Vyasa Layar Release Notes
Support documentation covering updates, bug fixes, and new software features for Vyasa Layar.
Vyasa Layar 2.60.0
September 2022
What’s New
- Automated Language Detection! Every document indexed with Layar is now run through our language detection model to identify the language used in the text. This prediction is then tagged as an annotation on the document and can also be filtered on using Layar search or through Curate’s metadata filters.
- PubMed metadata now available as Layar filters. Attributes including publication type, MeSH terms, and IDs such as the DOID and PMID can be used as filterable properties in a Layar search or through Curate’s metadata filters.
- New Layar Classification Model Available. Users can now leverage the popular fasttext classification model as a base model in their Layar projects. Classify texts into categorical groups with a quick and iterative classification model in Layar or Curate.
- Text Embedding Service (now available via the API). Transform your texts into vectorized embeddings using our new text embedding service. Use these embeddings as part of your data analytics pipeline to find texts of similar content, style, and more.
- Table Similarity (now available via the API). Compare two or more tables (or specific columns within two or more tables) to determine the semantic similarity between the tables. This feature utilizes the text embedding service mentioned above.
- Additional improvements made to the software:
- Users can now bulk delete projects within Layar
- Small improvements to the grouping algorithm have been made available in Layar
Vyasa Layar 2.59.0
August 2022
What’s New
- New Layar Models Available for Molecule Generation:
- ChemFormer Molecule Generation: generate novel compounds with a transformer model.
- ChemProp Molecular Property Prediction: predict molecular properties for an input of compounds using a message passing neural network.
- ChemVector 2.0: a new and improved ChemVector autoencoder model that vectorizes molecular structures and optimizes for specific properties.
- Render SMILE strings in the browser. Any tables where a SMILE string column is present can be configured to render the compounds directly within the browser.
- Compound Analytics 3D Plot Visualization! Any compound generation or property prediction model you run in Layar now has a 3D scatterplot to visualize results. You can configure any of the three axes directly within the UI!
- Tox21 Visualizations are Back! You guys asked for it, we brought it back! The toxicity prediction self-organizing map is back in Layar for any project run with the Tox21 Graph Convolutional Neural Network module!
Vyasa Layar 2.58.0
July 2022
What’s New
- New Layar Models Available:
- Text Clustering: In Curate and API
- Grouping Improvements! Any results generated by our QA models now implement an improved grouping algorithm, which has been applied to all applications where question answering is utilized (Axon, Curate, Layar, Layar API, Signal, and Synapse).
- VCF files are now supported! If you’re working in genomics, this is a big deal! You can now upload large VCF files and preview the file as a table directly within Layar!
- Please note that due to the size of these files, users will only be able to view the first 100K rows of these files within the UI. For genomic analyses, please utilize the Layar API to query the full file.
- Any user can now add additional users. Previously, only administrators could generate the registration link for new users. Now, anyone can create this link and share out.
- This feature can be changed back to be an admin-only capability. If you wish to change this to admin-only, let us know by reaching out to Vyasa’s Support team.
- Bulk delete projects in Layar. No more one-by-one deletions. Be careful! You’ll get a warning signal before deletion, but once they’re gone you will not be able to access them again.
Vyasa Layar 2.57.0
June 2022
What’s New
- New HPE Greenlake Connector! Connect data within your Greenlake cloud storage using our new Layar connector offering.
- SQL Querying Now Available in Layar Tables. Run real-time SQL queries directly within Layar. Click into tables that originate from your SQL connector, and run a query directly on the table of interest.
- Clinical Trials Now Captures All Record Changes. Instead of capturing the first record of a clinical trial, we now capture the first record of a registered trial, and also scan for modifications/updates to that record. Any time a record is changed, the document is added as a new file in Layar, and linked back to the original registered trial.
- New Layar Models Available:
- Multi-Document Summarization – In addition to our standard summarization models, users can now auto-generate one summary for multiple input documents. For example, if a user is looking to build a single summary of clinical outcomes for the year, based on twelve monthly reports.
- Tense Conversion – Convert a snippet of text from past to future tense, or vice versa.
