BACKGROUNDSpeech recognition has been the subject of a significant amount of research and commercial development. For example, speech recognition systems have been incorporated into mobile telephones, desktop computers, automobiles, and the like in order to provide a particular response to speech input provided by a user. For instance, in a mobile telephone equipped with speech recognition technology, a user can speak a name of a contact listed in the mobile telephone and the mobile telephone can initiate a call to the contact. Furthermore, many companies are currently using speech recognition technology to aid customers in connection with identifying employees of a company, identifying problems with a product or service, etc.
Research in ASR has explored layered architectures to perform speech recognition, motivated partly by the desire to capitalize on some analogous properties in the human speech generation and perception systems. In these studies, learning of model parameters has been one of the most prominent and difficult problems. In parallel with the development in ASR research, recent progresses made in learning methods from neural network research has ignited interest in exploration of deep-structured models. Examples of neural network research are given in
Ronan Collobert ET AL: "A unified architecture for natural language processing: deep neural networks with multitask learning", Paper accepted at the 25th International Conference on Machine Learning (ICML'08) to be held 5-9 July 2008, 2 May 2008 (2008-05-02), XP055104984, as well as
Trinh-Minh-Tri Do ET AL: "Neural conditional random fields", 13th international conference on artificial intelligence and statistics (AISTATS), Chia Laguna resort, Sardinia, Italy, Vol 9 of JMLR, 2010, pages 177-184, XP055130947. Hidden conditional random field models for phonetic classification and speech recognition are described in
US7627473. The article
BENGIO et al.: "GLOBAL OPTIMIZATION OF A NEURAL NETWORK HIDDEN MARKOV MODEL HYBRID" (IEEE TRANSACTIONS ON NEURAL NETWORKS, vol. 3, no. 2, 1 March 1992, pages 252-259) specifically relates to the integration of multilayered and recurrent artificial neural networks (ANNs) with hidden Markov models (HMMs) for speech recognition, where the parameters of the ANN and HMM subsystems can influence each other; accordingly. The optimization of the ANN/HMM system is directed towards driving a network gradient descent with parameters computed in the HMM. In the paper
LAFFERTY et al.: "Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data" (MACHINE LEARNING. PROCEEDINGS OF THE INTERNATIONAL CONFERENCE, 28 June 2001, pages 1-8), a framework of conditional random fields is discussed for building probabilistic models to segment and label sequence data. Lafferty sets forth iterative parameter estimation algorithms for conditional random fields, and further compares the performance of HMMs and maximum entropy Markov models (MEMMs) on synthetic and natural-language data. One particular advance is the development of effective learning techniques for deep belief networks (DBNs), which are densely connected, directed belief networks with many hidden layers. In general, DBNs can be considered as a highly complex nonlinear feature extractor with a plurality of layers of hidden units and at least one layer of visible units, where each layer of hidden units learns to represent features that capture higher order correlations in original input data.
Although DBNs typically have higher modeling power than their more shallow counterparts, learning in DBNs is difficult partly because a back-propagation algorithm often does not perform effectively due to the significantly increased chance of trapping into a local optimum.
Accordingly, improved learning techniques with respect to DBNs are desirable.
SUMMARYThe following is a brief summary of subject matter that is described in greater detail herein. This summary is not intended to be limiting as to the scope of the claims.
Described herein are various technologies pertaining to automatic speech recognition (ASR). With more specificity, various technologies pertaining to utilization of deep-structured models to perform ASR are described herein. With still more specificity, various technologies pertaining to performing full-sequence training of deep-structured models for speech recognition are described herein.
An exemplary deep-structured model that can be utilized in connection with ASR is a deep belief network (DBN). A pretraining procedure can be undertaken on a DBN, wherein such pretraining procedure can pertain to learning initial weights between layers of variables (visible and hidden) in the DBN. In an example, such pretraining procedure can learn the initial weights for each layer of the DBN greedily by treating each pair of layers in the DBN as a Restricted Boltzmann Machine (RBM).
Subsequent to the DBN being subjected to pretraining, the DBN weights, transition parameters, and language model (LM) scores can be substantially optimized jointly through utilization of a discriminative training criterion at the sequence level designed for DBNs. More particularly, speech recognition can be referred to as a sequential or full-sequence learning problem, and it has been well known that discriminative information at the sequence level contributes to improving recognition accuracy. In previous approaches, only frame-level information was utilized for training DBN weights, and transition parameters and LM scores are obtained separately.
