Two studies investigated the effects of various motor constraints of reaching on infants’ search performance on the A-not-B task. These studies were motivated by the idea that motor memories for reaching lead to A-not-B errors. The 2 motor constraints that were evaluated included barriers that blocked the path of the hand and hand-use preferences. Each of these motor constraints was examined separately. In Experiment 1, infants (N = 40, 20 8-month olds, 20 16-month olds) were given the A-not-B task twice. One condition was analogous to the traditional A-not-B task (i.e., using 2 hiding locations) and the other was modified such that a barrier (i.e., an opaque screen) blocked the infants’ reaching path of location A on A trials only. On A trials, all infants searched correctly less often when a barrier was present, and younger infants searched correctly less often than older infants. On B trials, younger infants made more errors in the no barrier condition, whereas older infants did not show any significant difference in B trial performance across conditions. In Experiment 2, infants (N = 51) completed an adapted handedness test (Michel, Ovrut, & Harkins, 1985) followed by a modified A-not-B task. The test assessed infants’ hand-use preferences for reaching, which was used to group infants into their respective preference group (i.e., consistent or inconsistent). Infants with a consistent preference were randomly assigned to a hiding side group (i.e., A on preferred side or A on non-preferred side). Infants searched correctly more often when hiding side was congruent with their preferred reaching hand, and older infants searched correctly more often than younger infants. On the B trial, neither age nor hiding side affected the production of the A-not-B error. Collectively, these studies present data that address the theory that motor memories for reaching are the cause for the production of A-not-B error. These studies provide novel evidence that motor memories for reaching are present in infants aged 8- and 16-months, and that motor memories can influence the production of such errors in certain A-not-B contexts. Implications and directions for future research are also discussed.

Two studies investigated the effects of various motor constraints of reaching on infants’ search performance on the A-not-B task. These studies were motivated by the idea that motor memories for reaching lead to A-not-B errors. The 2 motor constraints that were evaluated included barriers that blocked the path of the hand and hand-use preferences. Each of these motor constraints was examined separately. In Experiment 1, infants (N = 40, 20 8-month olds, 20 16-month olds) were given the A-not-B task twice. One condition was analogous to the traditional A-not-B task (i.e., using 2 hiding locations) and the other was modified such that a barrier (i.e., an opaque screen) blocked the infants’ reaching path of location A on A trials only. On A trials, all infants searched correctly less often when a barrier was present, and younger infants searched correctly less often than older infants. On B trials, younger infants made more errors in the no barrier condition, whereas older infants did not show any significant difference in B trial performance across conditions. In Experiment 2, infants (N = 51) completed an adapted handedness test (Michel, Ovrut, & Harkins, 1985) followed by a modified A-not-B task. The test assessed infants’ hand-use preferences for reaching, which was used to group infants into their respective preference group (i.e., consistent or inconsistent). Infants with a consistent preference were randomly assigned to a hiding side group (i.e., A on preferred side or A on non-preferred side). Infants searched correctly more often when hiding side was congruent with their preferred reaching hand, and older infants searched correctly more often than younger infants. On the B trial, neither age nor hiding side affected the production of the A-not-B error. Collectively, these studies present data that address the theory that motor memories for reaching are the cause for the production of A-not-B error. These studies provide novel evidence that motor memories for reaching are present in infants aged 8- and 16-months, and that motor memories can influence the production of such errors in certain A-not-B contexts. Implications and directions for future research are also discussed.

Páginas-objeto são páginas que representam exatamente um objeto inerente do mundo real na web, considerando um domínio específico, e a busca por essas páginas é chamada de busca-objeto. Os motores de busca convencionais (do Inglês, General Search Engine - GSE) conseguem responder, de forma satisfatória, à maioria das consultas realizadas na web atualmente, porém, isso dificilmente ocorre no caso de buscas-objeto, uma vez que, em geral, a quantidade de páginas-objeto recuperadas é bastante limitada. Essa dissertação propõe um novo método para a identificação e a busca de páginas-objeto, denominado OPIS (acrônimo para Object Page Identifying and Searching). O cerne do OPIS está na adoção de técnicas de realimentação de relevância e aprendizagem de máquina na tarefa de classificação, baseada em conteúdo, de páginas-objeto. O OPIS não descarta o uso de GSEs e, ao invés disso, em sua etapa de busca, propõe a integração de um classificador a um GSE, adicionando uma etapa de filtragem ao processo de busca tradicional. Essa abordagem permite que somente páginas identificadas como páginas-objeto sejam recuperadas pelas consultas dos usuários, melhorando, assim, os resultados de buscas-objeto. Experimentos, considerando conjuntos de dados reais, mostram que o OPIS supera o baseline com ganho médio de 47% de precisão média. / Object pages are pages that represent exactly one inherent real-world object on the web, regarding a specific domain, and the search for these pages is named as object search. General Search Engines (GSE) can satisfactorily answer most of the searches performed in the web nowadays, however, this hardly occurs with object search, since, in general, the amount of retrieved object pages is limited. This work proposes a method for both identifying and searching object pages, named OPIS (acronyms to Object Page Identifying and Searching). The kernel of OPIS is to adopt relevance feedback and machine learning techniques in the task of content-based classification of object pages. OPIS does not discard the use of GSEs and, instead, in his search step, proposes the integration of a classifier to a GSE, adding a filtering step to the traditional search process. This simple approach allows that only pages identified as object pages are retrieved by user queries, improving the results for object search. Experiments with real datasets show that OPIS outperforms the baseline with average boost of 47% considering the average precision.

