Abstract: A segment detecting device according to an embodiment includes at least one memory; and at least one processor. The at least one processor receives at least one of (i) an input signal including a first signal and a second signal or (ii) feature data representing one or a plurality of features of the input signal, estimates a level of the second signal by inputting the input signal or the feature data into a neural network, and determines a segment including the second signal in the input signal based on the level of the second signal.

Abstract: A control device includes at least one memory, and at least one processor configured to detect a voice segment from sound data, the sound data being detected while a controlled object operates, and stop the controlled object based on following conditions: a speaking speed is a predetermined speed threshold or greater, the speaking speed being calculated based on a portion of the sound data in the voice segment; and a length of the voice segment is a predetermined length threshold or less.

Abstract: A controller includes at least one memory, and at least one processor. The at least one processor is configured to acquire speech, recognize the speech, determine whether the speech is uttered in a quiet voice, and control a movable part of a controlled apparatus in accordance with a result of the speech recognition. The at least one processor is configured to control the movable part of the controlled apparatus such that a sound pressure level of a sound generated by the movable part of the controlled apparatus is lower when it is determined that the speech is uttered in the quiet voice than when it is determined that the speech is not uttered in the quiet voice.

Abstract: According to an embodiment, a speech recognition apparatus includes a calculation unit that calculates, based on a speech signal, a score vector sequence including score vectors including an acoustic score for each of input symbols, a search unit that generates an input symbol string by searching for a path of the input symbol tracing the acoustic score having a high likelihood in the score vector sequence and that generates an output symbol representing a recognition result of the speech signal based on a recognition target symbol representing linguistic information as a recognition target among the input symbols, an additional symbol acquisition unit that obtains an additional symbol representing paralinguistic information and/or non-linguistic information from among the input symbols included in a range corresponding to the output symbol, and an output unit that outputs the output symbol and the obtained additional symbol in association with each other.

Abstract: A generation device includes a receiving unit and a generating unit. The receiving unit receives a model representing correspondence between one or more phonetic symbols and one or more words. The generating unit generates a first finite state transducer based on the model, the first finite state transducer at least including, as outgoing transitions from a first state representing transition destination of a first transition which has a first phonetic symbol of a predetermined type as input symbol, a second transition that has a second phonetic symbol, which is different than a particular symbol representing part or whole of input symbol of the first transition, as input symbol, and a third transition that has a third phonetic symbol, which represents the particular symbol or silence, as input symbol.

Abstract: According to an embodiment, a lattice finalization device finalizes a portion of a lattice that is generated by pattern recognition with respect to a signal on a frame-by-frame basis in chronological order. The device includes a detector and a finalizer. The detector is configured to detect, as a splitting position, a frame in the lattice in which the number of nodes and passing arcs is equal to or smaller than a reference value set in advance. The finalizer is configured to finalize nodes and arcs in paths from a start node to the splitting position in the lattice.

Abstract: According to an embodiment, an automaton deforming device includes a transforming unit and a deforming unit. The transforming unit generates second values by transforming first values, which either represent weights assigned to transitions in a weighted finite state automaton or represent values that are transformed into weights assigned to transitions in a weighted finite state automaton, in such a way that number of elements of a set of the first values are reduced and an order of the first values is preserved. The deforming unit deforms a weighted finite state automaton in which weights according to the second values are assigned to transitions.

Abstract: According to an embodiment, a recognition apparatus includes one or more processors. The one or more processors are configured to calculate, based on the input signal, a score vector sequence in which a plurality of score vectors each including respective scores of symbols are arranged; and cause, among: a first score vector in which a representative symbol corresponding to a best score is a recognition-target symbol; a second score vector in which a representative symbol is a non-target symbol, and a score of the representative symbol is worse than a first threshold; and a third score vector in which a representative symbol is a non-target symbol, and a score of the representative symbol is equal to the first threshold or better than the first threshold, a third score vector satisfying a predefined first condition, to pass through to filter the score vector sequence.

Abstract: An information processing device includes a phonetic converting unit, an HMM converting unit, and a searching unit. The phonetic converting unit converts a phonetic symbol sequence into a hidden Markov model (HMM) state sequence in which states of an HMM are aligned. The HMM converting unit converts the HMM state sequence into a score vector sequence indicating the degree of similarity to a specific pronunciation using a similarity matrix defining the similarity between the states of the HMM. The searching unit searches for a path having a better score for the score vector sequence than that of the other paths out of paths included in a search network and outputs a phonetic symbol sequence corresponding to the retrieved path.

Abstract: According to one embodiment, a transcription support system supports transcription work to convert voice data to text. The system includes a first storage unit configured to store therein the voice data; a playback unit configured to play back the voice data; a second storage unit configured to store therein voice indices, each of which associates a character string obtained from a voice recognition process with voice positional information, for which the voice positional information is indicative of a temporal position in the voice data and corresponds to the character string; a text creating unit that creates the text in response to an operation input of a user; and an estimation unit configured to estimate already-transcribed voice positional information indicative of a position at which the creation of the text is completed in the voice data based on the voice indices.

Abstract: According to an embodiment, a data processing device generates result data which represents a result of performing predetermined processing on series data. The device includes an upper-level processor and a lower-level processor. The upper-level processor attaches order information to data blocks constituting the series data. The lower-level processor performs lower-level processing on the data blocks having the order information attached thereto, and attaches common order information, which is in common with the data blocks, to values obtained as a result of the lower-level processing. The upper-level processor integrates the values, which have the common order information attached thereto, based on the common order information and performs upper-level processing to generate the result data.