Other aspects will be appreciated upon reading and understanding the attached figures and description.
BRIEF DESCRIPTION OF THE DRAWINGS- Fig. 1 is a functional block diagram of an exemplary system that facilitates performing automatic speech recognition (ASR) through utilization of a Deep Believe Network (DBN).
- Fig. 2 is a functional block diagram of an exemplary system that facilitates initializing weights of a DBN.
- Fig. 3 is a functional block diagram of an exemplary system that facilitates jointly substantially optimizing DBN weights, transition parameters, and language model (LM) scores.
- Fig. 4 is an exemplary DBN.
- Fig. 5 is a flow diagram that illustrates an exemplary methodology for jointly learning DBN weights, transition parameters, and LM scores.
- Fig. 6 is a flow diagram that illustrates an exemplary methodology for jointly learning DBN weights, transition parameters, and LM scores.
- Fig. 7 illustrates an exemplary Deep Hidden Conditional Random Field.
- Fig. 8 is an exemplary computing system.
DETAILED DESCRIPTIONVarious technologies pertaining to automatic speech recognition (ASR) systems will now be described with reference to the drawings, where like reference numerals represent like elements throughout. In addition, several functional block diagrams of example systems are illustrated and described herein for purposes of explanation; however, it is to be understood that functionality that is described as being carried out by certain system components may be performed by multiple components. Similarly, for instance, a component may be configured to perform functionality that is described as being carried out by multiple components, and some steps in methodologies described herein may be omitted, re-ordered, or combined.
With reference toFig. 1, an exemplary system 100 that facilitates performing ASR is illustrated. The system 100 includes a speech recognition system 102 that receives a sample 104. The sample can be spoken words from an individual over a particular amount of time (e.g., captured through utilization of a microphone). The sample 104 can be digitized through utilization of an analog to digital converter, and can be subject to some form of normalization if desired. While the examples provided herein indicate that the sample 104 is a spoken utterance, it is to be understood that the system 100 may be configured to perform online handwriting recognition and/or real-time gesture recognition. Thus, the sample 104 may be an online handwriting sample or a video signal describing movement of an object such as a human being.
The speech recognition system 102 comprises a deep-structured model 106. In an example, the deep-structured model 106 can be a Deep Belief Network (DBN), wherein the DBN is temporally parameter-tied. A DBN is a probabilistic generative model with multiple layers of stochastic hidden units above a single bottom layer of observed variables that represent a data vector. With more specificity, a DBN is a densely connected, directed belief network with many hidden layers for which learning is a difficult problem. The deep-structured model 106 can receive the sample 104 and can output state posterior probabilities with respect to an output unit, which can be a phone, a senone, or some other suitable output unit. As will be described in more detail below, the deep-structured model 106 can be generated through a pretraining procedure, and thereafter weights of the deep-structured model 106, transition parameters in the deep-structured model 106 and language model scores can be substantially optimized jointly through sequential or full-sequence learning.
The speech recognition system 102 additionally includes a decoder 108, which can decode output of the deep-structured model to generate an output 110. The output 110, pursuant to an example, can include an indication of a word or word sequence that was received as the sample 104. In another example, the output 110 may be a gesture that pertains to a gesture captured in a video sample. In yet another example, the output 110 can be an indication of a word or word sequence that is being written on a pressure-sensitive screen.
Pursuant to an example, the speech recognition system 102 can be deployed in a variety of contexts. For instance, the speech recognition system 102 can be deployed in a mobile telephone, such that the mobile telephone can act responsive to spoken commands of a user. In another example, the speech recognition system 102 can be deployed in an automobile, such that the automobile can act responsive to spoken commands of a user. Other systems within which the speech recognition system 102 can be employed include automated transcribing systems, industrial automation systems, banking systems, and other suitable systems that employ ASR technology.