- Jupyter IDE in Layar Has Been Removed. Based on our feedback from users who utilize the API, we’ve decided to remove the embedded Jupyter notebook from Layar. Users can still leverage the Jupyter IDE in their own workspaces, and can utilize the Layar SDK as they had before.
Bug Fixes
- Minor bugs around ontology tagging in Layar tables have been addressed.
- Copy/paste with NER highlighted text has been refactored to cleanly copy raw text, without the concept labelling.
Vyasa Layar 2.56.0
May 2022
We’ve been busy this month! Here’s a few of the many updates we’ve made to the software & model collection.
What’s New
- Vyasa’s Developer Hub is now live for our technical users. This hub provides in depth information about the endpoints, tutorials, support, and more!
- New Sharepoint Connector Available! Connect data stored within your Sharepoint account using our new Layar connector offering.
- Concept Details Page in Layar. Any highlighted word or phrase you see in Layar has its own detail page for that specific concept! Learn about its synonyms, which ontologies it exists in, what documents or tables it can be found in, or view any related concepts based on the other data you’ve connected in your data fabric (see featured image above).
- Buckets of Layar Table Enhancements!
- Concept Type Prediction & Assignment – If you’ve been interested in tagging your table with specific concept types, we now enable automated concept type prediction. Simply select “Tag Concept Types” for a specific column to start this tagging process. Users can also override these predictions for validation, and to further train the NER models with their suggestions.
- Ontology Highlighting Now Available – The same ontology highlighting feature available in Layar Documents is now available in Layar Tables! Similarly to how you tag for NER, click the “Tag Concepts”, and select any Layar Ontologies you’d also wish to tag in the modal.
- Filter tables to view results based on a given column, including keywords, empty/non empty cells, etc.
- Sort your table based on a specified column.
- New Layar Models Available:
- NER Model for Study Endpoint – e.g. overall survival, progression-free survival, objective response rate, or other symptom derived endpoints)
- NER Model for Trial Arm – This model will tag snippet of text it believes to be a cohort or clinical trial arm. For example, an experimental group, active comparator, placebo, etc.
- Text Classification Module – Classify documents into one or many categories based on their text, such as emails, invoices, clinical records, etc.
- Boolean Question Answering – Users now have the option to return true/false results for a question (e.g. “Is this document in English?”) rather than trying to find an answer directly from the text.
Bug Fixes
- We’ve made some small but noticeable performance improvements for our user interfaces and data retrieval capabilities.
- Clinical Trials had a brief lapse in indexing new records – this has been resolved, and all missing records are accounted for.
Vyasa Layar 2.55.0
April 2022
What’s New
- Relationship Extraction in Layar. We’ve enabled a new visual feature that shows you relationships between two or more concepts within a text-based document! For example, identifying the reactants and chemical reactions needs to generated a specific product (see featured image above).
- Layar Search Expansion Using Ontologies. Expand your search to include any synonyms for a term, based on Layar Ontologies! When submitting a keyword or boolean search in Layar, terms will be run against any ontologies that exist in their instance. Any terms with synonyms/are synonyms can be incorporated into the search by clicking “Expand search” underneath the search bar.
- Dynamic Connections to Sybase & Oracle Connectors. Users now have the option to reference their Sybase or Oracle tables dynamically, rather than a snapshot of those datasets at the time of creating the connector.
- Clinical Trial Eligibility Criteria. Oftentimes, the eligibility criteria of a registered clinical trial bundles the inclusion and exclusion criteria together into the same snippet of text. We now automatically parse these two properties out into separate distinct metadata sections for QA or other Layar pursuits. We also keep the original eligibility criteria property, as well.
- Major Question Answering Additions!
- Document Section QA. Users can now focus their queries on specific areas of the text! Unstructured texts are automatically parsed by Vyasa into their respective headings / subheadings, so you can submit a question on specific sections. For example, you can ask “What protocol was used for staining the protein targets?” in the ˆmethods section of a publication, rather than the entire document.
- This not only improves the speed of your bulk query jobs, but also gives you more refined responses to the specific areas you wish to examine in your documents.
