We all have too many spreadsheets. Managing and aggregating data can be time consuming, and extracting new pieces of data from unstructured sources is often a manual, tedious process. The result is a large amount of ‘dark data’ that can’t be efficiently leveraged.
Vyasa Synapse is a novel, deep learning AI powered tool that enables you to derive value in dark data. Synapse AI agents efficiently dive into large-scale unstructured (document and data stream) content to extract information relevant to what’s currently in your spreadsheets.
Synapse enables this connection between data tables and unstructured content with deep learning text analytics powered “Smart Table” technology. This novel technology allows users to build tables of structured data with information from unstructured sources. Smart table functions include the ability to find related concepts with vector proximity, visualize trending terms, automatically determine synonyms, and ask natural language questions of content from a wide variety of data sources.
Learn more about how Synapse can fundamentally change the way you connect structured information within spreadsheets to unstructured content and shine a light on the dark data in your organization.
Features & Benefits
The power of Synapse is its ability to make connections between structured spreadsheet data and unstructured text data sources.
Volume, velocity, and variety are king when it comes to generating value from data, and Vyasa Layar works beside Synapse as a deep learning data fabric solution to deliver all three. We’ve pre-modeled publicly available data and established industry partnerships to ensure that Synapse can provide a wide array of quality sources alongside your existing data as soon as you start working with it.
Synapse can connect to external data sources including scientific publications, medical journals, patent office records, twitter streams, RSS feeds, web crawlers, and more. Users can also drag and drop their own structured and unstructured data sources (text, PDFs, CSVs, S3, etc.) directly into the application.
Sourced across a variety of domains, a number of large unstructured databases come pre-loaded into Synapse including:
- Pubmed: Draw from over 29 million citation for biomedical literature from MEDLINE, life science journals, and other online texts. Maintained by the NCBI, Pubmed is a database containing citations and abstracts from the fields of biomedicine and health, life sciences, behavioral sciences, chemical sciences, and bioengineering.
- Patent: Access the US Patent and Trademark full-text database. The database contains information on all US patents from 1790 to the most recent issue week. Full text data is available on all patents from 1976 to present.
- Clinical Trials: Explore over 300,000 clinical research studies in all 50 US states and 209 countries. Maintained by the U.S. National Library of Medicine, this trove of clinical trials includes trial meta-data, participant flow, baselines characteristics, outcome measures and statistical analyses, and adverse events.
In addition to our out-of-the-box databases, Synapse can:
- Connect to a range of more specialized sources through partnerships forged by Vyasa and access high quality data from organizations like Wiley.
- Tap into RSS and Twitter feeds and generate insights at high velocity, incorporating up-to-the-moment context in your analysis.
- Leverage search data to investigate term interactions with web search analytics.
Once Smart Tables are connected to sources of unstructured data, such as patent records or scientific journals, Synapse Radar allows users to start to uncover content that is semantically similar. Radar leverages deep learning to expand a list of input terms from Smart Tables to include those found within the unstructured text data that are neighbors in vector space. This provides efficient data augmentation and insight generation that are generally the result of tedious manual searching.
Generating radar terms is the first step in utilizing the power of Synapse to extract insights and value from large unstructured text databases. This is especially powerful in domains such as patent law and scientific research where unstructured text data is densely packed with domain specific language. Synapse Radar reduces this to a language that a computer can understand, a numeric vector, and finds similar concept vectors across the entire corpus of documents. Radar expedites finding the needle in the haystack, quickly drilling down on hidden information that can then be added to a Smart Table.
Using Radar is as simple selecting one or more terms in a Smart Table. The system immediately starts to work to discover what new terms can be added to the table.
Radar also uses Synapse Trendlines to visualize the frequency over time of the discovered concepts. This ensures that quickly identifying the most pertinent terms and adding them to Smart Tables is a straightforward task.
Synapse includes an Evidence feature that provides provenance and context behind terms that Radar finds.
Smart Tables need a smart search and Synapse provides just that. In addition to being able to expand terms in your spreadsheets and the content they are connected to, Synapse makes advanced querying easy with Smart Columns. Built on state-of-the-art language understanding networks, Smart Columns can answer natural language questions based on the content connected to your Synapse system. Smart Columns is an active learning system and continues to improve at answering questions as it learns your domain.
Vyasa has taken full advantage of recent breakthroughs in deep learning architecture and natural language understanding to deliver Smart Columns, a natural language query system. Once Smart Tables have been packed full of augmented data, inevitable questions will surface. Smart Columns can take any natural language question, scour the unstructured data sources connected to Smart Tables, and deliver an answer, along with the Evidence supporting the answer.
Smart Columns ships as a pre-trained system within Synapse, and can learn as it’s exposed to domain specific data. To give Smart Columns the ability to keep delivering value to users, it’s been designed as an active learning module. This means that as users flag answers as correct or incorrect, Smart Columns will routinely retrain itself and improve performance.
In addition to uncovering static data, Synapse can also detect trending terms in connected data sources to help identify whether novel terms are gaining or losing prominence relative to areas of a user’s interest and use case.
Utilize Trendlines across Synapse to provide insights into which novel terms have been trending and are most applicable to your use case.
Evidence allows users to trace the steps of Synapse AI agents as they collect information about terms in a spreadsheet. Agents document where they collected each piece of information, allowing the system to provide a bibliographic resource for everything that Synapse shows you about your spreadsheet.
Ready to dig into discovered literature, or just curious about where your Radar terms are coming from? Synapse untangles and organizes all discovered documents into a searchable corpus of Evidence.
- Drill down on documents published between a set of dates.
- Quickly assess relevance of documents. Radar terms are tagged within each document, allowing the user to quickly gauge of the value of each document to their use case.
- Improve your workflow and continue to build your Smart Tables as you read. Reading a document in Evidence and see a new and relevant term? Add it to a column in your Smart Table with a single click.
Manual research is time consuming and modern search tools are limited in scope. Synapse utilizes its understanding of semantics to drive efficient discovery and allows the focus to return to the current case.
Rare diseases are those that affect less than 200,000 people and, as can be expected, studying these elusive diseases can be difficult. Synapse allows efficient and deep exploration of massive amounts of text data that might otherwise remain hidden.
Automated identification of emergent technologies, patents, and companies from unstructured text can be challenging. Synapse can uncover even the most obscured similarities between technologies hidden deep in documents, making discovery efficient and effective.
Business intelligence focuses on the acquisition of huge quantities of data and leveraging that data into actionable insights. Synapse Radar and Trendlines enable deep, efficient research to uncover information hidden in text.