Search

The language modeling approach to IR is attractive and promising because it connects the problem of retrieval with that of language model estimation. A core technique for language model estimation is smoothing, which adjusts the maximum likelihood estimator so as to correct the inaccuracy due to data sparseness. In this paper we propose a General Jelinek-Mercer method (GJM) by using a document-dependent mixture coefficient to control the influence of maximum likelihood model and the collection model. Utilizing the number of unique terms in the document to improve the accuracy of language model estimation, we further develop GJM-2 smoothing method as a special case of GJM. Experimental results show that using GJM-2 for the language modeling approach can achieve better retrieval performances than the existing three popular methods both on short and long queries. [ABSTRACT FROM AUTHOR]

The KL divergence framework, the extended language modeling approach has a critical problem with estimation of query model, which is the probabilistic model that encodes user's information need. At initial retrieval, estimation of query model by translation model had been proposed that involves term co-occurrence statistics. However, the translation model has a difficulty to applying, because term co-occurrence statistics must be constructed in offline. Especially in large collection, constructing such large matrix of term co-occurrences statistics prohibitively increases time and space complexity. More seriously, because translation model comprises noisy non-topical terms in documents, reliable retrieval performance cannot be guaranteed. This paper proposes an effective method to construct co-occurrence statistics and eliminate noisy terms by employing parsimonious translation model. Parsimonious translation model is a compact version of translation model and enables to drastically reduce number of terms that includes non-zero probabilities by eliminating non-topical terms in documents. From experimentations, we show that query model estimated from parsimonious translation model significantly outperforms not only baseline language modeling but also non-parsimonious model. [ABSTRACT FROM AUTHOR]

Previous research has shown that using term associations could improve the effectiveness of information retrieval (IR) systems. However, most of the existing approaches focus on query reformulation. Document reformulation has just begun to be studied recently. In this paper, we study how to utilize term association measures to do document modeling, and what types of measures are effective in document language models. We propose a probabilistic term association measure, compare it to some traditional methods, such as the similarity co-efficient and window-based methods, in the language modeling (LM) framework, and show that significant improvements over query likelihood (QL) retrieval can be obtained. We also compare the method with state-of-the-art document modeling techniques based on latent mixture models. [ABSTRACT FROM AUTHOR]

In language engineering, language models are employed in order to improve system performance. These language models are usually N-gram models which are estimated from large text databases using the occurrence frequencies of these N-grams. An alternative to conventional frequency-based estimation of N-gram probabilities consists on using neural networks to this end. In this paper, an approach to language modeling with a hybrid language model is presented as a linear combination of a connectionist N-gram model, which is used to represent the global relations between certain linguistic categories, and a stochastic model of word distribution into such categories. The hybrid language model is tested on the corpus of the Wall Street journal processed in the Penn Treebank project. [ABSTRACT FROM AUTHOR]

Genomic IR, characterized by its highly specific information need, severe synonym and polysemy problem, long term name and rapid growing literature size, is challenging IR community. In this paper, we are focused on addressing the synonym and polysemy issue within the language model framework. Unlike the ways translation model and traditional query expansion techniques approach this issue, we incorporate concept-based indexing into a basic language model for genomic IR. In particular, we adopt UMLS concepts as indexing and searching terms. A UMLS concept stands for a unique meaning in the biomedicine domain; a set of synonymous terms will share same concept ID. Therefore, the new approach makes the document ranking effective while maintaining the simplicity of language models. A comparative experiment on the TREC 2004 Genomics Track data shows significant improvements are obtained by incorporating concept-based indexing into a basic language model. The MAP (mean average precision) is significantly raised from 29.17% (the baseline system) to 36.94%. The performance of the new approach is also significantly superior to the mean (21.72%) of official runs participated in TREC 2004 Genomics Track and is comparable to the performance of the best run (40.75%). Most official runs including the best run extensively use various query expansion and pseudo-relevance feedback techniques while our approach does nothing except for the incorporation of concept-based indexing, which evidences the view that semantic smoothing, i.e. the incorporation of synonym and sense information into the language models, is a more standard approach to achieving the effects traditional query expansion and pseudo-relevance feedback techniques target. [ABSTRACT FROM AUTHOR]

