Publicação
Information search in web archives
| Resumo: | Web archives preserve information that was published on the web or digitized from printed publications. Many of that information is unique and historically valuable. However, users do not have dedicated tools to find the desired information, which hampers the usefulness of web archives. This dissertation investigates solutions towards the advance of web archive information retrieval (WAIR) and contributes to the increase of knowledge about its technology and users. The thesis underlying this work is that the search results can be improved by exploiting temporal information intrinsic to web archives. This temporal information was leveraged from two different angles. First, the long-term persistence of web documents was analyzed and modeled to better estimate their relevance to a query. Second, a temporal-dependent ranking framework that learns and combines ranking models specific for each period was devised. This approach contrasts with a typical single-model approach that ignores the variance of web characteristics over time. The proposed approach was empirically validated through various controlled experiments that demonstrated their superiority over the state-of-the-art in WAIR. |
|---|---|
| Autores principais: | Costa, Miguel Ângelo Leal da, 1979- |
| Assunto: | Arquivos digitais World Wide Web Pesquisa de informação Aprendizagem automática Teses de doutoramento - 2014 |
| Ano: | 2014 |
| País: | Portugal |
| Tipo de documento: | tese de doutoramento |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade de Lisboa |
| Idioma: | português |
| Origem: | Repositório da Universidade de Lisboa |
| Resumo: | Web archives preserve information that was published on the web or digitized from printed publications. Many of that information is unique and historically valuable. However, users do not have dedicated tools to find the desired information, which hampers the usefulness of web archives. This dissertation investigates solutions towards the advance of web archive information retrieval (WAIR) and contributes to the increase of knowledge about its technology and users. The thesis underlying this work is that the search results can be improved by exploiting temporal information intrinsic to web archives. This temporal information was leveraged from two different angles. First, the long-term persistence of web documents was analyzed and modeled to better estimate their relevance to a query. Second, a temporal-dependent ranking framework that learns and combines ranking models specific for each period was devised. This approach contrasts with a typical single-model approach that ignores the variance of web characteristics over time. The proposed approach was empirically validated through various controlled experiments that demonstrated their superiority over the state-of-the-art in WAIR. |
|---|