TTI COIN Lab. External Server
This page contains research resources produced by the
Computational Intelligence Laboratory, Toyota Technological Institute (Nagoya, Japan).
Machine Learning Tools
Dual Coordinate Ascent SVM 0.03 (tgz) (License: GPL2.0)
The training time of DCASVM-0.03 over the RCV1.test data (Liblinear data) is 13s on our Xeon 3GHz server. Liblinear-1.94 takes 17s for the training of the same data.
Eze Hierarchical Classification System 0.01 (tgz) (License: GPL2.0)
Eze is a fast hierarchical classification system based on DCASVM.
Given a hierarchical class structure (tree or DAG), Eze can train
hierarchical classification models very fast and also can classify
new data very fast.
Eze can learn from the LSHTC3 Wikipedia Medium training data (456,866 data)
in around 30 minutes and can classify new data (81,262 data) in seven minutes
based on a large-scale class hierarchy (50,312 DAG-structured classes).
Smart Vehicle Ontology (v0.1) (here)
SC-CoMIcs (Superconductivity Corpus for Materials Informatics)
Materials Informatics (MI) needs textual datasets to accelarate the studies in this area, but there is no sizable datasets
suitable for our superconducting material search. In this respect, we decided to create a new corpus from scratch for MI information extraction.
SC-CoMIcs is the corpus that can contribute to the advancement of MI studies, especially in supercon.
Descriptions and experiment tools over the dataset can be found at
Suercondictivity information extraction dataset can be downloaded as follows.
Note that you need to agree the license displayed here. The set of 1000 MI abstracts are specially
permited to share in the research community under CreativeCommons BY-NC 3.0 by Elsevier under
a written agreement (#200221-005626). This is the strict condition you MUST obey.
1000 abstracts (License: CC BY-NC 3.0, Elsevier permission: 200221-005626) (here)
Separate stand-off annotations (License: CC BY 4.0) (here)
Brat (annotation visualizer) conf files (License: CC BY 4.0) (here)
Web Demo Systems
Word Embeddings with Capturing Antonymy
Embeddings of Bibliographic Information
Last update 2017-08-31 by Y. Sasaki