In information retrieval, the word mismatch problem is a critical issue. To resolve the problem, several techniques have been developed, such as query expansion, cluster-based retrieval, and dimensionality reduction. Of these techniques, this paper performs an empirical study on query expansion and cluster-based retrieval. We examine the effect of using parsimony in query expansion and the effect of clustering algorithms in cluster-based retrieval. In addition, query expansion and cluster-based retrieval are compared, and their combinations are evaluated in terms of retrieval performance. By performing experimentation on seven test collections of NTCIR and TREC, we conclude that 1) query expansion using parsimony is well performed, 2) cluster-based retrieval by agglomerative clustering is better than that by partitioning clustering, and 3) query expansion is generally more effective than cluster-based retrieval in resolving the word-mismatch problem, and finally 4) their combinations are effective when each method significantly improves baseline performance. [ABSTRACT FROM AUTHOR]

Although n-gram models are still the de facto standard in language modeling for speech recognition, it has been shown that more sophisticated models achieve better accuracy by taking additional information, such as syntactic rules, semantic relations or domain knowledge into account. Unfortunately, most of the effort in developing such models goes into the implementation of handcrafted inference routines. What lacks is a generic mechanism to introduce background knowledge into a language model. We propose the use of dynamic Bayesian networks for this purpose. Dynamic Bayesian networks can be seen as a generalization of the n-gram models and hmms traditionally used in language modeling and speech recognition. Whereas those models use a single random variable to represent state, Bayesian networks can have any number of variables. As such they are particularly well-suited for the construction of models that take additional information into account. In this paper language modeling with belief networks is discussed. Examples of belief network implementations of well-known language models are given and a new model is presented that models dependencies between the content words in a sentence. [ABSTRACT FROM AUTHOR]

This paper presents a multi-dependency language modeling approach to information retrieval. The approach extends the basic KL-divergence retrieval approach by introducing the hybrid dependency structure, which includes syntactic dependency, syntactic proximity dependency and co-occurrence dependency, to describe dependencies between terms. Term and dependency language models are constructed for both document and query. The relevant between a document and a query is then evaluated by using the KL-divergence between their corresponding models. The new dependency retrieval model has been compared with other traditional retrieval models. Experiment results indicate that it produces significant improvements in retrieval effectiveness. [ABSTRACT FROM AUTHOR]

L. Breiman recently introduced the concept of random forests (randomly constructed collection of decision trees) for classification. We have modified the method for regression and applied it to language modeling for speech recognition. Random forests achieve excellent results in both perplexity and error rate. They can be regarded as a language model in HMM form and have interesting properties that achieve very robust smoothing. [ABSTRACT FROM AUTHOR]

In this paper we present a novel multiple-taxonomy question classification system, facing the challenge of assigning categories in multiple taxonomies to natural language questions. We applied our system to category search on faceted information. The system provides a natural language interface to faceted information, detecting the categories requested by the user and narrowing down the document search space to those documents pertaining to the facet values identified. The system was developed in the framework of language modeling, and the models to detect categories are inferred directly from the corpus of documents. [ABSTRACT FROM AUTHOR]

In the past, Maximum Entropy based language models were constrained by training data n-gram counts, topic estimates, and triggers. We will investigate the obtainable gains from imposing additional constraints related to linguistic clusters, such as parts of speech, semantic/syntactic word clusters, and semantic labels. It will be shown that there substantial profit is available provided the estimates use Gaussian a priori statistics. [ABSTRACT FROM AUTHOR]

In this paper we propose a new type of syllable-based unit for recognition and language model to improve recognition rate for Korean phones, syllables and characters. We propose ‘combined' units for which both Korean characters and syllable units realized in speech are taken into consideration. We can obtain character, syllable and phone sequences directly from the recognition results by using proposed units. To test the performance of the proposed approach we perform two types of experiments. First, we perform language modeling for phones, characters, syllables and propose combined units based on the same text corpus, and we test the performance for each unit. Second, we perform a vector space model based retrieval experiment by using the proposed combined units. [ABSTRACT FROM AUTHOR]