Now referring toFig. 2, an exemplary 200 system that facilitates initializing weights of a DBN is illustrated. The system 200 comprises an initializer component 202 that receives a DBN 204. As described previously, DBNs are densely connected, directed belief network with many hidden layers for which learning is a difficult problem. The initializer component 202 can act to learn each layer of the DBN 204 greedily by treating each pair of layers as a Restricted Boltzmann Machine (RBM). The initializer component 202 can access training data in a data repository 206 to perform the aforementioned training. With more detail, an RBM is a particular type of Markov random field (MRF) that has one layer of (typically Bernoulli) stochastic hidden units and one layer of (typically Bernoulli or Gaussian) stochastic visible units. RBMs can be represented as bipartite graphs since all visible units are connected to all hidden units, but there are no visible-visible or hidden-hidden connections.
In the RBMs, the joint distribution p(v, h; θ) over the visible units v and hidden unitsh, given the model parameters θ, can be defined in terms of an energy function E(v,h; θ) of the following algorithm: whereZ = ∑u∑hexp(-E(v,h;θ)) is a normalization factor or partition function, and the marginal probability that the model assigns to a visible vector v can be defined as follows:
For a Bernoulli (visible)-Bernoulli (hidden) RBM, the energy is as follows: wherewij represents the symmetric interaction term between visible unitvi and hidden unithj, bi andaj represent the bias terms, and V andH are the numbers of visible and hidden units. The conditional probabilities can be calculated as follows: where
Similarly, for a Gaussian-Bernoulli RBM, the energy is as follows after assuming that the variance is unity: The corresponding conditional probabilities become: wherevi can take real values and can follow a Gaussian distribution with mean and variance of one. Gaussian-Bernoulli RBMs can be used to convert real-valued stochastic variables to binary stochastic variables which can then be further processed using the Bernoulli-Bernoulli RBMs.
Following the gradient of the log likelihood logp (v; θ) the update rule for the weights can be obtained by the initializer component 202 as follows: where 〈vihj〉data is the expectation observed in the training data and 〈vihj〉model is that same expectation under the distribution defined by the DBN 204. Unfortunately, 〈vihj〉model can be extremely expensive to compute exactly so the contrastive divergence (CD) approximation to the gradient may be used where 〈vihj〉model is replaced by running a Gibbs sampler initialized at the data for one full step.
From a decoding point of view, the DBN 204 can be treated as a multilayer perceptron with many layers. The input signal (from the training data) can be processed layer by layer through utilization of equation (4) until the final layer. The final layer can be transformed into a multinomial distribution using the following softmax operation: wherel = k denotes the input been classified into thek-th class, and λik is the weight between hidden unithi at the last layer and class label k.
Pursuant to an example, the initializer component 202 can utilize a conventional frame-level data to train the DBN 204. For instance, the initializer component 202 can train a stack of RBMs in a generative manner, resulting in output of a pretrained DBN 208. As will be described below, the DBN weights, transition parameters, and language model scores can be learned through utilization of a back-propagation algorithm by substantially maximizing the frame-level or utterance-level cross-entropy between the true and the predicted probability distributions over class labels. Furthermore, while the initializer component 202 has been described above as performing pretraining on a DBN in a particular manner, it is to be understood that weights of a DBN can be initialized through other methods, including but not limited to denoising/ autoencoding.
Now referring toFig. 3, an exemplary system 300 that facilitates jointly substantially optimizing DBN weights, transition parameters, and language model (LM) scores is illustrated. The system 300 includes a receiver component 301 that receives the pretrained DBN 208. A trainer component 302 that is in communication with the receiver component receives the pretrained DBN 208 and the training data in the data store 206 (which can be different training data than what was employed by the initializer component 202 or the same training data employed by the initializer component 202). The trainer component 302 can be configured to jointly substantially optimize weights of the pretrained DBN 208, state transition parameters, and language model scores. For instance, the trainer component 302 can utilize back-propagation to perform such joint fine-tuning of the DBN 208.
Conventional discriminative back-propagation methods optimize the log posterior probabilityp(lt|vt) of class labels given the current input, both at time-frame t (which may be a fixed local block of frames). This method of training DBNs can be referred to as a frame-based approach, because it only uses the frame (or frame-block) of an input sample to predict the class labels. The method does not explicitly use the fact that the neighboring frames (or frame-blocks) have smaller distances between the assigned probability distributions over class labels. To take this fact into account, the probability of the whole sequence of labels given the whole utterancep(l1:T|v1:T) can be modeled.
