# HG changeset patch # User Yasutaka Higa # Date 1434635452 -32400 # Node ID c93a37ff6c793cac8d5893fa94e31c591601cbb5 # Parent 45f5a93790dbec839523717c5e5c6ac9bde00629 Wrote slides to section 2 diff -r 45f5a93790db -r c93a37ff6c79 slide.md --- a/slide.md Thu Jun 18 16:06:36 2015 +0900 +++ b/slide.md Thu Jun 18 22:50:52 2015 +0900 @@ -87,7 +87,7 @@ * Only two individuals (a robot and a human participant): we do not take in consideration a higher number of individuals. * Eye contact is taken for granted. * Age is considered part of 'power relationship' -* Regionality is not considered. +* Regionally is not considered. * Setting is not considered # Model of Greetings: Assumptions (6 - 10) @@ -97,6 +97,67 @@ * Time since the last interaction is partially included in 'social distance' * Intimacy and politeness are not necessary +# Model of Greetings: Basis of classification +* Input + * All the other factors are then considered features of a mapping problem + * They are categorical data, as they can assume only two or three values. +* Output + * The outputs can also assume only a limited set of categorical values. + +# Model of Greetings: Features, mapping discriminants, classes, and possible status +TODO: FIGURE 2 + +# Model of Greetings: Overview of the greeting model +* Greeting model takes context data as input and produces the appropriate robot posture and speech for that input. +* The two outputs evaluated by the participants of the experiment through written questionnaires. +* These training data that we get from the experience are given as feedback to the two mappings. + +# Greeting selection system training data +* Mappings can be trained to an initial state with data taken from the literature of sociology studies. +* Training data should be classified through some machine learning method or formula. +* We decided to use conditional probabilities: in particular the Naive Bayes formula to map data. +* Naive Bayes only requires a small amount of training data. + +# Model of Greetings: Details of training data +* While training data of gestures can be obtained from the literature, data of words can also be obtained from text corpora. +* English: English corpora, such as British National Corpus, or the Corpus of Historical American English, are used. +* Japanese: extracted from data sets by [24, 37, 41-43]. Analyze Corpus on Japanese is difficult. + +# Model of Greetings: Location Assumption +* The location of the experiment was Germany. +* For this reason, the only dataset needed was the Japanese. +* As stated in the motivations at the beginning of this paper, the robot should initially behave like a foreigner. +* ARMAR-IIIb, trained with Japanese data, will have to interact with German people and adapt to their customs. + +# Model of Greetings: Mappings and questionnaires +* The mapping is represented by a dataset, initially built from training data, as a table containing weights for each context vector corresponding to each greeting type. +* We now need to update these weights. + +# feedback from three questionnaires +* Whenever a new feature vector is given as an input, it is checked to see whether it is already contained in the dataset or not. +* In the former case, the weights are directly read from the dataset +* in the latter case, they get assigned the values of probabilities calculated through the Naive Bayes classifier. +* The output is the chosen greeting, after which the interaction will be evaluated through a questionnaires. + +# Model of Greetings: Three questionnaires for feedback +* answers of questionnaires are five-point semantic differential scale: +1. How appropriate was the greeting chosen by the robot for the current context? +2. (If the evaluation at point 1 was <= 3) which greeting type would have been appropriate instead? +3. (If the evaluation at point 1 was <= 3) which context would have been appropriate, if any, for the greeting type of point 1? + +# Model of Greetings: feedback and terminate condition +* Weights of the affected features are multiplied by a positive or negative reward (inspired by reinforcement learning) which is calculated proportionally to the evaluation. +* Mappings stop evolving when the following two stopping conditions are satisfied +* all possible values of all features have been explored +* and the moving average of the latest 10 state transitions has decreased below a certain threshold. + +# Model of Greetings: Summary +* Thanks to this implementation, mappings can evolve quickly, without requiring hundreds or thousands of iterations +* but rather a number comparable to the low number of interactions humans need to understand and adapt to social rules. + +# TODO: Please Add slides over chapter (3. implementation of ARMAR-IIIb) + +