Abstract: An authoring tool (or process) to facilitate the performance of an annotation function and an indexing function. The annotation function may generate informational annotations and word annotations to a database design schema (e.g., an entity-relationship diagram or “ERD”). The indexing function may analyze the words of the annotations by classifying the words in accordance with a concordance and dictionary, and assign a normalized weight to each word of each of the annotations based on the classification(s) of the word(s) of the annotation. A query translator (or query translation process) to (i) accept a natural language query from a user interface process, (ii) convert the natural language query to a formal command query (e.g., an SQL query) using the indexed annotations generated by the authoring tool and the database design schema, and (iii) present the formal command query to a database management process for interrogating the relational database.

Abstract: A method is proposed for processing software or Website code associated with a primary locale by an automatic, or semi-automatic, parsing process. The method separates the code into a file of international code which is not locale dependent, and a resource pack of items specific to the primary locale. The international code and the resource pack co-operate to perform the function of the original code. The resource pack can be converted for any number of other locales. Each converted resource pack together with the international code has the effect of the original code adapted for another locale.

Abstract: A method reduces the number of diacritics and other intermediate symbols occurring between two factors that result from any factorization such as extraction of infinite ambiguity, factorization of finitely ambiguous finite-state transducer, or bimachine factorization. The method a posteriori removes all redundant intermediate symbols. The method can be used with any two finite-state transducers (FSTs) that operate in a cascade. With longer cascades, the method can be applied pair-wise to all FSTs, preferably starting from the last pair.

Abstract: A homophonic neologisms generator can include a dictionary table (10) having one or more entries including an orthography and an associated pronunciation including one or more phonemes; and a weightings table (14) having one or more entries specifying a cluster including one or more letters, a cluster pronunciation including one or more phonemes, and a weighting for the pronunciation of the cluster. A user interface can receive a word for which neologisms can be generated. A clustering mechanism can divide the pronunciation into a plurality of phonemes having one or more orthographic representations. Each orthographic representation can include one or more graphemes. Orthographic representations of the pronunciation can be ordered according to the associated weightings of the cluster graphemes in the weightings table and the dictionary can be searched to check that a generated well-formed orthography does not exist.

Abstract: A method factors an ambiguous finite state transducer (FST) into two finite state transducers. The first FST is functional (i.e., unambiguous). The second FST retains the ambiguity of the original FST but is fail-safe (i.e., no failing paths) when applied to the output of the first FST. That is, the application of the second FST to an input string never leads to a state that does not provide a transition for the next symbol in the input. Subsequently, the first FST can be factorized into a left-sequential FST and a right-sequential FST that jointly represent a bi-machine.

Abstract: A method factors an input finite state transducer (FST) with unknown symbols into a left-sequential FST and a right-sequential FST while avoiding direct factorization of the unknown symbols. The left-sequential FST is formed by replacing each occurrence of the unknown symbol in the input FST with a sequence of the unknown symbol and a diacritic. The right-sequential FST is formed by replacing each occurrence of the diacritic with a symbol representative of an empty string and an output symbol.

Abstract: A method and apparatus estimate additive noise in a noisy signal using an iterative technique within a recursive framework. In particular, the noisy signal is divided into frames and the noise in each frame is determined based on the noise in another frame and the noise determined in a previous iteration for the current frame. In one particular embodiment, the noise found in a previous iteration for a frame is used to define an expansion point for a Taylor series approximation that is used to estimate the noise in the current frame.

Abstract: A suppression system of background noise of speech signals uses an adaptive filter of long-time and short-time statistical characteristics of the speech signals. Since the statistical characteristics of the speech signals vary with time, the associated coefficents of the filter also have to be adjusted according to the varitation of the speech signals to eliminate the unnecessary background noise. High frequency attenuation of the speech signals is compensated for by passing the signal through a high frequency booster to elevate the degree of brightness of the speech signals and to improve their quality.