- Query Understanding Enhancements. This secondary query understanding pipeline kicks into gear when an input question returns low or zero results with our default pipeline. This secondary pipeline can also be leveraged via the API as its own endpoint (see the query_expansion endpoint for the QuestionApi class).
- QA Benchmarking. Users can now follow along with specific QA projects & models to understand the precision and recall of their models. Currently only available as API endpoints.
- Document Section QA. Users can now focus their queries on specific areas of the text! Unstructured texts are automatically parsed by Vyasa into their respective headings / subheadings, so you can submit a question on specific sections. For example, you can ask “What protocol was used for staining the protein targets?” in the ˆmethods section of a publication, rather than the entire document.
Vyasa Layar 2.54.0
March 2022
What’s New
- Highlighting for Ontologies Now Available! In addition to highlighting named entities in a document, Layar will now automatically tag any terms (or their relevant synonyms) based on an uploaded ontology. Simply select the ontology (or ontologies) of interest under the Concept Types legend, and Layar will automatically highlight any terms in the text that are from those ontologies.
- Hierarchical Table-Converted Ontologies. Take a table based file (csv, tsv, xml, etc.) and convert it into a flat data dictionary or tree-based hierarchical ontology! To learn more, check out this article.
- Expansive Keyword Search. When you run a keyword or boolean search in Layar, you now have the option to expand your search to return all results for synonyms as well as the initial term!
- Simpler Filter Menu. We’ve narrowed down the the filter menu to have the most widely used filters available for quick and easy utilization. That being said, those more advanced filters haven’t gone away – they still exist! If there’s a more advanced filter you’re interested in, you can click “Load More”, and you will see all of the previous filters available.
Bug Fixes
- We’ve implemented additional security improvements to the Layar system.
Vyasa Layar 2.53.0
February 2022
What’s New
- Air-gapped Deployments Now Available! Customers who require a disconnected network environment for installation and upgrades can now use our air-gapped Layar package. Please refer to this documentation for more information on how to install our software.
- PDF Preview Updates. Our PDF viewer has been updated! You may see a different layout than in previous Layar releases. The viewer still contains all prior functionality, but loads much quicker (especially for those pesky larger PDFs), has a higher resolution, and includes a built in search. Users can also view the Table of Contents for a PDF, if the PDF was provided with Table of Contents metadata.
- PDF Table Extraction Updates. We’ve made some major improvements to extracting tables from PDFs. Here are the highlights:
- Instead of using the “Extract Table” button and scrolling to the page of interest, users can scroll over to the page of interest, click and drag to create a boundary of the table they would like to extract, and click “extract table” to trigger our automated table extraction feature.
- Tables extracted in this manner will generate a downloadable CSV, and will be linked to the parent document as a “Related Document”, file type: table.
- Users can accurately extract tables with merged columns, invisible cell boundaries, or multiple header titles!
- Automated table extraction can also handle tables in landscape orientation, as well as tables that spread across multiple pages!
- We’ve made some improvements to processing time for large PDF uploads.
Bug Fixes
- Documents from our Canonical Data Fabric were being given an incorrect URL, making it difficult to link back to a document in Layar from the other Vyasa applications. This has been resolved.
Vyasa Layar 2.52.0
January 2022
Happy New Year!
What’s New
- New SAP HANA Connector! Connect data stored within your HANA database using our new Layar connector offering. Any information you can view using a HANA SQL query, you can connect to Layar. See the documentation here for a quick walkthrough of how to set up the connector.
Vyasa Layar 2.51.0
December 2021
What’s New
- We’ve added arXiv as a new data source in our canonical data catalog! An open-access archive of nearly 2M scholarly articles in diverse fields of physics, mathematics, computer science, quantitative biology, quantitative finance, statistics, and more.
- Custom keyword highlighting! Users can now input a list of terms or phrases into a document and view highlighting of any term in that list for rapid insights.
- A handful of improvements have been made to the packaged version of Layar for on-prem installations. AWS Marketplace and NVIDIA GTC Catalogs have been updated with the new Layar packages.
