[Japanese | Thesis | Researches in Minoh Lab | Minoh Lab]
It is dangerous for a cooker to cook while checking a recipe book because he/she has to take his/her eyes off hand. We propose the system that teaches what the cooker should do next according to the recipe only in a voice. The system gives voice navigation consisted of ”alias of process” and “alias of food product” for each cooking procedure by speech synthesis engine. The system can assign the expression used in the recipe book such as "Cut" as the alias of process. But for the alias of food product, the system cannot find its expression in the recipe book sometimes because the recipe book can specify a food product by referring to the previous procedures such as “(1) Cut an apple in small peaces. (2) Mix (1) with yoghurt.” Moreover, the recipe book can specify several food products by a single symbol as “A: soy, sake, and sugar. …Mix A”. When the cooker reads the recipe book, he/she can refer to the previous procedure and the list of the food materials whenever he/she wants. But when he/she cooks according to the system’s voice instruction, he/she cannot memorize all of the instruction completely. The system has to assign a suitable alias for each food product based on how the cooker understands the food product. Otherwise he/she cannot specify the food product to be cooked surely.
At the first, we analyzed cooking programs and recipe books, and found that food products can be called by two types of aliases. One is “Recent Process Name” that refers a food product by the name of most recent process on the food like “the mixed one” or “the cut one”. Another is “Composition Name” that refers a food product by the combination of material names that composed it like “the mixture of an apple and yoghurt”. However, not always these two types of aliases can specify only one food product. Recent Process Name cannot specify the food product when other food products that added the same process recently exist at the same time. Composition Name cannot specify the food product when other products that contain the same material exist at the same time. In addition, when both two types of aliases can specify the food product, we should choose one of them depend on the cooker’s preference.
To define which alias is preferred for each food product in such situations that either or both types of aliases can specify the food product, we experimented as follows. We constructed the cooking simulator that subjects can cook on personal computer according to voice instructions. The subjects cooked on this simulator and answered the questionnaire whether they can specify the food product to be cooked by each alias and the alias is comprehensible or not.
The subjects were 18 housewives in their thirties to seventies. We prepared two kinds of recipes that one of two types of aliases was chosen at random for each instruction. As a result, 61.1％ subjects preferred same type of alias for food product composed of one material in average on both recipes, and only 5.6％ subjects preferred different type of alias on one recipe from another. Moreover, 55.6％ subjects preferred same type of alias for food products composed of several materials on both recipes, and only 16.7％ subjects preferred different kind of alias on one recipe from another. It means that the favor of the individual exists and is consistent regardless of recipe. Composition Name was preferred as alias of food product which cannot be specified by Recent Process Name, but the evaluation of Composition Name came down as the number of component materials increased, and the evaluation of Composition Name came down to the same level with Recent Process Name when the number of component materials came to eight. For food product in such a situation that Composition Name cannot specify it, when subjects had cooked the same food product at the previous procedure, Recent Process Name was preferred. On the other hand, when subjects had cooked other food product at the previous procedure, the evaluations were changed on individual.
In conclusion, we clarified the criteria to decide comprehensible alias for each food product depending on individual cooker for the system that gives cooking instruction in a voice. As a future work, we will construct the system that generates voice cooking instruction automatically based on the criteria.