A Framework for Processing Corrections in Task-Oriented Dialogues

نویسندگان

  • Philip J. Hayes
  • Jaime G. Carbonell
چکیده

Mundane discourse abounds with utterances referring to other utterances. These meta-language utterances appear with surprising frequency in task-oriented dialogues, such as those arising in the context of a natural language interface to an operating system. This paper identifies some simpler types of dialogue-level meta language utterance and provides a computational framework to process such phrases in the context of a case-frame parser exploiting strongly-typed domain semantics. 1 . I n t r o d u c t i o n ivteta-language utterances, i.e. utterances about other utterances in the discourse context, have received little attention in the computational linguistic literature. Yet, as a recent attempt [1] to create a taxonomy of discourse-level metal inguist ic devices shows, such utterances are ubiquitous in all kinds of dialogues, including task-oriented man-machine dialogues through a natural language interface. Indeed, in a recent informal experiment with a simulated natural language command interface to a computer operating system, we found that one out of every twelve inputs was a meta language utterance of one type or another. Meta language utterances fall into two primary categories, intra sentential and extra sentential (dialogue level) utterances. The former class may be divided into referential and interpretive meta-linguistic devices. For instance: "Load, read and rewind the tape in tha t o rde r . " is a typical referential intra-sentential meta language utterance of the kind that has received some attention in the linguistics literature [8, 6]. "In that order" refers to the surface structure sequencing of the lexical items denoting the commands. Interpretive meta language utterances make a statement about how another portion of that utterance should be interpreted, for instance: " M e t a p h o r i c a l l y s p e a k i n g , John's ideas are out of this wor ld. " Discourse-level metal inguist ic utterances refer to utterances or portions of utterances outside the scope of the immediate sentence. For instance, the third utterance in the dialogue fragment below is essentially a correction of a previous utterance. USER: Print gauss.rel on the diablo printer. SYSTEM: GAUSS.REL is not a printable file. USER: Oops, I meant gauss.for This paper focuses on developing a process model for certain classes of dialogue-level meta language utterances, which occur in task-oriented dialogues with greater frequency than intrasentential ones. With the exception of the analysis and classification of dialogue-level metal inguist ic utterances i n [1 ] , there is little prior analysis of this phenomenon, although work in discourse focus [2, 10] and indirect speech acts [9, 7] is of direct relevance in analyzing the general phenomenon. We currently have no general process-level theory for metalanguage comprehension, but we have developed a framework for dealing with an important common subset of discourse-level meta language utterances: those that repair errors or misunderstandings that arose earlier in the dialogue, such as "I meant to say transfer, not copy" ,

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Modelsaz: An Object-Oriented Computer-Aided Modeling Environment

Modeling and simulation of processing plants are widely used in industry. Construction of a mathematical model for a plant is a time-consuming and error-prone task. In light of extensive advancements in computer science (both hardware and software), computers are becoming a necessary instrument in industrial activities. Many software tools for modeling, simulation and optimization of proces...

متن کامل

The Involvement Load Hypothesis and Vocabulary Learning: The Effect of Task Types and Involvement Index on L2 Vocabulary Acquisition

This study builds on Laufer and Hulstijn’s (2001) motivational-cognitive construct of task-induced involvement in learning vocabulary and addresses itself to its strong claim that the depth of processing is the overriding factor in learning words. The paper first re-examines the effect of processing load and then of task type on the initial learning and retention of words. To do so, 60 EFL lear...

متن کامل

Logical Form of Complex Sentences in Task-Oriented Dialogues

Introduction Although most NLP researchers agree that a level of "logical form" is a necessary step toward the goal of representing the meaning of a sentence, few people agree on the content and form of this level of representation. An even smaller number of people have considered the complex action sentences that are often expressed in taskoriented dialogues. Most existing logical form represe...

متن کامل

سوگیری حافظه ناآشکار و آشکار در افراد افسرده بر اساس پردازش انتقال مناسب

The aim of the present research was to examine implicit and explicit memory bias in depressed individuals based on the Transfer Appropriate Processing (TAP) framework. For this purpose, 60 participants (30 outpatient depressed participants for the experimental group and 30 non-depressed participants for the control group) were selected as research sample based on psychiatric interviews of DSM-I...

متن کامل

Syntax, Semantics, and Pragmatics in Concert: An Incremental, Multilevel Approach in Reconstructing Task-Oriented Dialogues

This paper gives an overview of a model for the reconstruction of task-oriented dialogues based on an in teract ive , mult i level parsing formalism. It is applied to route description dialogues. It w i l l be shown, how the pragmatic aspects of such dialogues are taken into account on d i f ferent levels of processing. The approach described is based on an extension of the concept of cascaded ...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1983