- We’ve added some object detection in our table extraction feature to improve identification of text and cell borders within tables.
Bug Fixes
- Log4j security patches were quickly deployed to all Layar systems in response to the Apache log4j security vulnerability.
- Data derived from Twitter connectors were not being processed and fed into our QA pipeline. This has been resolved.
Vyasa Layar 2.50.0
November 2021
What’s New
- We’ve added bioRxiv and medRxiv as a new data source in our canonical data catalog! Two open-access archives with a collective 1M scholarly articles covering all aspects of research in clinical research and the life sciences.
- We’ve also buffed up our document ingest process to scale with these new large data sources, and more on their way.
- Layar is now compatible with LaTeX-processed documents!
Bug Fixes
- We’re continuing to improve paragraph processing for PDFs to ensure the most comprehensive text extraction of these unstructured documents.
Vyasa Layar 2.49.0
October 2021
What’s New
- Table extraction got a huge upgrade!
- Users can extract highly complex tables from PDFs, including those that stretch across multiple pages, are displayed in a landscape page orientation, or have dozens of merged rows & columns.
- Extracted tables for PDF are available as “Related Documents” for the parent document. Each Related Doc table provides a viewer that displays the original table in the PDF to compare to the extracted output.
- Quickly reference which PDF pages had a table! Document pages where a table is found by our algorithms gets tagged with a document page classifier, provided as an annotation in Layar.
- Developers also have access to this new and improved table extraction capability via the Layar API
- PDF documents from CT.gov (including protocols, statistical analysis plans, and more) are now automatically extracted and added as part of the Clinical Trials data source in our canonical data catalog!
- On-prem packaged installations of Layar are now much easier, with some improved user onboarding to boot!
- System administrators can now install several Layar instances onto the same machine.
- We’ve swapped out our PDF viewer in the single document view with something new and improved!
- Our question answering pipelines are getting some major feature updates over the next few months. A larger announcement to come!
Bug Fixes
- That ever elusive double-login bug? Squashed.
- We’ve implemented a handful of small improvements to our data import pipeline for JDBC drivers.
Vyasa Layar 2.48.0
September 2021
What’s New
- We can now connect SQL tables via our new SQL Server connector!
Bug Fixes
- On-premise stack builds now have a startup page for Layar installations. Now, you can see a 404 or 503 error until the stack is stood up. After Layar has been instantiated, you will see the Layar login/registration page.
- We’ve implemented a handful of small improvements to our data import pipeline for PDF documents.
Vyasa Layar 2.47.0
August 2021
What’s New
- We’ve added a few new logs that are accessible through the Layar API for question-answering events on documents that reside in Layar.
Bug Fixes
- There was an authentication issue preventing users from downloading Layar documents to their local device. This has been resolved.
Vyasa Layar 2.46.0
July 2021
Bug Fixes
- We updated some thresholds in the question answering service to make our life science results a little less noisy.
- We’ve implemented a handful of small improvements to our data import pipeline for PDF documents.
Vyasa Layar 2.45.3
June 20, 2021
Bug Fixes
- Layar has been updated on AWS Marketplace to AMI version 21.07.1
- Dependency Update! Kubernetes has now been updated to 1.19
- We polished some rough edges and fixed up some minor bugs.
Vyasa Layar 2.45.2
June 14, 2021
What’s New
- We heard you! You love Sets! Now, Layar search has a new filter capability based on Sets. You can narrow down your search results to files that reside in a specific Set you’ve made.
- NER concept tagging has also been added to trials from our ClinicalTrials.gov canonical data source! See automatic highlighting for disease indications, organizations, small compounds or more!
Bug Fixes
- Turns out tables are quite stubborn, and they don’t like being edited or changed, at least in Safari & Firefox. We’ve had a sit-down chat with them and talked out our feelings, they’re going to be more accommodating now.
Vyasa Layar 2.45.1
June 8, 2021
What’s New
- Layar is now available on AWS Marketplace as a self-deployed AMI!
- We’ve added Dell’s EMC Isilon servers as a new Layar connector!
- You may have noticed a new “Questions” section in the Layar navigation menu. That’s right, we’ve brought Question Answering into Layar.
