We encourage you to try out AIKO SciLingual yourself and experience everything it has to offer
AIKO SciLingual is an AI translation platform that specializes in
the fields of life sciences and pharmaceuticals It was developed and is offered by ASCA Corporation (ASCA) to fulfill the company's mission of "delivering new drugs to patients as quickly as possible ."
We encourage you to try out AIKO SciLingual yourself and experience everything it has to offer.
We encourage you to try out AIKO SciLingual yourself and experience everything it has to offer
A translation engine independently developed by ASCA, SciLingual has learned from bilingual data created by ASCA's own translations based on publicly available data such as clinical trial registries, pharmaceutical databases, as well as clinical and basic medical literature. This enables the engine to produce highly accurate AI translations specializing in the field of life sciences
AIKO SciLingual has been designed to place the utmost priority on protecting user data. None of the translation engines used re-purpose the input documents or data, thoroughly eliminating any risk of translation content being leaked externally. We make every effort to thoroughly protect privacy and confidentiality. Our safe and secure translation service provides our customers with peace of mind.
In addition to SciLingual, AIKO SciLingual is also equipped with general-purpose AI translation engines like Google and Microsoft, as well as other translation engines that utilize generative AI. It comes standard with nine translation engines, supporting translations in up to 36 languages. Users can select each translation engine according to their usage needs.
Support for a total of 8 file formats for translation, including Word, Excel, PowerPoint, and PDF. Also equipped with an OCR tool, allowing users to download translated files while retaining the layout and original file format, including scanned PDFs.
*For PDF files only, translated files can be downloaded in either docx or pptx format.
Translation results are displayed sentence by sentence, paired with the original text. This allows for smoother, more detailed verification and editing of translation results.
The results from multiple built-in translation engines can be compared side-by-side on the same screen, making it easier to choose the optimal translation and helping to improve quality.
Users can register glossaries and phrase collections on a word or sentence basis to further enhance translation accuracy.
The original texts and translations translated with AIKO SciLingual are stored in an AIKO SciLingual document list. Only the user themself can view the translation drafts and translated texts, including the document list. Past translation results can be checked anytime in a secure environment.
With AIKO SciLingual, users can easily share translation result screens, glossaries, and phrase collections with designated users within the same company. Ideal not only for individual translation tasks but also for information sharing and collaborative work within teams.
Administrator users can easily check each user’s monthly usage status and latest login date from the management dashboard. Since the AIKO SciLingual contract is based on the number of accounts, the accounts of users who do not use the platform frequently can be reassigned to others as needed.
ASCA has independently developed this translation engine specialized for life sciences. It supports translations between Japanese and English, and between Japanese and Simplified Chinese. In terms of evaluation metrics, it scores over 60 points in BLEU and over 0.8 in COMET, with an accuracy rate of 94% in correctly translating pharmaceutical product names, demonstrating high precision.
BLEU is the most commonly used indicator in research fields. It gives a score that evaluates the output quality of natural language processing models, including MT Engines. Translations are quantified based on how closely they match correct answers, allowing for an objective measurement of expressiveness and accuracy. Scores range from 0 to 100, with higher numbers indicating better quality. Generally, a score above 30 is considered high quality.
COMET is a metric that uses neural network technology to evaluate the quality of machine translation models. It is characterized by a high correlation with human evaluation and has increasingly been used in recent years as a replacement for traditional metrics like BLEU. Using a pre-trained language model, translation quality is scored based on the original text, the translation, and the correct reference translation. COMET scores range from -1 to 1, where a score close to 1 indicates a high-quality translation, around 0 indicates average quality, and negative values indicate a low-quality translation.
This is a translation engine developed by Google LLC.
It supports 36 languages and provides high-quality, speedy translations.
This is a translation engine developed by Microsoft Corporation.
It supports 34 languages other than Sinhala and Tagalog, and can be used for a wide range of purposes.
This is a translation engine developed by NAVER Corporation in South Korea. Its strengths particularly lies in accurate Korean translations. Supported languages include Japanese, English, Korean, Chinese (Simplified and Traditional), Vietnamese, Thai, Indonesian, French, Spanish, Russian, German, and Italian.
This is a proprietary engine developed and provided by Yaraku Corporation. It features simple and concise translations, making it ideal for post-editing tasks.
“ChatGPT GPT-4.1” is a large language model (LLM) developed by OpenAI LLC. Translation accuracy and expressions can be freely adjusted depending on the prompt. The model installed in AIKO SciLingual is operated in a secure environment that does not store data for fraud monitoring, compliant with the HIPAA law protecting personal patient information in the U.S., based on a BAA contract with OpenAI LLC.
The faster version of GPT-4.1. This is also operated in a secure environment that complies with the HIPAA law protecting U.S. patients' personal information under the BAA contract with OpenAI, ensuring that no data storage occurs for fraud monitoring purposes.
“Claude 3.5 Sonnet” is a large language model (LLM) developed by Anthropic in the United States. It is generally said to produce more fluent Japanese output compared to other LLMs. No data storage occurs for fraud monitoring purposes. * The Claude embedded in AIKO SciLingual is provided via Amazon Bedrock.
※AIKO SciLingualに搭載のClaudeは、Amazon Bedrock経由で提供しています。
“Gemini” is a large-scale language model (LLM) developed by Google DeepMind. It is said to have a deeper understanding compared to other major generative AI engines, enabling the generation of translations that consider context. The Gemini integrated into AIKO SciLingual has obtained opt-out consent for data storage to monitor misuse.
We create an original AI translation engine by sharing data from customers and provide it as part of an integrated package in AIKO SciLingual. We can build not only company-wide content such as clinical, non-clinical, CMC, clinical trial protocols, and investigator brochures, but also create original frameworks tailored to specific fields and document types.