sentAls

Themed research of jj1lis, IT Department, Tama High School of Science and Technology.

To Japanese reader

日本語版READMEはこちら

Overview

This is Sentiment Classification Analyzer for Japanese. "Sentiment classification" is the degree to which text is positive or negative terms in the emotion and impression. sentAls analyzes morpheme, syntax, and dependency that text includes and calculates its classification as score from them.

Dependencies

sentAls depends and is confirmed its motion following environments and libraries.

dlang: dmd v2.090.0 over (https://dlang.org/)
cabocha: cabocha of 0.69 (https://taku910.github.io/cabocha/)
Japanese Sentiment Polarity Dictionary: Natural Language Processing Lab (Inui-Suzuki Lab), Tohoku University (https://www.cl.ecei.tohoku.ac.jp/index.php?Open%20Resources/Japanese%20Sentiment%20Polarity%20Dictionary)

Feedback

contact me at:

issues on Github
email: rinta.nambuline205@gmail.com

about research

name: jj1lis
affiliation: Tokyo Metropolitan Tama High School of Science and Technology
research title: Sentiment Classification of Sentence in Japanese (日本語における文章の感情極性判定)

Usage

Clone this repo to your local.

$ git clone https://github.com/jj1lis/sentAls.git

Before using

Compile the part written in C++. Move to sentAls/cpp.

$ cd cpp/
$ g++ -c dependecy.cpp

Build project by DUB: package manager for D package. Move to the root directory (sentAls/) and execute following:

$ dub build

Refer to https://code.dlang.org if you don't have.

I'll also create makefile (someday).

Execute

$ ./sent-als [<option> <arguments>]

# example

$ ./sent-als -h
    #print help
$ ./sent-als -t おはようございます。 -i data/hogedata
    # input text "おはようございます。" and contents of file "data/hogedata"
$ ./sent-als -i fugatext -o fugaout
    # input from "fugatext" and output to file "fugaout.(sum|als|ctx)"

Options

There are these options now:

Option	Description	Supplement
-v	Display version	Ignore arguments
-h	Display help information	Ignore arguments
-t	Input texts from commandline	You can specify multiple texts
-i	Specify input files	You can specify multiple files
-o	Specify output file	You cannot specify multiple files; only one file
-n	No output	sentAls doesn't output results anywhere

About output option

sentAls defaultly output results to standard output. It creates three output files if you specify option -o:

analysis file (.als)
- Detailed analysis contents such as phrase weight and score for each sentences.
summary file (.sum)
- Summary that is texts and score only.
context file (.ctx)
- Basic format.

Executing -o sample make sentAls create sample.als sample.sum sample.ctx.