- If you noticed the “Questions” tab, you must have also noticed the “Develop” section. Quite astute, you are! We’ve brought in a Jupyter notebook IDE for custom code & model management, as an homage to our fellow data scientists.
Vyasa Layar 2.45.0
June 1, 2021
Bug Fixes
- Our document ingest performance hit a new personal record! It’s quite zippy now.
Vyasa Layar 2.44.2
May 24, 2021
Bug Fixes
- Product Tour inbound! Your first time registering/logging into Layar, you’ll get a walkthrough of the major features within the software.
- We’ve leveled up our table viewing and analytics – more loaded rows, more added columns, and our NER concept tagging doesn’t get tripped up by the large volume of cells!
- Layar’s DICOM viewer got a touch-up
- You can now add a new parent node into your ontologies
Vyasa Layar 2.44.1
May 17, 2021
What’s New
- We have launched our Layar helm chart on NVIDIA’s NGC Catalog!
- We made some internal architecture changes to how we serve our deep learning models, including using our multi-GPU systems more efficiently!
- Added an “Exit” drop down option to leave the System Admin Panel
- Incorporated additional metrics for API tracking and event logging
Bug Fixes
- Linked more APIs in the Layar Swagger documentation to the Layar web app
- We’ve resolved some hiccups in our parallel file uploading pipeline!
- Some of the Layar filters were acting up, but we’ve made improvements to how the filters render in Layar search.
- Squashed some other, smaller, data import bugs.
Vyasa Layar 2.44.0
May 10, 2021
What’s New
- Layar search can now search for a date range using the metadata from any document’s annotations.
- Users can now upload a single local file system as a Layar connector, instead of uploading individual documents.
Vyasa Layar 2.43.0
April 2021
What’s New
- The Vyasa Canonical Data Fabric now has [email protected] and Biotech News!
- We’ve embedded a Python SDK into Layar, for your friendly neighborhood scientists
- You can now export NER Concepts from documents and tables
Bug Fixes
- We’ve improved the overall onboarding process for SaaS and helm chart users.
- You may have noticed the Data Connectors menu has changed. Everything is still there, it’s just in alphabetical order now 🙂
- System admin user privileges were playing hide and seek, really well. We’ve increased the size of the overlay in the admin panel dashboard to make them easier to find.
- The dreaded double login has returned! We’ve gone on a quest to vanquish it for good. In the meantime, we’ve resolved some other accomplices to the issue.
- One contributor was a pesky Registration button that refused to let users into Layar until they refreshed the browser page. We walked it off the plank.
Vyasa Layar 2.42.2
March 29, 2021
What’s New
- We can now run address parsing on unstructured documents to pull out any important addresses for clinical sites, sponsors, etc.
Bug Fixes
- We’ve made updates to our Question Answering models to improve our clinical trial protocol analytics and ontology mapping capabilities
Vyasa Layar 2.42.1
March 22, 2021
What’s New
- .Layar’s table view has a fresh new look. When you extract tables from PDFs in Layar, you can view them side-by-side with the original PDF to compare!
- You can also filter down the types of files you see in the “Related Documents” modal. Want to only see the images automatically extracted from a PDF? Here’s a good solution!
Bug Fixes
- Layar’s Q/A got a new speedy personal record. We’ve tuned it up and taking it for a drive!
- Some file thumbnails were MIA – we’ve tracked them down and all is well.
Vyasa Layar 2.42.0
March 8, 2021
What’s New
- Layar got a new and fresh font style
- Ontologies got a handful of new capabilities! Users can now create ontologies from CSVs, add new or additional attributes/properties, and search within a given ontology to quickly find terms of interest.
Bug Fixes
- Connected Amazon S3 buckets can also have their children folders connected to Layar without repeating steps.
- Local documents can be renamed within Layar
- Some hiccups in the Firefox 68 supported Layar version have been resolved
Vyasa Layar 2.41.0
March 1, 2021
What’s New
- When you create a whole bundle of Sets, it’s difficult to find a specific Set on the left hand menu. So we’ve created a new modal that lets you search all of your Sets, or create a brand new one, without scrolling through all of your Sets on the left hand menu.
