet, or fail to meet, respectively, conditions of the labels' rules, and to also create a corpus of non-attachments by appending a weight of 1??p(x), or ?p(x), to labels of entries that meet, or fail to meet conditions of the labels' rules.

Abstract: The presence of speech in a filtered speech signal is detected for the purpose of suspending noise level calculations during periods of speech. A received speech signal is split into a plurality of subband signals. A subband variable gain is determined for each subband based on an estimation of the noise level in the received voice signal and on an envelope of the received signal in each subband. Each subband signal is multiplied by the subband variable gain for that subband. The subband signals are combined to produce an output voice signal.

Abstract: A method for responding to an electronic mail message with a limited input device such as a phone includes audibly rendering the question and a set of proposed answers typically provided in the electronic mail message by the sender of the electronic mail message. A language model indicative of the proposed answers is provided to a speech recognizer. The response from the user is obtained and converted to a textual response using the speech recognizer and language model. A second electronic e-mail message is then sent back to the sender. The second electronic mail message includes the textual response.

Abstract: A method for preparing rules used for generating text to describe structured data comprises providing a sample of structured data, and providing a text corresponding to the sample of structured data to a pattern matching learning engine, the text comprising at least one value from within the sample of structured data. The method further comprises organizing the text into a hierarchy of syntactic components, and building rules based on the text and the sample of structured data, wherein the rules define an organization of the hierarchy of syntactic components.

Abstract: A spoken dialog system having a dialog management module is disclosed. The dialog management module includes a plurality of dialog motivators for handling various operations during a spoken dialog. The dialog motivators comprise error-handling, disambiguation, assumption, confirmation, missing information, and continuation. The spoken dialog system uses the assumption dialog motivator in either a-priori or a-posteriori modes. A-priori assumption is based on predefined requirements for the call flow and a-posteriori assumption can work with the confirmation dialog motivator to assume the content of received user input and confirm received user input.