Nec Corporation

Kaichiro Hatazaki

Takao Watanabe

Shinichi Ando

Shinichiro Kamei

Takahiro Okonogi

Noriko Watanabe

Hiroto Teranishi

It is possible to efficiently evaluate utterance while ensuring the effectiveness of the result of the evaluation. A selection unitselects recognition data, which is used in agent evaluation from a data storage unit, from recognition data generated by a voice recognition uniton the basis of voice data of utterance by applying a selection operation corresponding to evaluation viewpoint information with reference to an evaluation viewpoint management tablethrough an evaluation viewpoint management unit. An evaluation unitdisplays recognition data and plays the voice data by an evaluation viewpoint management unitby applying an output operation corresponding to evaluation viewpoint information with reference to an evaluation viewpoint management table

Monitoring device, evaluation data selecting device, agent evaluation device, agent evaluation system, and program

A speech dialogues system comprises a voice analyzing portion converting a voice input signal into a time series feature vector, a voice detecting portion detecting the leading end and the trailing end of a voice period in the voice input signal on the basis of the time series feature vector, a recognition and understanding portion translating the time series feature vector into a meaning expression per each voice period, an input dividing portion dividing the meaning expression string into partial strings respectively aggregated in meaning and outputting each partial strings, and an application executing portion for performing corresponding application process for each partial string of the meaning expressions.

Speech dialogue system in which a recognition and understanding process,

The input pattern to a pattern recognition system is defined as a candidate graph composed of several possibly valid category strings. The most valid category string is selected as the input pattern. Each category string is composed of linked category segments, each category segment being bounded by category nodes or branching points to which are connected one or more category segments. The most valid category string is developed by establishing the most valid midway path from the start node and extending to the terminal node. A most valid midway path to each midway node (n) is established by using the minimum valid estimation value determined for all possible midway paths from the start node to the midway node (n).

Growing community of inventors

Tokyo, Japan

Kaichiro Hatazaki