Bug Fixes
- We’ve improved our text and image capture for image-native PDFs
- Not to mention the TIFF image capture performance enhancements
- Finally, we touched up the user registration screens with some small modifications
Vyasa Layar 2.40.2
February 16, 2021
What’s New
- We’ve heard a lot of excitement from folks about our ClinicalTrial.gov data, so we’ve enhanced our filters to also enable users to search based on specific trial metadata, such as whether the trial has an attached Protocol or Statistical Analysis Plan.
- There have also been a handful of updates made to the “Related Documents” section for each file, including a new navigation capability to allow to to quickly scroll through related documents to find what you’re looking for.
Bug Fixes
- We’ve made big improvements our automated extraction of tables from unstructured PDF documents!
Vyasa Layar 2.40.1
February 8, 2021
Bug Fixes
- Filters for NER tags on tables works similarly to docs now!
- The double-click username login bug has emerged, and vanquished, once more!
- We’ve spruced up the Account Billing page with API usage tracking
- We’ve hit the keys and improved our Question Answering pipeline for better performance on pre- and post-processed text
Vyasa Layar 2.40.0
February 1, 2021
Bug Fixes
- Nothing major in this release, but we have improved navigation between related documents. Now, it’s easier to switch between parent and child files within Layar.
Vyasa Layar 2.39.4
January 25, 2021
What’s New
- We now ingest clinical trial protocol documents as PDFs from the ClinicalTrials.gov data repository! Provided through our Vyasa Canonical Data Fabric.
- “Related Documents” for a given file shows all of the images and tables extracted from a document automatically – all in one place.
- Question answering in Synapse and Axon now includes a thumbs up/ thumbs down manual validation capability that will report back to our models for fine-tuning and performance improvements.
- Synapse and Axon also have an ontology filter to narrow down results based on whether those terms/phrases exist in a given ontology.
Bug Fixes
- We’ve hit the keys and improved our Question Answering pipeline for better performance on pre- and post-processed text
- We’ve squashed a couple small bugs for table extraction in PDFs
- Search results weren’t saving for some users – it’s been tracked down and fixed!
- PDF previews in Layar were also misbehaving – it was a browser-based issue, and quickly resolved.
Vyasa Layar 2.39.3
January 23, 2021
What’s New
- Each Layar instance is now billed on a “Per Node” basis. See the pricing details for more information.
Bug Fixes
- We’ve hit the keys and improved our document classification pipeline for better performance on life science and clinical texts.
- We’ve re-trained our Question Answering pipeline to focus on to particular sections of documents for different life science tasks.
Vyasa Layar 2.39.2
January 19, 2021
Bug Fixes
- Our NER tagging API (on the Layar API Demo App) now highlights multiple NER concept types for a term.
- Users can now also log in to the API Demo App through their Layar instance
- We’ve put in some solid developer hours into our term grouping algorithms
- Layar search no longer has a problem with hyphens and other special characters
- A handful of ontology search capabilities have been updated — now, search looks through parent nodes,child nodes, and term synonyms. Results are also whittled down to view the final terms results, instead of opening up the full tree in a hierarchical ontology.
Vyasa Layar 2.39.1
January 11, 2021
Bug Fixes
- We’ve improved the speed and performance of dynamically adding Layar nodes via the System Admin panel
- Node-to-node sharing has become a much more streamlined process for users
- Billing is fully self-contained to the Layar node, on a per instance license
- We’ve hit the keys and improved our Question Answering pipeline for better performance on pre- and post-processed text
- We’ve squashed a couple small bugs for table extraction in PDFs
Vyasa Layar 2.39.0
January 4, 2021
What’s New
- Layar can now share data from one Layar node to another – even if that node is in someone else’s data fabric! That’s pretty stellar news, guys!
- We can also offer fully self-contained stack testing through HTTPS/HTTP support!
We’ve built out a new API for the Layar API Demo app, which identifies the similarity between two different columns of data!
Happy New Year! Welcome to 2021!
