What is LexMachina?
LexMachina is the first neural machine translation engine specialized in Swiss law available to lawyers. Designed specifically to meet the legal translation needs of Swiss lawyers and legal practitioners…
LexMachina is the first neural machine translation engine specialized in Swiss law available to lawyers.
Designed specifically to meet the legal translation needs of Swiss lawyers and legal practitioners, LexMachina was primarily trained with Swiss legal texts, thus ensuring coherence to the phraseology and terminology specific to our country.
Hosted on servers in Switzerland, LexMachina guarantees data security and confidentiality. Find out more.
In order to address the linguistic needs of our customers and continuously improve LexMachina, we are committed to offering unmatched customer service – by answering your questions and helping you overcome difficulties associated with legal machine translation.
Find out more about our team here.
What are the advantages of using LexMachina?
LexMachina allows you to translate any Swiss legal document within a matter of minutes for only CHF 75/month. Whether or not you choose to perfect the machine translation through human revision…
LexMachina allows you to translate any Swiss legal document within a matter of minutes for only CHF 75/month. Whether or not you choose to perfect the machine translation through human revision, LexMachina is the most time and cost-efficient solution. If you choose to review the translation yourself, you can request access to our full translator toolbox.
Since LexMachina was designed for Swiss lawyers, our users benefit from the three following professional must-haves:
What is the difference between statistical and neural machine translation, and which technology is LexMachina using?
Statistical machine translation has existed for more than 40 years and is at the origin of machine translation’s “dubious” reputation (and amusing accidents of gobbledygook). Neural machine translation (NMT)…
Statistical machine translation has existed for more than 40 years and is at the origin of machine translation’s “dubious” reputation (and amusing accidents of gobbledygook). Neural machine translation (NMT) is a breakthrough from 2018 powered by deep learning. As a matter of course, LexMachina is based on NMT and would never have been possible without this exciting new technology.
Statistical machine translation engines merely collect statistics on the training data (frequency of a term, co-occurrence of two terms). While NMT systems also use statistics, they are modeled after the way humans translate, regularly glancing back at the entire sentence, so as to incorporate contextual knowledge, and keeping linguistic knowledge of the source language in mind. This dynamic and innovative approach clearly outperforms mere statistical translation.
Neural Machine Translation (NMT) has been rendered possible by “deep learning” technology. A huge set of data (e.g. bilingual texts) is “fed” to powerful computers (Graphical Processing Units or GPUs), which try to recognize patterns by using different random trials and “spraying” the data into a slew of “vector spaces”. At the end of the process, the training data is captured by to a single mathematical formula (the “translation engine”), which will be used to translate further texts.
Obviously, the closer the texts to be translated are to the texts used to “train” the engine, the better the result will be: if an algorithm has been trained to predict the weather in Geneva, it will not work well to predict weather in London.
It is therefore essential that we carefully select the resources used to train the different sub-engines. This is why we are continuously gathering high-quality Swiss legal texts and expanding the “field-specific” sub-engines offered to our users.
How can the machine learn?
The expression “deep-learning” only refers to the creation of a translation engine (see “How does Neural Machine Translation work?” above). Once defined…
The expression “deep-learning” only refers to the creation of a translation engine (see “How does Neural Machine Translation work?” above). Once defined, the algorithm in principle stays the same, until the next re-training.
Public machine translation websites (Google, etc.) usually save all texts entered by their users in order to use this data in subsequent trainings. On the contrary, LexMachina NEVER saves your data. We augment our data sets by collecting publicly available specialized data and carefully categorizing it into the appropriate legal field. Furthermore, our partners share precious proprietary data with us to help us improve the performance of LexMachina.
There are also ways to adapt and improve NMT results between two trainings. When you send us feedback, we investigate the source of the problem and seek to resolve it. In some cases, we can correct the mistake directly in the source text, before it passes through the translation algorithm (pre-processing adjustment), or in the translation itself, after it has been generated by the machine (post-processing adjustment).
Do not hesitate to inform us of any significant problems encountered, by clicking on the “Edit” button in the machine translation interface or by contacting us directly.
You will find an explanation of the most frequent problems on our dedicated page.
How do we guarantee the confidentiality of your data?
LexMachina is the most confidential Neural Machine Translation (NMT) system available to Swiss lawyers. It has been designed to adhere to Swiss lawyers’ professional secrecy standards…
Would you like to know more about our security measures? Please see the following sections or speak with one of our specialists.
Texts entered or documents uploaded tothe machine translation interface are processed through the LexMachina translation system (e.g. for format conversion, segmentation, application of pre-processing rules, processing through the NMT-algorithm, post-processing, and re-formatting). However, they only “pass” through our system for the required processing time and they are not stored on our servers for any further purposes. Similarly, the translation produced by LexMachina is not saved on our servers. If you forget to save the translation, it will be lost and you will have to translate your document again.
What happens when you order a review service?
When you order a review service in your client interface, the quote is generated automatically by our system, without the need for any human intervention. After quote approval, …
When you order a review service in your client interface, the quote is generated automatically by our system, without the need for any human intervention. After quote approval, the reviewer is automatically given secured online access to the document to be translated for the duration of the project. The documents never leave our server.
All LexMachina reviewers undergo a strict vetting process (review of criminal and debt records, references, qualifications, and CV). They are bound by strict confidentiality agreements and have received training on security and confidentiality obligations in the legal and financial fields.
What happens when you use the full translator toolbox?
All our servers are located in Tier 4 centers in Switzerland.
Our security processes and measures have been implemented in accordance with ISO 27001 (Information Security). Although we have comprehensive documentation on this subject, the strict security standards we follow do not allow us to publish more detailed information online. For this reason, we only disclose such information as part of a personal discussion
LexMachina is the first translation platform to offer its users the opportunity to communicate privately with the reviewer in charge of their project. Our system automatically assigns a professional translator with proven expertise in the legal field indicated by the user. The user can post questions or comments on the translation directly in the Computer Assisted Translation (CAT) interface. The reviewer is thus personally responsible for quality assurance and client feedback.
In addition, the users benefit from LexMachina’s guarantees:
What are the most frequent errors? What are their causes? How can they be addressed?
You will find a description of the most frequent problems (as well as their solutions) here.
You will find a description of the most frequent problems (as well as the solutions that can be implemented) here.
You sent feedback, but the problem persists… why?
We take into consideration all feedback received and implement the requested changes or adjustments as soon as possible…
We take into consideration all feedback received and implement the requested changes or adjustments as soon as possible.
Some errors can easily be corrected and will be addressed in a matter of days. Others require the creation of complex rules or patches, the implementation of new technology, or a complete re-training of the engine.
We will keep you posted about the status of your enquiry.
How can you help LexMachina to help you?