The approach described herein is to consider the top-most layer of the DBN as a linear-chain conditional random field (CRF) withht as input features from the lower layer at time t. This model can be viewed as a modification of a deep-structured CRF where the lower multiple layers of CRFs are replaced by DBNs.
The conditional probability of the full-sequence labels given the full-sequence input features in this sequential model can be given as follows: where the transition feature is as follows:γij is the parameter associated with this transition feature,htd is the d-th dimension of the hidden unit value at the t-th frame at the last layerht, andD is the dimension of (or number of units) at that hidden layer.To optimize the log conditional probability of the n-th utterance, the trainer component 302 can take the gradient over the activation parameters λkd, transition parameters γij, and M-th-layer weights as follows: It can be noted that the gradient can be considered as back-propagating the error versus in a frame-based training algorithm.
While the basic optimization algorithm with gradient descent can be succinctly described by equations (13), (14), and (15), which compute the gradients in analytical forms, several practical issues can be considered in the algorithm implementation. First, the top-layer CRF's state transition parameters can form a transition matrix, which is different from that of a Hidden Markov Model (HMM). In fact, such state transition parameters are a combination of the transition matrix and bi-phone/senone LM scores. Without proper constraints, the transition matrix may have low likelihoods of being transitioning between states that are prohibited in the left-to-right three-state HMMs even though the training data does not support such transitions. To prevent this from happening so that a sharper model may be built, this constraint can be enforced in the training by setting transition weights that are prohibited in the HMMs to have a very large negative value.
Second, since the weights in the DBNs are jointly optimized together with CRF's transition parameters, the optimization problem is no longer convex. For this reason, good initialization is crucial. The DBN weights can be initialized by the initializer component 202 (Fig. 2) described above. For example, the transition parameters can be initialized from the combination of the HMM transition matrices and the LM scores, and can be further optimized by tuning the transition features while fixing the DBN weights prior to the trainer component 302 performing joint optimization.
Third, there are two ways of doing decoding using a DBN that has been trained as described above. A first approach is to feed the log marginal probability logp(lt|v1:T) as the activation scores to the conventional HMM decoder and use the HMM transition matrices and LM scores in a conventional manner. This approach may work when the full-sequence training can improve the quality of logp(lt|v1:T). A second approach is to generate the state sequence first and then to map the state sequence to the phoneme/senone sequence. The decoding result may be further improved if insertion penalties are contemplated, which can be integrated into the decoder component 108 (Fig. 1) by modifying the transition parameters by the following: if statei is the final state of a phone and statej is the first state of a phone, whereϕ is the insertion penalty, andρ is the scaling factor.
The pretrained DBN 208 can be configured with the jointly optimized DBN weights, LM scores, and transition probabilities as the parameters. The trainer component 302 can further train the DBN 208 by way of back propagation.
Now referring toFig. 4, an exemplary DBN 400 is illustrated. A top level 402 of the DBN can be a linear-chain CRF 404, and the architecture of the DBN 400 can be viewed as shared DBNs unfolding over time (an exemplary shared DBN 406 is illustrated inFig. 4). The DBN 400 can receive the sample 104 or some derivation thereof, which can be partitioned into a plurality of observed variables 404 over time t. The observed variables 408 can represent data vectors at different instances in time. The DBN 400 further comprises multiple layers of stochastic hidden units 410. The DBN 400 has undirected connections 412 between the top two layers of the stochastic hidden units 410 and directed connections 414 to all other layers from the layers above. Weights w can be initially assigned to the directed and undirected connections 412 and 414, respectively, during the pretraining described above.λik is the weight between hidden unithi at the last layer in the DBN 400 and class labelk, andγij are transition probabilities between classes. In this exemplary embodiment, the DBN 400 can be trained such that the output units in the uppermost layer (theMth layer) can be modeled as a phonetic unit or subunit, such as a phone or senone.
With reference now toFigs. 5 and6, exemplary methodologies are illustrated and described. While the methodologies are described as being a series of acts that are performed in a sequence, it is to be understood that the methodology is not limited by the order of the sequence. For instance, some acts may occur in a different order than what is described herein. In addition, an act may occur concurrently with another act. Furthermore, in some instances, not all acts may be required to implement a methodology described herein.
