Advances in the Theory of Distributional Learning of Formal Languages
| Project/Area Number |
17K00026
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Theory of informatics
|
|---|
| Research Institution | Hosei University (2018-2022)
National Institute of Informatics (2017) |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2017-04-01 – 2023-03-31
|
| Project Status |
Completed (Fiscal Year 2022)
|
| Budget Amount *help |
¥2,990,000 (Direct Cost: ¥2,300,000、Indirect Cost: ¥690,000)
Fiscal Year 2019: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2018: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2017: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000)
|
|---|
| Keywords | 文脈自由文法 / 文法推論 / 正例と所属性質問からの極限同定 / 分布学習 / 拡張正規閉包 / 拡張正規表現 / 有限文脈特性 / 所属性質問 / 正規木言語 / 正規演算 / 閉包性 / 等価性判定 / 情報基礎 / 形式言語 |
|---|
| Outline of Final Research Achievements |
Distributional learning algorithms for context-free languages work by assigning to each nonterminal of the hypothesized grammar a string set that can be decided by making queries to the membership oracle for the target language. In previous works, these string sets were limited to those that can be represented by finite conjunctions of membership queries. The present study presented two generalizations. The first generalization allows arbitrary Boolean combinations of membership queries in place of finite conjunctions. The second generalization allows regular operations in addition Boolean operations, and represents each nonterminal by an extended regular expression containing atoms for membership queries. These generalizations greatly extend the class of context-free languages that can be targeted by distributional learning algorithms.
|
| Academic Significance and Societal Importance of the Research Achievements |
いまだに謎に包まれている人間の母語習得のメカニズムの解明のためには，研究の指針となるような学習の数理モデルの確立が欠かせない。この観点から，母語習得のモデルとして一定の説得力を持つ学習の枠組みのもとで，どれだけ広い文脈自由言語の部分クラスが学習可能になるのかを調べることは，非常に重要な課題である。本研究は，正例と所属性質問からの極限同定の枠組みのもとで，従来の分布学習のアルゴリズムで目標言語とすることができる文脈自由言語のクラスを飛躍的に拡大することに成功した。
|
Report
(7 results)
Research Products
(11 results)