This paper presents the Cross Language Information Retrieval (CLIR) experiments of Language Technologies Research Centre (LTRC, IIIT-Hyderabad) as part of our participation in the ad-hoc track of CLEF 2006. This is our first participation in the CLEF evaluation campaign and we focused on Afaan Oromo, Hindi and Telugu as source (query) languages for retrieval of documents from English text collection. We have used a dictionary based approach for CLIR. After a brief description of our CLIR system we discuss the evaluation results of various experiments we conducted using CLEF 2006 dataset. [ABSTRACT FROM AUTHOR]

For the 2005 Cross-Language Evaluation Forum, Thomson Legal and Regulatory participated in the Hungarian, French, and Portuguese monolingual search tasks as well as French-to-Portuguese bilingual retrieval. Our Hungarian participation focused on comparing the effectiveness of different approaches toward morphological stemming. Our French and Portuguese monolingual efforts focused on different approaches to Pseudo-Relevance Feedback (PRF), in particular the evaluation of a scheme for selectively applying PRF only in the cases most likely to produce positive results. Our French-to-Portuguese bilingual effort applies our previous work in query translation to a new pair of languages and uses corpus-based language modeling to support term-by-term translation. We compare our approach to an off-the-self machine translation system that translates the query as a whole and find the latter approach to be more performant. All experiments were performed using our proprietary search engine. We remain encouraged by the overall success of our efforts, with our main submissions for each of the four tasks performing above the overall CLEF median. However, none of the specific enhancement techniques we attempted in this year's forum showed significant improvements over our initial result. [ABSTRACT FROM AUTHOR]

Recently, some machine learning techniques like support vector machines are employed for question classification. However, these techniques heavily depend on the availability of large amounts of training data, and may suffer many difficulties while facing various new questions from the real users on the Web. To mitigate the problem of lacking sufficient training data, in this paper, we present a simple learning method that explores Web search results to collect more training data automatically by a few seed terms (question answers). In addition, we propose a novel semantically related feature model (SRFM), which takes advantage of question focuses and their semantically related features learned from the larger number of collected training data to support the determination of question type. Our experimental results show that the proposed new learning method can obtain better classification performance than the bigram language modeling (LM) approach for the questions with untrained question focuses. Keywords: Question Answering, Question Classification, Web Search Results, Language Model, Question Focus, Semantically Related Feature. [ABSTRACT FROM AUTHOR]

This paper proposes a word sense language model based method for information retrieval. This method, differing from most of traditional ones, combines word senses defined in a thesaurus with a classic statistical model. The word sense language model regards the word sense as a form of linguistic knowledge, which is helpful in handling mismatch caused by synonym and data sparseness due to data limit. Experimental results based on TREC-Mandarin corpus show that this method gains 12.5% improvement on MAP over traditional tf-idf retrieval method but 5.82% decrease on MAP compared to a classic language model. A combination result of this method and the language model yields 8.92% and 7.93% increases over either respectively. We present analysis and discussions on the not-so-exciting results and conclude that a higher performance of word sense language model will owe to high accurate of word sense labeling. We believe that linguistic knowledge such as word sense of a thesaurus will help IR improve ultimately in many ways. [ABSTRACT FROM AUTHOR]

We report experiments on automatic essay grading using Latent Dirichlet Allocation (LDA). LDA is a "bag-of-words" type of language modeling and dimension reduction method, reported to outperform other related methods, Latent Semantic Analysis (LSA) and Probabilistic Latent Semantic Analysis (PLSA) in Information Retrieval (IR) domain. We introduce LDA in detail and compare its strengths and weaknesses to LSA and PLSA. We also compare empirically the performance of LDA to LSA and PLSA. The experiments were run with three essay sets consisting in total of 283 essays from different domains. On contrary to the findings in IR, LDA achieved slightly worse results compared to LSA and PLSA in the experiments. We state the reasons for LSA and PLSA outperforming LDA and indicate further research directions. [ABSTRACT FROM AUTHOR]

