By Antal van den Bosch
This ebook is the results of a gaggle of researchers from diverse disciplines asking themselves one query: what does it take to advance a working laptop or computer interface that listens, talks, and will resolution questions in a website? First, evidently, it takes really expert modules for speech attractiveness and synthesis, human interplay administration (dialogue, enter fusion, andmultimodal output fusion), uncomplicated query realizing, and solution discovering. whereas all modules are researched as self reliant subfields, this booklet describes the advance of cutting-edge modules and their integration right into a unmarried, operating software able to answering clinical (encyclopedic) questions similar to "How lengthy is anyone with measles contagious?" or "How am i able to hinder RSI?".
The contributions during this booklet, which grew out of the IMIX undertaking funded via the Netherlands supplier for medical examine, record the advance of the program, but in addition tackle extra basic matters in common language processing, akin to the improvement of multidimensional discussion structures, the purchase of taxonomic wisdom from textual content, resolution fusion, series processing for domain-specific entity acceptance, and syntactic parsing for query answering. jointly, they give an outline of crucial findings and classes realized within the scope of the IMIX venture, making the e-book of curiosity to either educational and advertisement builders of human-machine interplay structures in Dutch or the other language.
Highlights contain: integrating multi-modal enter fusion in discussion administration (Van Schooten and Op den Akker), cutting-edge ways to the extraction of time period editions (Van der Plas, Tiedemann, and Fahmi; Tjong Kim Sang, Hofmann, and De Rijke), and multi-modal resolution fusion (two chapters by way of Van Hooijdonk, Bosma, Krahmer, Maes, Theune, and Marsi).
Watch the IMIX motion picture at www.nwo.nl/imix-film .
Like IBM's Watson, the IMIX method defined within the publication provides clearly phrased responses to obviously posed questions. the place Watson can in basic terms generate man made speech, the IMIX method additionally acknowledges speech. nevertheless, Watson is ready to win a tv quiz, whereas the IMIX method is domain-specific, answering purely to scientific questions.
"The Netherlands has constantly been one of many leaders within the common box of Human Language expertise, and IMIX isn't any exception. It used to be a truly bold software, with a remarkably winning functionality resulting in fascinating effects. The groups coated a outstanding quantity of territory within the common sphere of multimodal query answering and knowledge supply, query answering, info extraction and part technologies."
Eduard Hovy, USC, united states, Jon Oberlander, collage of Edinburgh, Scotland, and Norbert Reithinger, DFKI, Germany
Read Online or Download Interactive Multi-modal Question-Answering PDF
Similar user experience & usability books
This ebook is the 5th legit archival book dedicated to RoboCup. It records the achievements provided on the fifth robotic global Cup football video games and meetings held in Seattle, Washington, united states, in August 2001. The publication comprises the subsequent elements: creation, champion groups, problem award finalists, technical papers, poster displays, and crew descriptions (arranged in line with a variety of leagues).
This publication constitutes the refereed complaints of the fee 2102 overseas convention on Verbal and Nonverbal positive factors of Human-Human and Human-Machine interplay, held in Patras, Greece, October 29 -31, 2007. The 21 revised complete papers have been conscientiously reviewed and chosen. The papers are geared up in topical sections on static and dynamic processing of faces, facial expressions and gaze in addition to emotional speech synthesis and popularity.
Such a lot programmers' worry of consumer interface (UI) programming comes from their worry of doing UI layout. they suspect that UI layout is like image design—the mysterious procedure during which artistic, latte-drinking, all-black-wearing humans produce cool-looking, inventive items. so much programmers see themselves as analytic, logical thinkers instead—strong at reasoning, susceptible on inventive judgment, and incapable of doing UI layout.
The two-volume complaints of the ACIIDS 2015 convention, LNAI 9011 + 9012, constitutes the refereed lawsuits of the seventh Asian convention on clever info and Database structures, held in Bali, Indonesia, in March 2015. the entire of 117 complete papers authorized for book in those complaints was once conscientiously reviewed and chosen from 332 submissions.
Extra resources for Interactive Multi-modal Question-Answering
This is also the case with the individual modules. Most collaboration between subprojects was not related to the Demonstrator, but to the use of supplementary resources, such as NLP tools (for example the use of the Alpino parser) and shared data, such as the collections of annotated questions and annotated pictures. All in all, the conclusion is that the Demonstrator itself did not work as a research tool. So, was the Demonstrator a good idea? It served as a collaboration tool, but there may be more efficient methods.
For each FQ, one of a small set of relatively basic rewriting strategies was chosen. The most successful strategy by far, for both unimodal and multimodal FQ, was found to be the anaphora substitution strategy. Here, an anaphor in the sentence has to be located and identified as being substitutable. In practice, it was found that only a fraction of typical FQ can be rewritten using this method, due to the lack of proper rewriting methods to cover the entire range of FQ satisfactorily. Performance is measured by assuming that the first step was done correctly, and by counting the percentage of correct (intelligible and syntactically correct) sentences.
Table 3 summarises existing systems in terms of these steps, and give performance figures where available. It has to be concluded that the performances are hard to compare. Not only do the systems have different languages and domains, they also use different performance criteria and corpora. It was found that the same system tested on different corpora can give quite different results. It was also found that no distinction was made anywhere between harmful and insufficient. Vidiam: Dialogue Manager for Multimodal Question Answering 35 3 The Corpora This section describes the three corpora collected and analysed.