Abstract: According to an embodiment, a generation device includes a receiver and a generator. The receiver is configured to receive a first model that converts subwords serving as elements of words into the words. The generator is configured to produce, on the basis of the first model, a first finite state transducer that includes a first path having transitions converting one or more subwords into one or more words and a second path, whose first state is the first state of the first path, having cyclic paths to which the subwords are assigned and a transition to which a class classifying a word is assigned.

Abstract: According to an embodiment, a search device searches paths of a digraph and includes a retriever, and an expander. The retriever is configured to, from among hypotheses stored in a storage, retrieve, as a target hypothesis, a single hypothesis for which a weight obtained by addition of a weight of an already-searched path corresponding to each hypothesis and a weight of the best path from a state of the head of concerned path to a final state is the best weight. The expander is configured to, when the retrieved target hypothesis is not a final hypothesis for which the search has been completed up to a final state, generate new hypotheses each holding an input symbol string that is present in a path in which the search has been performed from states held by the target hypothesis until finding a single input symbol, and write the generated hypotheses in the storage.

Abstract: According to an embodiment, a conversion device converts a first automaton into a second automaton, which both are weighted finite state automatons. The first automaton has a boundary of a path assigned with an input symbol, an appearance position of the boundary, and identifiers for identifying paths. The second automaton has path(s) except unnecessary path(s). The device includes a specifying unit and a search unit. The specifying unit is configured to specify, as a start position, a state of the head of a retrieved path in which a combined weight, which is obtained by adding an accumulated weight from an initial state to the state of the head of the retrieved path in the first automaton and a weight of the best path from the state of the head of the retrieved path to a final state, is best. The search unit is configured to search for a path in which a weight from the start position to a final state in the first automaton is best until reaching next boundary.

Abstract: According to an embodiment, a decoder searches a finite state transducer and outputs an output symbol string corresponding to a signal that is input or corresponding to a feature sequence of signal that is input. The decoder includes a token operating unit and a duplication eliminator. The token operating unit is configured to, every time the signal or the feature is input, propagate each of a plurality of tokens, which is assigned with a state of the head of a path being searched, according to the finite state transducer. The duplication eliminator is configured to eliminate duplication of two or more tokens which have same state assigned thereto and for which respective previously-passed transitions are assigned with same input symbol.

Abstract: According to an embodiment, a speech recognition apparatus includes a calculation unit that calculates, based on a speech signal, a score vector sequence including score vectors including an acoustic score for each of input symbols, a search unit that generates an input symbol string by searching for a path of the input symbol tracing the acoustic score having a high likelihood in the score vector sequence and that generates an output symbol representing a recognition result of the speech signal based on a recognition target symbol representing linguistic information as a recognition target among the input symbols, an additional symbol acquisition unit that obtains an additional symbol representing paralinguistic information and/or non-linguistic information from among the input symbols included in a range corresponding to the output symbol, and an output unit that outputs the output symbol and the obtained additional symbol in association with each other.

Abstract: According to an embodiment, a recognition apparatus includes one or more processors. The one or more processors are configured to calculate, based on the input signal, a score vector sequence in which a plurality of score vectors each including respective scores of symbols are arranged; and cause, among: a first score vector in which a representative symbol corresponding to a best score is a recognition-target symbol; a second score vector in which a representative symbol is a non-target symbol, and a score of the representative symbol is worse than a first threshold; and a third score vector in which a representative symbol is a non-target symbol, and a score of the representative symbol is equal to the first threshold or better than the first threshold, a third score vector satisfying a predefined first condition, to pass through to filter the score vector sequence.

Abstract: A generation device includes a receiving unit and a generating unit. The receiving unit receives a model representing correspondence between one or more phonetic symbols and one or more words. The generating unit generates a first finite state transducer based on the model, the first finite state transducer at least including, as outgoing transitions from a first state representing transition destination of a first transition which has a first phonetic symbol of a predetermined type as input symbol, a second transition that has a second phonetic symbol, which is different than a particular symbol representing part or whole of input symbol of the first transition, as input symbol, and a third transition that has a third phonetic symbol, which represents the particular symbol or silence, as input symbol.

Abstract: According to an embodiment, an information processing apparatus includes a storage unit, a detector, an acquisition unit, and a search unit. The storage unit configured to store therein voice indices, each of which associates a character string included in voice text data obtained from a voice recognition process with voice positional information, the voice positional information indicating a temporal position in the voice data and corresponding to the character string. The acquisition unit acquires reading information being at least a part of a character string representing a reading of a phrase to be transcribed from the voice data played back. The search unit specifies, as search targets, character strings whose associated voice positional information is included in the played-back section information among the character strings included in the voice indices, and retrieves a character string including the reading represented by the reading information from among the specified character strings.

Abstract: According to an embodiment, a decoder includes a token operating unit, a node adder, and a connection detector. The token operating unit is configured to, every time a signal or a feature is input, propagate each of a plurality of tokens, which is an object assigned with a state of the of a path being searched, according to a digraph until a state or a transition assigned with a non-empty input symbol is reached. The node adder is configured to, in each instance of token propagating, add, in a lattice, a node corresponding to a state assigned to each of the plurality of tokens. The connection detector is configured to refer to the digraph and detect a node that is connected to a node added in an i-th instance in the lattice and that is added in an i+1-th instance in the lattice.