Yoshihiro Sekiguchi

Keyword Extraction using Term-Domain Interdependence for Dictation of Radio News

In this paper, we propose keyword extraction method for dictation of radio news which consists of several domains. In our method, newspaper articles which are automatically classified into suitable domains are used in order to calculate feature vectors. The feature vectors shows term-domain interdependence and are used for selecting a suitable domain of each part of radio news. Keywords are extracted by using the selected domain. The results of keyword extraction experiments showed that our methods are robust and effective for dictation of radio news.

Keyword extraction of radio news using term weighting with an encyclopedia and newspaper articles

In this paper, we propose a method for keyword extraction of radio news. Using our method, data sparseness problem and false alarm problem was Iightened even for short discourse or document. Also, our method is robust for partial errors of phoneme recognition. In our method, there are two procedures: i.e. term weighting and keyword extraction. In procedure of term weighting, a feature vector of each domain is calculated using an encyclopedia and newspaper articles. In procedure of keyword extraction, keywords are extracted using feature vectors and result of domain identification. The results of experiments demonstrate the appbcability of the method.

Spoken Term Detection Using Phoneme Transition Network from Multiple Speech Recognizers' Outputs

Spoken Term Detection (STD) that considers the out-of-vocabulary (OOV) problem has generated signifi- cant interest in the field of spoken document processing. This study describes STD with false detection control using phoneme transition networks (PTNs) derived from the outputs of multiple speech recognizers. PTNs are similar to subword-based confusion networks (CNs), which are originally derived from a single speech recognizer. Since PTN- formed index is based on the outputs of multiple speech recognizers, it is robust to recognition errors. Therefore, PTN should also be robust to recognition errors in an STD task, when compared to the CN-formed index from a single speech recognition system. Our PTN-formed index was evaluated on a test collection. The experiment showed that the PTN-based approach effectively detected OOV terms, and improved the F-measure value from 0.370 to 0.639 when compared with a baseline approach. Furthermore, we applied two false detection control parameters, one is based on the majority voting scheme. The other is a measure of the ambiguity of CN, to the calculation of detection score. By introducing these parameters, the performance of STD was found to be better (0.736 for the F-measure value) than that without any parameters (0.639).

Entropy-based false detection filtering in spoken term detection tasks

This paper describes spoken term detection (STD) and inexistent STD (iSTD) methods using term detection entropy based on a phoneme transition network (PTN)-formed index. Our previously reported STD method uses a PTN derived from multiple automatic speech recognizers (ASRs) as an index. A PTN is almost the same as a sub-word-based confusion network, which is derived from the output of an ASR. In the previous study, our PTN was very effective in detecting query terms. However, the PTN generated many false detection errors. In this study, we focus on entropy of the PTN-formed index. Entropy is used to filter out false detection candidates in the second pass of the STD process. Our proposed method was evaluated using the Japanese standard test-set for the STD and iSTD tasks. The experimental results of the STD task showed that entropy-based filtering is effective for improving STD at a high-recall range. In addition, entropy-based filtering was also demonstrated to work well for the iSTD task.

Word error correction of continuous speech recognition using WEB documents for spoken document indexing

This paper describes an error correction method of continuous speech recognition using WEB documents for spoken documents indexing. We performed an experiment of error correction for news speech automatically transcribed, where we focused on especially proper nouns. Two LVCSR systems were used to detect correctly and incorrectly recognized words. Keywords for the Internet search engine were selected among the correctly transcribed words, then correct candidates for the mis-recognized words were obtained in retrieved documents. A Dynamic Programming (DP) technique with a confusion matrix was utilized to compare the candidates with the mis-recognized words. In results of experiment of error correction, recognition rate of proper nouns achieved improvement of about 10% by using WEB documents.

Evaluation of the usefulness of spoken term detection in an electronic note-taking support system

The usefulness of a spoken term detection (STD) technique in an electronic note-taking support system is assessed through a subjective evaluation experiment. In this experiment, while listening to a lecture, subjects recorded electronic notes using the system. They answered questions related to the lecture while browsing the recorded notes. The response time required to correctly answer the questions was measured. When the subjects browsed the notes, half of them used the STD technique and half did not. The experimental results indicate that the subjects who used the STD technique answered all questions faster than those who did not use the STD technique. This indicates that the STD technique worked well in the electronic note-taking system.

Topic tracking using subject templates and clustering positive training instances

Topic tracking, which starts from a few sample stories and finds all subsequent stories that discuss the same topic, is a new challenge for the text categorization task and is useful for timeline-based IR systems. Much previous research on topic tracking use machine learning techniques. However, the small size of the training data, especially positive training stories, presents difficulties in training the parameters of the topic tracking system to produce optimal results. In this paper, we present a method for topic tracking using subject templates and k -means clustering algorithm to select a suitable training set. The method was tested on the TDT1 corpus, and the result shows the effectiveness of the method.

Speech recognition system for spoken Japanese sentences

A speech recognition system for continuously spoken Japanese simple sentences is described. The acoustic analyser based on a psychological assumption for phoneme identification can represent the speech sound by a phoneme string in an expanded sense which contains acoustic features such as buzz and silence as well as ordinary phonemes. Each item of the word dictionary is written in Roman letters of Hepburn system, and the reference phoneme string and the reference characteristic phoneme string necessary for matching procedure of input phoneme sequences are obtained from the word dictionary using a translating routine. In syntax analysis, inflexion of verbs and adjectives and those of some main auxiliary verbs are taken into account. The syntax analyser uses a network dealing with state transition among Parts of speech, predicts following words and outputs their syntactic interpretation of the input phoneme string. The semantic knowledge system deals with semantic definition of each verb, semantic nature of each word and the schema of the sentence, and conconstructs a semantic network. The semantic analyser examines semantic validity of the recognized sentence as to whether each word in the sentence meets the definition of the recognized verb or others. The present object of recognition is a Japanese fairy tale composed of simple sentences alone. The syntactic and semantic analysers work well and can recognize simple sentences provided that the acoustic analyser outputs correct phoneme strings. For real speech, though the level of semantic processing is yet low, it can recognize 25 blocks out of 33 blocks (A block means a part of speech sound uttered in a breath.), and 9 sentences out of 16 sentences uttered by an adult male.

Domain identification and keyword extraction of radio news using term weighting

We propose a method for domain identification and keyword extraction using term weighting for radio news. In our method, feature vectors whose elements are /spl chi//sup 2/ values between each keyword and each domain are calculated from newspaper articles automatically. Using the feature vectors, a domain of each part of radio news is selected. Then keywords are extracted by using the selected domain. The results of experiments show that our methods are robust and effective for the speech recognition system.

Associative information for speech recognition using semantic attributes

Speech recognition has been studied extensively in recent years with regard to human interface and the goal has been realization of speech recognition systems. However, the problem is that these systems have numerous limitations, despite being accompanied by implementation of speech-based interactive systems. There have been a few attempts to use associative information in Japanese speech recognition and understanding. As a result, this paper introduces a new concept of association between semantic attributes and associations between words under the assumption of so-called “indirect associations” in human communication. Hence, nearly successful processing of associations close to common sense relations is achieved.