Information retrieval techniques have been used for a long time to identify links between textual items for the automatic construction of hypertexts and electronic books where sought information can be accessed by browsing. While research work in this area has been steadily decreasing in recent years, some of the techniques developed in that context are proving very valuable in a number of new application areas. In this paper we present an approach to automatic linking of textual items that is used to prioritise criminal suspects in a police investigation. A free-text description of an unsolved crime is compared to previous offence descriptions where the offender is known. By linking the descriptions, inferences about likely suspects can be made. Language Modeling is adapted to produce a Bayesian model which assigns a probability to each suspect. An empirical study showed that the linking of free text descriptions of burglaries enables prioritisation of offenders. The model presented in this paper could be easily extended to take account of additional crime and suspect linking data, such as geographical location of crimes or suspect social networks. This would enable large networks of investigative information automatically constructed from police archives to be browsed. [ABSTRACT FROM AUTHOR]

An expert finding is a very common task among enterprise search activities, while its usual retrieval performance is far from the quality of the Web search. Query modeling helps to improve traditional document retrieval, so we propose to apply it in a new setting. We adopt a general framework of language modeling for expert finding. We show how expert language models can be used for advanced query modeling. A preliminary experimental evaluation on TREC Enterprise Track 2006 collection shows that our method improves the retrieval precision on the expert finding task. [ABSTRACT FROM AUTHOR]

Measuring the similarity between documents and queries has been extensively studied in information retrieval. However, there are a growing number of tasks that require computing the similarity between two very short segments of text. These tasks include query reformulation, sponsored search, and image retrieval. Standard text similarity measures perform poorly on such tasks because of data sparseness and the lack of context. In this work, we study this problem from an information retrieval perspective, focusing on text representations and similarity measures. We examine a range of similarity measures, including purely lexical measures, stemming, and language modeling-based measures. We formally evaluate and analyze the methods on a query-query similarity task using 363,822 queries from a web search log. Our analysis provides insights into the strengths and weaknesses of each method, including important tradeoffs between effectiveness and efficiency. [ABSTRACT FROM AUTHOR]

The paper describes automatic speech recognition experiments and results on the spontaneous Hungarian MALACH speech corpus. A novel morph-based lexical modeling approach is compared to the traditional word-based one and to another, previously best performing morph-based one in terms of word and letter error rates. The applied language and acoustic modeling techniques are also detailed. Using unsupervised speaker adaptations along with morph based lexical models 14.4%-8.1% absolute word error rate reductions have been achieved on a 2 speakers, 2 hours test set as compared to the speaker independent baseline results. [ABSTRACT FROM AUTHOR]

We propose and evaluate a query expansion mechanism that supports searching and browsing in collections of annotated documents. Based on generative language models, our feedback mechanism uses document-level annotations to bias the generation of expansion terms and to generate browsing suggestions in the form of concepts selected from a controlled vocabulary (as typically used in digital library settings). We provide a detailed formalization of our feedback mechanism and evaluate its effectiveness using the TREC 2006 Genomics track test set. As to the retrieval effectiveness, we find a 20% improvement in mean average precision over a query-likelihood baseline, whilst increasing precision at 10. When we base the parameter estimation and feedback generation of our algorithm on a large corpus, we also find an improvement over state-of-the-art relevance models. The browsing suggestions are assessed along two dimensions: relevancy and specifity. We present an account of per-topic results, which helps understand for what type of queries our feedback mechanism is particularly helpful. [ABSTRACT FROM AUTHOR]

In this paper, we describe our studies of cross-language and cross-media image retrieval at the ImageCLEF 2005. This is the first participation of our CUHK (The Chinese University of Hong Kong) group at ImageCLEF. The task in which we participated is the "bilingual ad hoc retrieval" task. There are three major focuses and contributions in our participation. The first is the empirical evaluation of language models and smoothing strategies for cross-language image retrieval. The second is the evaluation of cross-media image retrieval, i.e., combining text and visual contents for image retrieval. The last is the evaluation of bilingual image retrieval between English and Chinese. We provide an empirical analysis of our experimental results, in which our approach achieves the best mean average precision result in the monolingual query task in the campaign. Finally we summarize our empirical experience and address the future improvement of our work. [ABSTRACT FROM AUTHOR]

