Abstract This paper proposes flexible and multi-purpose web-based distributed architecture for a multilingual textto- speech synthesis system (TTS). The proposed architecture is complex client/server architecture, composed of several modules implemented as finite-state engines and located by different users worldwide. Data exchange (text/audio) between modules is implemented through the use of protocol standards that have wide support throughout the speech and telecommunications... This paper proposes flexible and multi-purpose web-based distributed architecture for a multilingual textto- speech synthesis system (TTS). The proposed architecture is complex client/server architecture, composed of several modules implemented as finite-state engines and located by different users worldwide. Data exchange (text/audio) between modules is implemented through the use of protocol standards that have wide support throughout the speech and telecommunications area. The integration of modules into complex client/server architecture provides a flexible, reliable, easily re-configurable and maintainable architecture for the TTS system. In the proposed architecture all modules use the same data structure (finitestate engines), all input/output formats are standardized and compatible, the structure is modular, interchangeable, easily maintainable, and object oriented. The proposed architecture has the advantage that researchers can concentrate their efforts on a single module and test its performance within a complete TTS system, composed of modules from different researchers worldwide. The benefit is that such an approach enables the composition of a high- performance TTS system by using high performance modules from different institutions. This paper describes a flexible web-based distributed architecture of the TTS system and exposes its great flexibility regarding numerous simultaneous configuration possibilities. Read More Read Less

Abstract Stemming is a technique for reducing variant forms of a word to their roots (or stems) by enabling extraction of common suffixes. Stem need not correspond to the linguistic root of a word. Stemming is predominantly used in IR system to enrich retrieval effectiveness and to reduce the size of index for information retrieval task. This paper presents a systematic... Stemming is a technique for reducing variant forms of a word to their roots (or stems) by enabling extraction of common suffixes. Stem need not correspond to the linguistic root of a word. Stemming is predominantly used in IR system to enrich retrieval effectiveness and to reduce the size of index for information retrieval task. This paper presents a systematic way of algorithm and implementation to develop an unsupervised Telugu stemmer using Take-All-Splits heuristic and improved paradigms. The performance of this stemmer is evaluated by taking two sets of 500 randomly extracted words. The trained Telugu corpus is taken from large set of documents containing 129066 words from CIIL Mysore. The accuracy is found to be 85.40% after applying normalization heuristic. The F-score is 92.94%, the recall is 91.97% and the precision is 96.91%. As the algorithm does not require any language specific information, it can be applied to other languages that are morphologically rich. The percentage reduction in index size for Telugu information retrieval task is also evaluated before and after normalization heuristic Read More Read Less

Abstract Arabic is a strongly structured and highly derivational and inflectional language. Arabic morphology and syntax provide the ability to add a large number of affixes to each word which makes combinatorial increment of possible words. Arabic morphological analysis has gained the focus of Arabic natural language processing research for a long time in order to achieve the automated understanding of... Arabic is a strongly structured and highly derivational and inflectional language. Arabic morphology and syntax provide the ability to add a large number of affixes to each word which makes combinatorial increment of possible words. Arabic morphological analysis has gained the focus of Arabic natural language processing research for a long time in order to achieve the automated understanding of Arabic. In this paper, we present a new approach for Arabic morphological analysis and generation based on morphological automata and using our lexicon. Examples of morphological analysis results from the proposed approach will be given and will illustrate how the system works. Our system will be very useful for NLP applications such as syntactic and semantic analysis, spell-checking, machine translation and information retrieval. Read More Read Less

Abstract In recent years, there has been much focus on the design and development of database management systems that incorporate and provide more flexible query operators that return data items which are dominating other data items in all attributes (dimensions). This type of query operations is named preference queries as they prefer one data item over the other data item if... In recent years, there has been much focus on the design and development of database management systems that incorporate and provide more flexible query operators that return data items which are dominating other data items in all attributes (dimensions). This type of query operations is named preference queries as they prefer one data item over the other data item if and only if it is better in all dimensions and not worse in at least one dimension. Several preference evaluation techniques for preference queries have been proposed including top-k, skyline, top-k dominating, k-dominance, and k- frequency. All of these preference evaluation techniques aimed at finding the “best” answer that meet the user preferences. This paper evaluates these five preference evaluation techniques on real application when huge number of dimensions is the main concern. To achieve this, a recipe searching application with maximum number of 60 dimensions has been developed which assists users to identify the most desired recipes that meet their preferences. Two analyses have been conducted, where execution time is the measurement used. Read More Read Less