Moreover, the acts described herein may be computer-executable instructions that can be implemented by one or more processors and/or stored on a computer-readable medium or media. The computer-executable instructions may include a routine, a sub-routine, programs, a thread of execution, and/or the like. Still further, results of acts of the methodologies may be stored in a computer-readable medium, displayed on a display device, and/or the like. The computer-readable medium may be a non-transitory medium, such as memory, hard drive, CD, DVD, flash drive, or the like.
With reference solely toFig. 5, an exemplary methodology 500 that facilitates training a deep-structured model for utilization in a speech recognition system is illustrated. The methodology 500 begins at 502, and at 504 parameters of a deep-structured model are provided through a pretraining step. For example, weights between layers of a DBN can be initialized during such pretraining step. At 506, labeled training data is provided to the deep structure, wherein the labeled training data may be labeled words or word sequences, labeled gestures, labeled handwriting samples, etc. At 508, weights between layers in the deep-structured model, language model parameters, and state transition probabilities are jointly substantially optimized, such that the resultant trained deep-structured model can be commissioned into a speech recognition system. The methodology 500 completes at 510.
Turning now toFig. 6, an exemplary methodology 600 that facilitates training a DBN for utilization in an automated speech recognition system is illustrated. The methodology 600 starts at 602, and at 604 each layer of a DBN that is configured for utilization in an automated speech recognition system is greedily learned. At 606, the log conditional probabilities of the output states/sequences of the DBN are substantially optimized through utilization of training data. At 608, weights in the DBN, transition parameters in the DBN, and language model scores are substantially simultaneously optimized based at least in part upon the log of the conditional probabilities of output states/sequences produced by the DBN. The methodology 600 completes at 610.
The systems and methodologies shown and described above have generally referred to utilizing a DBN in a speech recognition system; as indicated above, however, other deep structures can be employed. An exemplary deep structure that can be utilized is a Deep Hidden Conditional Random Field (DHCRF). Referring toFig. 7, an exemplary DHCRF 700 is illustrated. In an example, the Nth layer of the DHCRF can be a Hidden Conditional Random Field (HCRF), and the intermediate layers can be zero-th order CRFs that do not use state transition features.
In the exemplary DHCRF 700, the observation sequenceoj at layerj consists of two parts: the preceding layer's observation sequenceoj-1 and the frame-level log marginal posterior probabilities computed from the preceding layerj-1, where is the state value at layerj-1. The raw observations at the first layer can be denoted aso = [ot], t = 1,···,T.
Both parameter estimation and sequence inference in the DHCRF can be carried out bottom-up, layer by layer. The final layer's state sequence conditional probability can be shown as follows: whereN is the total number of layers, (·)T is the transposition of (·), is the observation sequence at the final layer, w is the output sequence (senone, phoneme, word, etc.), is a hypothesized state sequence,f(w, sN,oN) = [f1(w,sN,oN),···,fT(w, sN,oN)]T is the feature vector at the final layer,λN = is the model parameter (weight vector), andz(oN;λN) = ∑w,sN∈wexp ((λN)Tf(w,sN,oN)) is the partition function (normalization factor) to ensure probabilitiesp(w|oN;λN) sum to one. It can be ascertained that invalid sequences can be ruled out by summing over valid phoneme or word sequences only.
In contrast to the final layer, the state conditional probabilities at the intermediate layerj can be as follows: This is different from (17) in two ways. First, transition features are not used in (18) and observation featuresf(si,oj) can be simplified to which is defined below. Second, there is no summation over state sequences with all possible segmentations in (18).
Weights of the DHCRF can be learned utilizing a combination of supervised and unsupervised learning. The training supervision of the DHCRF 700 is available only at the final layer and can be directly determined by the problem to be solved. For example, in the phonetic recognition task, the phoneme sequence w is known at the final layer during the training phase. Parameter estimation at the final layer can thus be carried out in a supervised manner. The supervision, however, is not available for the intermediate layers, which play the role of converting original observations to some intermediate abstract representations. For this reason, an unsupervised approach can be utilized to learn parameters in the intermediate layers.
