Bahasa-Indo-NLP-Dataset

A catalog of Indonesian language processing software, dictionaries, and text corpora, grouped by task, with each entry listing its language, license, and author.

The repository gathers Bahasa Indonesia NLP material into one place, and its opening line invites pull requests. Entries are split between software libraries, dictionaries, and corpus, then grouped under task headings. The practical result is that a developer can scan the whole resource without bouncing between a dozen separate projects.

Each task gets its own list. There is a Python toolkit called bahasa that the README describes as pre alpha, a sentiment analysis library, an Open NLP POS tagger in Java, an indonesia ner library also in Java, and the sastrawi stemmer written in PHP. Rows carry the description, programming language, license, and an author link, so the attribution is visible on every line. A question answering system for Bahasa Indonesia appears too, listed under its own heading.

The data side is just as organized. MALINDO Morph is a morphological dictionary for Malay and Indonesian, and TALPCo is a parallel corpus with a Japanese to Indonesian pairing. The downloadable corpus section includes an Indonesian CoNLL 2017 universal dependency dataset around 29.64 GB, an opinion words list, a frequency list of common words, and a manually tagged corpus. WordNet Bahasa also shows up, described as a large scale semantic dictionary modeled on the Princeton WordNet.

Pre trained word vectors close out the list. A fastText skip gram model trained on Wikipedia offers 300 dimensions under a Creative Commons license, and a word2vec Indonesian model comes in at about 402 MB. The grammar section adds INDRA, an implemented HPSG grammar for Indonesian under the MIT license. Between the models and the grammars, the catalog covers both ready to use embeddings and more formal language resources.

The ordering is what makes the file worth keeping around. Someone starting on Indonesian NLP can land here, find the right library or corpus, read off the license and author, and move on. Additions are welcome, so the set keeps growing.

• Source 1 · Project repository
• Source 2 · README