We, Works Applications, authors and maintainers of Sudachi: a highly-featured Japanese Tokenizer & Morphological Analyzer, provide services to increase the accuracy of search in Japanese documents.

Our experienced members from Tokushima NLP Laboratory will support your efforts for creating accurate full-text search applications on AWS, using Elasticsearch or OpenSearch.

①. Technical support

　We provide technical support for Works Applications OSS,including Sudachi and elasticsearch-sudachi (a plu gin for Elasticsearch and OpenSearch).

Support is provided through a dedicated website and covers:

　・Technical questions

　・Technical issues

　・General product information

　　※Environments where we cannot freely configure plugins, such as OpenSearch Service, are not eligible for support.

②. Sudachi configuration for search

　We help you understand how Sudachi can improve your search pipeline. Support is provided through web conferences or email.

③. Search results evaluation

　We conduct a quantitative and a qualitative analysis of your search results to identify potential issues. Additionally, we can suggest solutions to improve your　search accuracy.

④. Search accuracy tuning

　We guide you through the steps to enhance your search results, including settings for Sudachi and full-text search engines (e.g., Elasticsearch, OpenSearch), finding the right segmentation units and normalization features, and expanding Sudachi’s synonym dictionary.

For more advanced tuning, consider using this service in conjunction with our search results evaluation service.

⑤. Custom dictionary development

　We provide customized dictionaries (user dictionaries) tailored to your specific needs.

　While Sudachi’s dictionary (morphological analysis, and synonyms) includes a large vocabulary, it may not comprehensively cover internal terminology or specialized vocabularies. In such cases, relying solely on Sudachi’s standard dictionary may not provide sufficient search accuracy.

　For example:

　・Drug names in the medical domain

　・Part names and model numbers in manufacturing

　・Technical terms in law and finance

　・Product names in e-commerce

　Creating a custom dictionary using Sudachi’s user dictionary feature can address this problem.

However, this process involves selecting segmentation units, adding normalization and segmentation information, and incorporating synonym information, which can be labor-intensive.

Our dictionary team handles these tasks for you.

　※Please note that you may have to provide reference materials, such as glossaries, as we may not have expertise in your target domain.

Amazon OpenSearch Service is not supported.

We can provide a customized solution tailored to your specific requirements.