We introduce a method for content-based advertisement selection for personal blog pages, based on combining multiple representations of the blog. The core idea behind the method is that personal blogs represent individuals, whose interests can be modeled by the language used in the blog itself combined with the language used in related sources of information, such as comments posted to a blog post or the blogger's community. An evaluation of our ad placement method shows improvement over state-of-the-art ad placement methods which were not designed for blog pages. [ABSTRACT FROM AUTHOR]

A new method for information retrieval which is on the basis of language model with relative entropy and feedback is presented in this paper. The method builds a query language model and document language models respectively for the query and the documents. We rank the documents according to the relative entropies of the estimated document language models with respect to the estimated query language model. The feedback documents are used to estimate a query model by the approach that we assume that the feedback documents are generated by a combined model in which one component is the feedback document language model and the other is the collection language model. Experimental results show that the method is effective for feedback documents and performs better than the basic language modeling approach. The results also indicate that the performance of the method is sensitive to both the smoothing parameters and the interpolation coefficients used to estimate the values of the language models. [ABSTRACT FROM AUTHOR]

This paper presents the results of the State University of New York at Buffalo (UB) in the Mono-lingual and Multi-lingual tasks at CLEF 2004. For these tasks we used an approach based on statistical language modeling. Our Adhoc retrieval work used the TAPIR toolkit developed in house by M Srikanth. Our approach focused on the validation and adaptation of the language model system to work in a multilingual environment and in exploring ways to merge results from multiple collections into a single list of results. We explored the use of a measure of query ambiguity, also known as clarity score, for merging results of the individual collections into a single list of retrieved documents. Our results indicate that the use of clarity scores normalized across queries gives statistically significant improvements over using a fixed merging order. [ABSTRACT FROM AUTHOR]

A database user session is a sequence of queries issued by a user (or an application) to achieve a certain task. Analysis of task-oriented database user sessions provides useful insight into the query behavior of database users. In this paper, we describe novel algorithms for identifying sessions from database traces and for grouping the sessions different classes. We also present experimental results. [ABSTRACT FROM AUTHOR]

This paper describes the approach of the University of Twente at its first participation in GeoCLEF. A large effort went into the construction of a geographic thesaurus which was utilized to add geographic knowledge to the documents and queries. Geographic filtering was applied to the results returned from a retrieval by content run. Employing such a geographic knowledge base however showed no added value - the content-only baseline outperformed all geographically filtered runs. [ABSTRACT FROM AUTHOR]

This paper presents the results of our experiments in the monolingual English, Spanish and Portuguese tasks and the Bilingual Spanish → English, Spanish → Portuguese, English → Spanish and Portuguese → Spanish tasks. We used the Terrier Information Retrieval Platform to run experiments for both tasks using the Inverse Document Frequency model with Laplace after-effect and normalization 2. Experiments included topics processed automatically as well as topics processed manually. Manual processing of topics was carried out using gazetteers. For the bilingual task we developed a component based on the transfer approach in machine translation. Topics were pre-processed automatically to eliminate stopwords. Then topics in the source language were translated to the target language. Evaluation of results show that the approach worked well in general, but applying more natural language processing techniques would improve retrieval performance. [ABSTRACT FROM AUTHOR]

We present a report on our participation in the mixed monolingual web task of the 2006 Cross-Language Evaluation Forum (CLEF). We compared the result of web page retrieval based on the page content, page title, and anchor page. The retrieval effectiveness for the combination of page content, page title, and anchor texts was better than that of the combination of page title and page title only. Applying the pseudo-relevance feedback improved the retrieval performance of the queries. [ABSTRACT FROM AUTHOR]

We report on the CLEF 2006 WebCLEF track devoted to crosslingual web retrieval. We provide details about the retrieval tasks, the used topic set, and the results of the participants. WebCLEF 2006 used a stream of known-item topics consisting of: (i) manual topics (including a selection of WebCLEF 2005 topics, and a set of new topics) and (ii) automatically generated topics (generated using two techniques). The results over all topics show that current CLIR systems are very effective, retrieving on average the target page in the top ranks. Manually constructed topics result in higher performance than and automatically generated ones. And finally, the resulting scores on automatic topics provide a reasonable ranking of the systems, showing that automatically generated topics are an attractive alternative in situations where manual topics are not readily available. [ABSTRACT FROM AUTHOR]