Abstract: An improved speech model and methods for estimating the model parameters, synthesizing speech from the parameters, and quantizing the parameters are disclosed. The improved speech model allows a time and frequency dependent mixture of quasi-periodic, noise-like, and pulse-like signals. For pulsed parameter estimation, an error criterion with reduced sensitivity to time shifts is used to reduce computation and improve performance. Pulsed parameter estimation performance is further improved using the estimated voiced strength parameter to reduce the weighting of frequency bands which are strongly voiced when estimating the pulsed parameters. The voiced, unvoiced, and pulsed strength parameters are quantized using a weighted vector quantization method using a novel error criterion for obtaining high quality quantization. The fundamental frequency and pulse position parameters are efficiently quantized based on the quantized strength parameters.

Abstract: An e-mail message process is provided for use with a personal digital assistant which allows for the use of input speech messaging which is converted to text using a focused language model which is downloaded by a cellular phone connection to an Internet server which provides the focused language model based upon a topic for the intended e-mail message. The text that is generated from the input speech method can be summarized by the e-mail message processor and can be edited by the user. The generated e-mail message can then be transmitted again via cellular connection to an Internet e-mail server for transmitting the e-mail message to a recipient.

Abstract: The invention relates to a method and apparatus for generating translations of natural language terms from a first language to a second language. A plurality of terms are extracted from unaligned comparable corpora of the first and second languages. Comparable corpora are sets of documents in different languages that come from the same domain and have similar genre and content. Unaligned documents are not translations of one another and are not linked in any other way. By accessing monolingual thesauri of the first and second languages, a category is assigned to each extracted term. Then, category-to-category translation probabilities are estimated, and using said category-to-category translation probabilities, term-to-term translation probabilities are estimated. The invention preferably exploits class-based normalization of probability estimates, bi-directionality, and relative frequency normalization.

Abstract: Successive pitch periods/frequencies are accurately determined in an audio equivalent signal. Using a suitable conventional pitch detection technique, an initial value of the pitch frequency/period is determined for so-called pitch detection segments of the audio equivalent signal. Based on the determined initial value, a refined value of the pitch frequency/period is determined. To this end, the signal is divided into a sequence of pitch refinement segments. Each pitch refinement segment is associated with at least one of the pitch detection segments. The pitch refinement segments are filtered to extract a frequency component with a frequency substantially corresponding to an initially determined pitch frequency of an associated pitch detection segment. The successive pitch periods/frequencies are determined in the filtered signal.

Abstract: Multimodal utterances contain a number of different modes. These modes can include speech, gestures, and pen, haptic, and gaze inputs, and the like. This invention use recognition results from one or more of these modes to provide compensation to the recognition process of one or more other ones of these modes. In various exemplary embodiments, a multimodal recognition system inputs one or more recognition lattices from one or more of these modes, and generates one or more models to be used by one or more mode recognizers to recognize the one or more other modes. In one exemplary embodiment, a gesture recognizer inputs a gesture input and outputs a gesture recognition lattice to a multimodal parser. The multimodal parser generates a language model and outputs it to an automatic speech recognition system, which uses the received language model to recognize the speech input that corresponds to the recognized gesture input.

Abstract: An audio type signal is encoded. The signal is first divided into bands. For each band, a yardstick signal element is selected. The yardstick may be the signal element having the largest magnitude in the band, the second largest, closest to the median magnitude, or having some other selected magnitude. This magnitude is used for various purposes, including assigning bits to the different bands, and for establishing reconstruction levels within a band. The magnitude of non yardstick signal elements is also quantized. The encoded signal is also decoded. Apparatus for both encoding and decoding are also disclosed. The location of the yardstick element within its band may also be recorded and encoded, and used for efficiently allocating bits to non-yardstick signal elements. Split bands may be established, such that each split band includes a yardstick signal element and each full band includes a major and a minor yardstick signal element.