Nikolaos Fakotakis

New algorithms for skewing correction and slant removal on word-level [OCR]

The word preprocessing is a crucial stage of OCR systems. Here we present two algorithms appropriate for this stage. The first one corrects the skewing of words and the second removes the slant from handwritten words. Both algorithms make use of the Wigner-Ville distribution and the projection profile technique. The algorithms have been tested on words taken from more than 200 writers and the results obtained can be considered very satisfactory, since the overall accuracy of our OCR system is notably improved.

Multimedia content analysis for emotional characterization of music video clips

Nowadays, tags play an important role in the search and retrieval process in multimedia content sharing social networks. As the amount of multimedia contents explosively increases, it is a challenging problem to find a content that will be appealing to the users. Furthermore, the retrieval of multimedia contents, which can match users’ current mood or affective state, can be of great interest. One approach to indexing multimedia contents is to determine the potential affective state, which they can induce in users. In this paper, multimedia content analysis is performed to extract affective audio and visual cues from different music video clips. Furthermore, several fusion techniques are used to combine the information extracted from the audio and video contents of music video clips. We show that using the proposed methodology, a relatively high performance (up to 90%) of affect recognition is obtained.

An unconstrained method for lip detection in color images

The use of visual information derived from accurate lip extraction, can provide features invariant to noise perturbation for speech recognition systems and can be also used in a wide variety of applications. Unlike many current automatic lip reading systems which impose several restrictions on users, our efforts are towards an unconstrained system. In this paper we introduce a method using k-means color clustering with automatically adapted number of clusters, for the extraction of the lip area. The methods performance is improved by previously applying nearest neighbor color segmentation. The extracted lip area is morphologically processed and fitted by a best-fit ellipse. The points of interest (keypoints) of the mouth area are extracted, while a corner detector for fine tuning of mouth corners is applied. Experimental tests have shown that the algorithm works very well under natural conditions and accurate extraction of lip keypoints is feasible.

Improving SMS Usability Using Bayesian Networks

During the last years, the significant increase of mobile communications has resulted in the wide acceptance of a plethora of new services, like communication via written short messages (SMS). The limitations of the dimensions and the number of keys of the mobile phone keypad are probably the main obstacles of this service. Numerous intelligent techniques have been developed aiming at supporting users of SMS services. Special emphasis has been provided to the efficient and effective editing of words. In the presented research, we introduce a predictive algorithm that forecasts Greek letters occurrence during the process of compiling an SMS. The algorithm is based on Bayesian networks that have been trained with sufficient Greek corpus. The extracted network infers the probability of a specific letter in a word given one, two or three previous letter that have been keyed by the user with precision that reaches 95%. An important advantage, compared to other predictive algorithms is that the use of a vocabulary is not required, so the limited memory resources of mobile phones can easily accommodate the presented algorithm. The proposed method achieves improvement in the word editing time compared to the traditional editing method by a factor of 34.72%, as this has been proven by using Keystroke Level Modeling technique described in the paper.

APPLYING SIMILARITY MEASURES FOR AUTOMATIC LEMMATIZATION: A CASE STUDY FOR MODERN GREEK AND ENGLISH

This paper addresses the problem of automatic induction of the normalized form (lemma) of regular and mildly irregular words with no direct supervision using language-independent algorithms. More specifically, two string distance metric models (i.e. the Levenshtein Edit Distance algorithm and the Dice Coefficient similarity measure) were employed in order to deal with the automatic word lemmatization task by combining two alignment models based on the string similarity and the most frequent inflectional suffixes. The performance of the proposed model has been evaluated quantitatively and qualitatively. Experiments were performed for the Modern Greek and English languages and the results, which are set within the state-of-the-art, have showed that the proposed model is robust (for a variety of languages) and computationally efficient. The proposed model may be useful as a pre-processing tool to various language engineering and text mining applications such as spell-checkers, electronic dictionaries, morphological analyzers etc.

Using the Levenshtein Edit Distance for Automatic Lemmatization: A Case Study for Modern Greek and English

In the present work we have implemented the Edit Distance (also known as Levenshtein Distance) on a dictionary-based algorithm in order to achieve the automatic induction of the normalized form (lemma) of regular and mildly irregular words with no direct supervision. The algorithm combines two alignment models based on the string similarity and the most frequent inflexional suffixes. In our experiments, we have also examined the language-independency (i.e. independency of the specific grammar and inflexional rules of the language) of the presented algorithm by evaluating its performance on the Modern Greek and English languages. The results were very promising as we achieved more than 95 % of accuracy for the Greek language and more than 96 % for the English language. This algorithm may be useful to various text mining and linguistic applications such as spell-checkers, electronic dictionaries, morphological analyzers, search engines etc.

A NOVEL AND EFFICIENT METHOD FOR PARSING UNRESTRICTED TEXTS OF QUASI FREE WORD ORDER LANGUAGES

This paper deals with the problems stemming from the parsing of long sentences in quasi free word order languages. Due to the word order freedom of a large category of languages including Greek and the limitations of rule-based grammar parsers in parsing unrestricted texts of such languages, we propose a flexible and effective method for parsing long sentences of such languages that combines heuristic information and pattern-matching techniques in early processing levels. This method is deeply characterized by its simplicity and robustness. Although it has been developed and tested for the Greek language, its theoretical background, implementation algorithm and results are language independent and can be of considerable value for many practical natural language processing (NLP) applications involving parsing of unrestricted texts.

Experimental evaluation of tree-based algorithms for intonational breaks representation

The prosodic specification of an utterance to be spoken by a Text-to-Speech synthesis system can be devised in break indices, pitch accents and boundary tones. In particular, the identification of break indices formulates the intonational phrase breaks that affect all the forthcoming prosody-related procedures. In the present paper we use tree-structured predictors, and specifically the commonly used in similar tasks CART and the introduced C4.5 one, to cope with the task of break placement in the presence of shallow textual features. We have utilized two 500-utterance prosodic corpora offered by two Greek universities in order to compare the machine learning approaches and to argue on the robustness they offer for Greek break modeling. The evaluation of the resulted models revealed that both approaches were positively compared with similar works published for other languages, while the C4.5 method accuracy scaled from 1% to 2,7% better than CART.