The paper describes the system built by the team from the University of West Bohemia for participation in the CLEF 2006 CL-SR track. We have decided to concentrate only on the monolingual searching in the Czech test collection and investigate the effect of proper language processing on the retrieval performance. We have employed the Czech morphological analyser and tagger for that purposes. For the actual search system, we have used the classical tf.idf approach with blind relevance feedback as implemented in the Lemur toolkit. The results indicate that a suitable linguistic preprocessing is indeed crucial for the Czech IR performance. [ABSTRACT FROM AUTHOR]

The use of visual features in text-based ad-hoc image retrieval is challenging. Visual and textual information have so far been treated equally in terms of their properties and combined with weighting mechanisms for balancing their contributions to the ranking. The use of visual and textual information in a single retrieval system sometimes limits its applicability due to the lack of modularity. In this paper, we propose an image retrieval method that separates the usage of visual information from that of textual information. Visual clustering establishes linkages between images and the relationships are later used for the re-ranking. By applying clustering on visual features prior to the ranking, the main retrieval process becomes purely text-based. Experimental results on the ImageCLEFphoto ad-hoc task show this scheme is suitable for querying multilingual collections on some search topics. [ABSTRACT FROM AUTHOR]

The paper describes QUANTICO, a cross-language open domain question answering system for German and English. The main features of the system are: use of preemptive off-line document annotation with syntactic information like chunk structures, apposition constructions and abbreviation-extension pairs for the passage retrieval; use of online translation services, language models and alignment methods for the cross-language scenarios; use of redundancy as an indicator of good answer candidates; selection of the best answers based on distance metrics defined over graph representations. Based on the question type two different strategies of answer extraction are triggered: for factoid questions answers are extracted from best IR-matched passages and selected by their redundancy and distance to the question keywords; for definition questions answers are considered to be the most redundant normalized linguistic structures with explanatory role (i.e., appositions, abbreviation's extensions). The results of evaluating the system's performance by CLEF were as follows: for the best German-German run we achieved an overall accuracy (ACC) of 42.33% and a mean reciprocal rank (MRR) of 0.45; for the best English-German run 32.98% (ACC) and 0.35 (MRR); for the German-English run 17.89% (ACC) and 0.17 (MRR). [ABSTRACT FROM AUTHOR]

Feature engineering is known as one of the most important challenges for knowledge acquisition, since any inductive learning system depends upon an efficient representation model to find good solutions to a given problem. We present an NLP-driven constructive learning method for building features based upon noun phrases structures, which are supposed to carry the highest discriminatory information. The method was test at the CLEF 2006 Ad-Hoc, monolingual (Portuguese) IR track. A classification model was obtained using this representation scheme over a small subset of the relevance judgments to filter false-positives documents returned by the IR-system. The goal was to increase the overall precision. The experiment achieved a MAP gain of 41.3%, in average, over three selected topics. The best F1-measure for the text classification task over the proposed text representation model was 77.1%. The results suggest that relevant linguistic features can be exploited by NLP techniques in a domain specific application, and can be used suscesfully in text categorization, which can act as an important coadjuvant process for other high-level IR tasks. [ABSTRACT FROM AUTHOR]

In this paper we compare n-grams and morphological normalization, two inherently different text-preprocessing methods, used for text classification on a Croatian-English parallel corpus. Our approach to comparing different text preprocessing techniques is based on measuring computational performance (execution time and memory consumption), as well as classification performance. We show that although n-grams achieve classifier performance comparable to traditional word-based feature extraction and can act as a substitute for morphological normalization, they are computationally much more demanding. [ABSTRACT FROM AUTHOR]

Our submission to the INEX 2006 Ad-hoc retrieval track is described. We study how to utilize the Wikipedia structure (XML documents with hyperlinks) by combining XML and Web retrieval. In particular, we experiment with different combinations of language models and the HITS algorithm. An important feature of our techniques is a filtering phase that identifies the relevant part of the corpus, prior to the processing of the actual XML elements. We analyze the effect of the above techniques based on the results of our runs in INEX 2006. [ABSTRACT FROM AUTHOR]