There are several approaches to learning the intermediate layer representations in the DHCRF 700. For example, the intermediate layer learning problem can be cast into a multi-objective programming (MOP) problem in which the average frame-level conditional entropy is minimized and the state occupation entropy is maximized at a substantially similar time. Minimizing the average frame-level conditional entropy can force the intermediate layers to be sharp indicators of subclasses (or clusters) for each input vector, while maximizing the occupation entropy guarantees that the input vectors be represented distinctly by different intermediate states. The MOP optimization algorithm alternates the steps in optimizing these two contradictory criteria until no further improvement in the criteria is possible or the maximum number of iterations is reached. The MOP optimization, however, can become difficult when the number of classes in the intermediate layers becomes higher (as in a phone recognition task) since it is hard to control when to switch to optimize the other criterion given the vastly increased probability of being trapped into a local optimum.
Alternatively, a GMM-based algorithm can be employed to learn parameters in the intermediate layers of the DHCRF 700. This algorithm can utilize a layer-by-layer approach: once a lower layer is trained, the parameters of that layer are fixed and the observation sequences of the next layer are generated using the newly trained lower-layer parameters. This process can continue until all the layers are trained.
With more specificity, to learn the parameters of an intermediate layer, a single GMM with diagonal covariance (initialized from the corresponding HMM model which is optimized using the Gaussian splitting strategy) can be trained. The following can then be assigned as the state value to each observation frame at layerj by assuming each Gaussian component is a state, where and are the mean and variance of the i-th Gaussian component at layerj: The parameters of the CRF at layerj can then be learned by maximizing the regularized log-conditional probability as follows: where k is the utterance ID, ||.∥1 is a L1-norm to enforce sparseness of the parameters associated with each state value, is the square of L2-norm to give preference to smaller weights, andσ1 andσ2 are positive values to determine the importance of each regularization term. A regularized dual averaging method can be used to solve this optimization problem with L1/L2 regularization terms.
Pursuant to an example, transition features may not be used in the intermediate layers. Instead, only the first- and second-order observation features can be used as follows: where ∘ is an element-wise product.
The final layer of the DHCRF 700 can be trained to optimize the following in a supervised manner: wherew(k) is the label for the output unit for thek-th utterance without segmentation information. In the final layer, the following can be used as features: whereδ(x) = 1 ifx is true, andδ(x) = 0 otherwise. are bi-gram language model (LM) features in which each output unit sequence w is consisted of I output units (e.g., senones, phonemes, or words), are state transition features, and and are the first- and second-order statistics generated from the observations, respectively.
Now referring toFig. 8, a high-level illustration of an example computing device 800 that can be used in accordance with the systems and methodologies disclosed herein is illustrated. For instance, the computing device 800 may be used in a system that supports ASR. In another example, at least a portion of the computing device 800 may be used in a system that supports training a DBN. The computing device 800 includes at least one processor 802 that executes instructions that are stored in a memory 804. The memory 804 may be or include RAM, ROM, EEPROM, Flash memory, or other suitable memory. The instructions may be, for instance, instructions for implementing functionality described as being carried out by one or more components discussed above or instructions for implementing one or more of the methods described above. The processor 802 may access the memory 804 by way of a system bus 806. In addition to storing executable instructions, the memory 804 may also store a training data set, a validation data set, a DBN, etc.
The computing device 800 additionally includes a data store 808 that is accessible by the processor 802 by way of the system bus 806. The data store may be or include any suitable computer-readable storage, including a hard disk, memory, etc. The data store 808 may include executable instructions, a DBN, a training data set, a validation data set, etc. The computing device 800 also includes an input interface 810 that allows external devices to communicate with the computing device 800. For instance, the input interface 810 may be used to receive instructions from an external computer device, from a user, etc. The computing device 800 also includes an output interface 812 that interfaces the computing device 800 with one or more external devices. For example, the computing device 800 may display text, images, etc. by way of the output interface 812.
Additionally, while illustrated as a single system, it is to be understood that the computing device 800 may be a distributed system. Thus, for instance, several devices may be in communication by way of a network connection and may collectively perform tasks described as being performed by the computing device 800.
As used herein, the terms "component" and "system" are intended to encompass hardware, software, or a combination of hardware and software. Thus, for example, a system or component may be a process, a process executing on a processor, or a processor. Additionally, a component or system may be localized on a single device or distributed across several devices. Furthermore, a component or system may refer to a portion of memory and/or a series of transistors.
It is noted that several examples have been provided for purposes of explanation. These examples are not to be construed as limiting the hereto-appended claims. Additionally, it may be recognized that the examples provided herein may be permutated while still falling under the scope of the claims.