Chinese character input is always a key issue in a variety of Chinese based applications especially when only a small number keypad is available. Though many kinds of Chinese character encoding schemes are proposed ac...
详细信息
ISBN:
(纸本)7801501144
Chinese character input is always a key issue in a variety of Chinese based applications especially when only a small number keypad is available. Though many kinds of Chinese character encoding schemes are proposed according to Chinese character characteristics, such as the shape, they are not straightforward and will take users a long time to learn. An easy way is to input via Chinese pinyins. In this paper, we establish the mapping between digit string and pinyin as well as the mapping between the pinyin string and the word, referred to as the Syllable-Digit search Tree (SDT) and the Word-Syllable search Tree (WST) respectively. By using these two search trees as well as the word N-gram language model and the syllable-synchronous network search (SSNS) algorithm, any digit string can be easily converted into Chinese word sequence or sentence. Without users' selecting from candidates, the character error rate (CER) of digit-to-character (D/C) conversion is 6.6% across a test text consisting 22,083 characters.
In this paper, we present an Equivalent-Class Based Maximum Mutual Information (ECB-MMI) learning method for our previously proposed Mixed Gaussian Continuous Probability Model (MGCPM). Similar to HMMs, the defined ob...
详细信息
ISBN:
(纸本)7801501144
In this paper, we present an Equivalent-Class Based Maximum Mutual Information (ECB-MMI) learning method for our previously proposed Mixed Gaussian Continuous Probability Model (MGCPM). Similar to HMMs, the defined object function for MGCPM training considers the mutual information among different models so as to maximally separate the Speech Recognition Units (SRUs) in model space. Experimental result shows that for MGCPM the MMI training method can improve the recognition rate by 5% compared to the traditional training method MLE (Maximum Likelihood Estimation). Because the computation amount of MM algorithm is very large, we propose an N-Best strategy to find the corresponding equivalent class (EC) in order to reduce complexity. Our experimental result shows that this criterion works very well.
Read and spontaneous discourses are two different but very significant speech styles to be investigated. So phonetic labeling on read and spontaneous discourse corpora are made one is ASCCD, a 10 hours read discourse ...
ISBN:
(纸本)7801501144
Read and spontaneous discourses are two different but very significant speech styles to be investigated. So phonetic labeling on read and spontaneous discourse corpora are made one is ASCCD, a 10 hours read discourse corpus and the other is CASS, a 4 hours spontaneous discourse corpus. First the principles and conventions of transcription are presented. Then, these two speech styles are compared from phonetic and syntactic point of view, including the statistic results of different phonetic units got from the annotated corpora.
In this paper, we propose a new ambiguity representation scheme;Structure Preference Relation (SPR), which consists of useful quantitative distribution information for ambiguous structures. Two automatic acquisition a...
In this paper, we present a new method for reliability evaluation of speech recognition in acoustic modeling. The new method incorporates the Integrated Model (IM), which is trained by all the speech data. For close-s...
详细信息
ISBN:
(纸本)7801501144
In this paper, we present a new method for reliability evaluation of speech recognition in acoustic modeling. The new method incorporates the Integrated Model (IM), which is trained by all the speech data. For close-set verification, the IM method has theoretically the lowest equal error rate;and for open-set, the method often performs well. At the same time, it costs much less computation than other verification methods commonly used. Experiment results show the method is feasible.
As the outcome of a 3-year joint effort of Department of Computer Science, Tsinghua University and Language Information Processing Institute, Beijing Language and Culture University, Beijing, China, a word-segmented a...
As the outcome of a 3-year joint effort of Department of Computer Science, Tsinghua University and Language Information Processing Institute, Beijing Language and Culture University, Beijing, China, a word-segmented and part-of-speech tagged Chinese corpus with size of 2 million Chinese characters, named HuaYu, has been established. This paper firstly introduces some basics about HuaYu in brief, as its genre distribution, fundamental considerations in designing it, word segmentation and part-of-speech tagging standards. Then the complete list of tag set used in HuaYu is given, along with typical examples for each tag accordingly. Several pieces of annotated texts in each genre are also included at last for reader's reference.
The dynamic modeling of two-link flexible manipulators is specially examined by means of Lagrange equations, assuming rigid arms and finite modal expansion. The obtained model is integrated with the dynamics of drivin...
详细信息
The dynamic modeling of two-link flexible manipulators is specially examined by means of Lagrange equations, assuming rigid arms and finite modal expansion. The obtained model is integrated with the dynamics of driving motors, as benefits the practical control. A sensor-based control strategy of separate joints is presented to suppress flexible vibrations, and the globally asymptotically stable closed-loop system is analyzed.
Based on the analyses of the limitation of common genetic algorithm and the model of the structure and development of human society,this paper discusses a modified genetic *** algorithm applies unidirectional inherita...
详细信息
Based on the analyses of the limitation of common genetic algorithm and the model of the structure and development of human society,this paper discusses a modified genetic *** algorithm applies unidirectional inheritance and hierarchical structure to solve the optimizing *** method can not only reserve the useful genetic information,but also make the succeeding inheritance more *** utilizes the intrinsic genetic knowledge to calculate.
In this paper,a two-stage approach for pattern recognition problems of large scale is *** approach consists of two steps,where two new technologies are *** first one is a neural network construction method that can be...
详细信息
In this paper,a two-stage approach for pattern recognition problems of large scale is *** approach consists of two steps,where two new technologies are *** first one is a neural network construction method that can be used to build very complex decision boundaries for difficult pattern classification *** second is a coarse classification method that takes speed and accuracy into consideration at the same *** the maximum size of the resulted cluster is controlled in order to avoid too much difference among the size of resulted *** of 1000 hand written Chinese characters is used to test the performance of the approach and the results are promising.
Structured rank-deficient matrices arise in many applications in signal *** inverse iteration algorithm was proposed to solve the socalled structured total least squares *** algorithm,however,converges to local-minimu...
详细信息
Structured rank-deficient matrices arise in many applications in signal *** inverse iteration algorithm was proposed to solve the socalled structured total least squares *** algorithm,however,converges to local-minimum under certain *** is well known that Genetic algorithms are stochastic optimization techniques that can often outperform classical methods of *** algorithms was utilized here to get the better solution of the STLS problems. Computer simulations show that our method ensures convergence to global minimum.
暂无评论