Multi-modal feature integration for story boundary detection in broadcast news

Abstract

Abstract-This paper investigates how to integrate multi-modal features for story boundary detection in broadcast news. The detection problem is formulated as a classification task, i.e., classifying each candidate into boundary/non-boundary based on a set of features. We use a diverse collection of features from text, audio and video modalities: lexical features capturing the semantic shifts of news topics and audio/video features reflecting the editorial rules of broadcast news. We perform a comprehensive evaluation on boundary detection performance for six popular classifiers, including decision tree (DT), Bayesian network (BN), naive Bayesian (NB) classifier, multi-layer peceptron (MLP), support vector machines (SVM) and maximum entropy (ME) classifier. Results show that BN and DT can generally achieve superior performances over other classifiers and BN offers the best F1-measure. Analysis of BN and DT reveals important inter-feature dependencies and complementarities that contribute significantly to the performance gain.

MAHIR System: Unsupervised Segmentation for Malay Spoken Broadcast News Stories

“…A second disadvantage is that the apriori algorithm requires multiple scans of the database to generate the candidates. Many algorithms based on existing apriori algorithms have been designed with modifications or enhancements [8], [22]. In this study, we modified and enhanced an existing apriori algorithm to meet our specific aims.…”

“…The vectors are scrutinized to search for which items are recurrently joined together by various objects (i.e., linked or associated together). These co-occurrences are given in the form of the rules of association [8], [20]:…”

“…Confidence and support serve as a yardstick for determining the quality of a particular rule in terms of its utility (strength) and certainty [8], [20]. Support provides an idea of how many of the examples (transactions) of a dataset are used to produce the rule, including elements of both LHS and RHS.…”

“…Accordingly, audio searching has become more popular [1], [5], [6]. However, a number of obstacles, including the lack of overt punctuation and formatting and the difficulty in detecting story boundaries or segments, make the process of retrieving spoken documents challenging [6]- [8]. In particular, identifying word errors generated by ASR is one of the major challenges facing SDR.…”

Data-Driven Audio Feature Space Clustering for Automatic Sound Recognition in Radio Broadcast News

“…[1][2][3][4][5][6][7][8][9] Concerning the audio data, the automatic analysis of the audio signals can offer the users useful information. In the case of broadcast news, automatic processing is related to tasks such as sound recognition, 10,11 speaker recognition, 12 anchor detection, 13 role detection, [14][15][16] story boundary detection, 2,17,18 summary construction from anchor talking, 9,19 channel's quality detection, 20 sound event detection, 21,22 non-linguistic humanproduced sounds detection, 5,6,[23][24][25] audio type segmentation in sport games, 4,26,27 highlight scene extraction from sports games, 3 violence scene detection, 28 music characteristics classification, 29,30 jingle detection, 1 commercial block detection, 8 voice activity detection, 31 language recognition, 32 emotion recognition 33 and speech recognition. 34 Sound recognition is the cornerstone of analysis as typically precedes the other stages.…”