Theory and practice of ambiguity labelling with a view to interactive disambiguation in text and speech MT

Theory and practice of ambiguity labelling with a view to interactive disambiguation in text and speech MT Christian Boitet GETA, CLIPS, IMAG (UJF & CNRS), 150 rue de la Chimie, BP 53 38041 Grenoble Cedex 9, France Christian. BoitetOimag. fr Abstract In many contexts, automatic analyzers cannot fully disambiguate a sentence or an utterance reliably, but can produce ambiguous results containing the correct interpretation. It is useful to study vatious properties of these ambiguities in the view of subsequent total or partial inter- active disambiguation. We have proposed a technique for labelling ambiguities in texts and in dialogue transcriptions, and experimented it on multilingual data. It has been first necessary to define formally the very notion of ambiguity relative to a representation system, as well as associated concepts such as ambiguity kernel, ambiguity scope, ambiguity occurrence. Keywords: interactive disambiguation, ambiguity labelling, ambiguity occurrence, ambiguity kernel Introduction We are interested in improving the quality of MT systems for monolinguals, where the input can be text or speech, no revision is possible, and the controlled language approach is not usable. In such contexts, the automatic analyzer cannot fully and reliably disambi- guate a sentence or an utterance, and the best available heuristics don't select the correct results often enough. Complete or partial interactive disambiguation, folio- wing a best possible automatic disambiguation, is an attractive way to raise quality and reliability. To develop good strategies for interactive disambi- guation, it is useful to study vatious properties of the ambiguities unsolvable by state of the art analyzers. To conduct such studies, it is necessary to gather data, that is, to perform "ambiguity labelling" on texts and transcriptions of spoken dialogues. Our motiva- tions and goals are explained in more detail in the first part. As the usual notion of ambiguity is too vague for our purpose, it is necessary to refine it. This is done in the second part, where we define formally the notion of ambiguity relative to a representation sys- tem, as well as associated concepts such as kernel, scope, occurrence and type of ambiguity. In the third part, we propose a format for ambiguity labelling, and illustrate it examples from a transcribed dialogue. This format is independent of the exact kind of output produced by any implemented analyzer, essentially because ambiguities are described with a view to generate human-oriented questions. We have experimented our technique on various kinds of dialogues and on some texts in several languages. In some cases, analysis results produced by automatic Mutsuko Tomokiyo ATR Interpreting Telecommunications Research Labs 2-2 Hikari-dai, Seika-cho, Soraku-gun Kyoto 619-02, Japan tomokiyo@itl, atr. co. jp analyzers were available, in others not. It is interesting to compare the intuition of the human labeller with results actually produced: most of the time, differences may be attributed to the fact that available analyzers don't yet match our expectations for "state of the art" analyzers, because they produce spurious, "parasite" ambiguities, and don't yet implement all types of sure 1 linguistic constraints. Motivations and Goals Interactive disambiguation …

Content truncated to 500 words