This paper describes the retrieval approaches used by Queen Mary, University of London in the INEX 2006 ad hoc track. In our participation, we mainly investigate element-specific smoothing method within the language modelling framework. We adjust the amount of smoothing required for each XML element depending on its number of topic shifts to provide a focused access to XML elements in the Wikipedia collection. We also investigate whether using non-uniform priors is beneficial for the ad hoc tasks. [ABSTRACT FROM AUTHOR]

In this paper we present an extension of information retrieval based on Kullback-Leibler divergence (with backoff smoothing) to support structured queries on structured documents. The proposed method applies to several common retrieval tasks characterized by an implication relationship among texts, including fielded topics and XML documents. We discuss how to choose the method parameters to make the computation of the ranking function efficient. We finally report some experimental results obtained using a loose approximation of the model based on a discriminative selection strategy. [ABSTRACT FROM AUTHOR]

Brain-Computer Interfaces (BCIs) are systems capable of decoding neural activity in real time, thereby allowing a computer application to be directly controlled by the brain. Since the characteristics of such direct brain-to-computer interaction are limited in several aspects, one major challenge in BCI research is intelligent front-end design. Here we present the mental text entry application ‘Hex-o-Spell' which incorporates principles of Human-Computer Interaction research into BCI feedback design. The system utilises the high visual display bandwidth to help compensate for the extremely limited control bandwidth which operates with only two mental states, where the timing of the state changes encodes most of the information. The display is visually appealing, and control is robust. The effectiveness and robustness of the interface was demonstrated at the CeBIT 2006 (world's largest IT fair) where two subjects operated the mental text entry system at a speed of up to 7.6 char/min. [ABSTRACT FROM AUTHOR]

Service discovery is a fundamental concept underpinning the move towards dynamic service-oriented business partnerships. The business process for integrating service discovery and underlying registry technologies into busi ness relationships, procurement and project management functions has not been examined and hence existing Web Service registries lack capabilities required by business today. In this paper we present a novel contextualized B2B registry that supports dynamic registration and discovery of resources within manage ment contexts to ensure that the search space is constrained to the scope of au thorized and legitimate resources only. We describe how the registry has been deployed in three case studies from important economic sectors (aerospace, automotive, pharmaceutical) showing how contextualized discovery can sup port distributed product development processes. [ABSTRACT FROM AUTHOR]

Traditional text classification studied in the information retrieval and machine learning literature is mainly based on topics. That is, each class represents a particular topic, e.g., sports and politics. However, many real-world problems require more refined classification based on some semantic perspectives. For example, in a set of sentences about a disease, some may report outbreaks of the disease, some may describe how to cure the disease, and yet some may discuss how to prevent the disease. To classify sentences at this semantic level, the traditional bag-of-words model is no longer sufficient. In this paper, we study semantic sentence classification of disease reporting. We show that both keywords and sentence semantic features are useful. Our results demonstrated that this integrated approach is highly effective. [ABSTRACT FROM AUTHOR]

This paper examines genre classification of documents and its role in enabling the effective automated management of digital documents by digital libraries and other repositories. We have previously presented genre classification as a valuable step toward achieving automated extraction of descriptive metadata for digital material. Here, we present results from experiments using human labellers, conducted to assist in genre characterisation and the prediction of obstacles which need to be overcome by an automated system, and to contribute to the process of creating a solid testbed corpus for extending automated genre classification and testing metadata extraction tools across genres. We also describe the performance of two classifiers based on image and stylistic modeling features in labelling the data resulting from the agreement of three human labellers across fifteen genre classes. [ABSTRACT FROM AUTHOR]

