2014 Fiscal Year Final Research Report
Calculating Autocorrelation Function for Word Occurrences in Texts and Its Modeling with Stochastic Processes
| Project/Area Number |
25580093
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Linguistics
|
|---|
| Research Institution | Showa University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2013-04-01 – 2015-03-31
|
| Keywords | 自己相関関数 / 拡張指数型関数 / 確率過程 / ポアソン過程 / 非定常ポアソン課程 / Cox過程 / 言語統計 / テキストマイニング |
|---|
| Outline of Final Research Achievements |
In this study,we attempt to offer a new analyzing point of view for texts in which occurrences of words are considered as dynamical time series. Based on this interpretation of texts, we propose a method for calculating autocorrelation function (ACF) which represents the correlation between occurrences of a considered word. In our method, the basic time unit of the stochastic process of word occurrence is taken to be one sentence and this allows us a suitable definition of ACF. The examples of ACF obtained through our method for 'conceptual words'and those for 'nonconceptual words' are given and their characteristic behaviors are discussed. Here, the term 'conceptual word' means the word which is deeply related with the central concepts or themes of text, and the 'nonconceptual word' represents the word which is not related with themes of text. It was found that the ACFs for 'conceptual words' and those for 'nonconceptual words' show entirely different characteristic behaviors.
|
| Free Research Field |
統計的機械学習，計量言語学
|