Páginas-objeto são páginas que representam exatamente um objeto inerente do mundo real na web, considerando um domínio específico, e a busca por essas páginas é chamada de busca-objeto. Os motores de busca convencionais (do Inglês, General Search Engine - GSE) conseguem responder, de forma satisfatória, à maioria das consultas realizadas na web atualmente, porém, isso dificilmente ocorre no caso de buscas-objeto, uma vez que, em geral, a quantidade de páginas-objeto recuperadas é bastante limitada. Essa dissertação propõe um novo método para a identificação e a busca de páginas-objeto, denominado OPIS (acrônimo para Object Page Identifying and Searching). O cerne do OPIS está na adoção de técnicas de realimentação de relevância e aprendizagem de máquina na tarefa de classificação, baseada em conteúdo, de páginas-objeto. O OPIS não descarta o uso de GSEs e, ao invés disso, em sua etapa de busca, propõe a integração de um classificador a um GSE, adicionando uma etapa de filtragem ao processo de busca tradicional. Essa abordagem permite que somente páginas identificadas como páginas-objeto sejam recuperadas pelas consultas dos usuários, melhorando, assim, os resultados de buscas-objeto. Experimentos, considerando conjuntos de dados reais, mostram que o OPIS supera o baseline com ganho médio de 47% de precisão média. / Object pages are pages that represent exactly one inherent real-world object on the web, regarding a specific domain, and the search for these pages is named as object search. General Search Engines (GSE) can satisfactorily answer most of the searches performed in the web nowadays, however, this hardly occurs with object search, since, in general, the amount of retrieved object pages is limited. This work proposes a method for both identifying and searching object pages, named OPIS (acronyms to Object Page Identifying and Searching). The kernel of OPIS is to adopt relevance feedback and machine learning techniques in the task of content-based classification of object pages. OPIS does not discard the use of GSEs and, instead, in his search step, proposes the integration of a classifier to a GSE, adding a filtering step to the traditional search process. This simple approach allows that only pages identified as object pages are retrieved by user queries, improving the results for object search. Experiments with real datasets show that OPIS outperforms the baseline with average boost of 47% considering the average precision.

Páginas-objeto são páginas que representam exatamente um objeto inerente do mundo real na web, considerando um domínio específico, e a busca por essas páginas é chamada de busca-objeto. Os motores de busca convencionais (do Inglês, General Search Engine - GSE) conseguem responder, de forma satisfatória, à maioria das consultas realizadas na web atualmente, porém, isso dificilmente ocorre no caso de buscas-objeto, uma vez que, em geral, a quantidade de páginas-objeto recuperadas é bastante limitada. Essa dissertação propõe um novo método para a identificação e a busca de páginas-objeto, denominado OPIS (acrônimo para Object Page Identifying and Searching). O cerne do OPIS está na adoção de técnicas de realimentação de relevância e aprendizagem de máquina na tarefa de classificação, baseada em conteúdo, de páginas-objeto. O OPIS não descarta o uso de GSEs e, ao invés disso, em sua etapa de busca, propõe a integração de um classificador a um GSE, adicionando uma etapa de filtragem ao processo de busca tradicional. Essa abordagem permite que somente páginas identificadas como páginas-objeto sejam recuperadas pelas consultas dos usuários, melhorando, assim, os resultados de buscas-objeto. Experimentos, considerando conjuntos de dados reais, mostram que o OPIS supera o baseline com ganho médio de 47% de precisão média. / Object pages are pages that represent exactly one inherent real-world object on the web, regarding a specific domain, and the search for these pages is named as object search. General Search Engines (GSE) can satisfactorily answer most of the searches performed in the web nowadays, however, this hardly occurs with object search, since, in general, the amount of retrieved object pages is limited. This work proposes a method for both identifying and searching object pages, named OPIS (acronyms to Object Page Identifying and Searching). The kernel of OPIS is to adopt relevance feedback and machine learning techniques in the task of content-based classification of object pages. OPIS does not discard the use of GSEs and, instead, in his search step, proposes the integration of a classifier to a GSE, adding a filtering step to the traditional search process. This simple approach allows that only pages identified as object pages are retrieved by user queries, improving the results for object search. Experiments with real datasets show that OPIS outperforms the baseline with average boost of 47% considering the average precision.

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第20026号 / 情博第621号 / 新制||情||108(附属図書館) / 33122 / 京都大学大学院情報学研究科社会情報学専攻 / (主査)教授 田中 克己, 教授 吉川 正俊, 教授 黒橋 禎夫 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

An autonomous robot that acts in a goal-directed fashion requires a world model of the elements that are relevant to the robot's task. In real-world, dynamic environments, the world model has to be created and continually updated from uncertain sensor data. The symbols used in plan-based robot control have to be anchored to detected objects. Furthermore, robot perception is not only a bottom-up and passive process: Knowledge about the composition of compound objects can be used to recognize larger-scale structures from their parts. Knowledge about the spatial context of an object and about common relations to other objects can be exploited to improve the quality of the world model and can inform an active search for objects that are missing from the world model. This thesis makes several contributions to address these challenges: First, a model-based semantic mapping system is presented that recognizes larger-scale structures like furniture based on semantic descriptions in an ontology. Second, a context-aware anchoring process is presented that creates and maintains the links between object symbols and the sensor data corresponding to those objects while exploiting the geometric context of objects. Third, an active perception system is presented that actively searches for a required object while being guided by the robot's knowledge about the environment.