Multiplicative update rules have proven useful in many areas of machine learning. Simple to implement, guaranteed to converge, they account in part for the widespread popularity of algorithms such as nonnegative matrix factorization and Expectation-Maximization. In this paper, we show how to derive multiplicative updates for problems in L1-regularized linear and logistic regression. For L1-regularized linear regression, the updates are derived by reformulating the required optimization as a problem in nonnegative quadratic programming (NQP). The dual of this problem, itself an instance of NQP, can also be solved using multiplicative updates; moreover, the observed duality gap can be used to bound the error of intermediate solutions. For L1-regularized logistic regression, we derive similar updates using an iteratively reweighted least squares approach. We present illustrative experimental results and describe efficient implementations for large-scale problems of interest (e.g., with tens of thousands of examples and over one million features). [ABSTRACT FROM AUTHOR]

We present an analysis of word use and sentence length in different types of Dutch speech, ranging from conversations over discussions and formal speech to read speech. We find that the distributions of sentence length and personal pronouns are characteristic for the type of speech. In addition, we analyzed differences in word use between male and female speakers and between speakers with high and low education levels. We find that male speaker use more fillers, while women use more pronouns and adverbs. Furthermore, gender specific differences turn out to be stronger than differences in language use between groups with different education levels. [ABSTRACT FROM AUTHOR]

Speech synthesis techniques have already reached a high level of naturalness. However, they are often evaluated on text reading tasks. New applications will request for conversational speech instead and disfluencies are crucial in such a style. The present paper presents a system to predict filled pauses and synthesise them. Objective results show that they can be inserted with 96% precision and 58% recall. Perceptual results even shown that its insertion increases naturalness of synthetic speech. [ABSTRACT FROM AUTHOR]

This paper reports on the evaluation of a system that allows the use of spoken queries to retrieve information from a textual document collection. First, a large vocabulary continuous speech recognizer transcribes the spoken query into text. Then, an information retrieval engine retrieves the documents relevant to that query. The system works for Spanish language. In order to increase performance, we proposed a two-pass approach based on dynamic adaptation of language models. The system was evaluated using a standard IR test suite from CLEF. Spoken queries were recorded by 10 different speakers. Results showed that the proposed approach outperforms the baseline system: a relative gain in retrieval precision of 5.74%, with a language model of 60,000 words. [ABSTRACT FROM AUTHOR]

In his LREC 2004 invited talk when awarded by the first ever Antonio Zampolli prize for his essential contributions to the use of spoken and written language resources, Frederick Jelinek has used the title "Some of My Best Friends Are Linguists". He did so for many reasons, one of them being that he wanted to remove the perception that he dislikes linguists and linguistics after so many people used to cite his famous line from an old presentation at a Natural Language Processing Evaluation workshop in 1988, in which he said "Whenever I fire a linguist our system performance improves." [ABSTRACT FROM AUTHOR]

This paper proposes a unified extractive approach based on affinity graph to both generic and topic-focused multi-document summarizations. By using an asymmetric similarity measure, the relationships between sentences are reflected in a directed affinity graph for generic summarization. For topic-focused summarization, the topic information is incorporated into the affinity graph using a topic-sensitive affinity measure. Based on the affinity graph, the information richness of sentences is computed by the graph-ranking algorithm on differentiated intra-document links and inter-document links between sentences. Lastly, the greedy algorithm is employed to impose diversity penalty on sentences and the sentences with both high information richness and high information novelty are chosen into the summary. Experimental results on the tasks of DUC 2002-2005 demonstrate the excellent performances of the proposed approaches to both generic and topic-focused multi-document summarization tasks. [ABSTRACT FROM AUTHOR]

With the exponential growth of the available information on the World Wide Web, a traditional search engine, even if based on sophisticated document indexing algorithms, has difficulty meeting efficiency and effectiveness performance demanded by users searching for relevant information. Users surfing the Web in search of resources to satisfy their information needs have less and less time and patience to formulate queries, wait for the results and sift through them. Consequently, it is vital in many applications - for example in an e-commerce Web site or in a scientific one - for the search system to find the right information very quickly. Personalized Web environments that build models of short-term and long-term user needs based on user actions, browsed documents or past queries are playing an increasingly crucial role: they form a winning combination, able to satisfy the user better than unpersonalized search engines based on traditional Information Retrieval (IR) techniques. Several important user personalization approaches and techniques developed for the Web search domain are illustrated in this chapter, along with examples of real systems currently being used on the Internet. [ABSTRACT FROM AUTHOR]